Datasets:
librarian-bots
/
model_cards_with_metadata

Dataset card Data Studio Files
xet
Community
2
modelId
stringlengths
5
139
author
stringlengths
2
42
last_modified
timestamp[us, tz=UTC]date
2020-02-15 11:33:14
2025-07-14 18:27:59
downloads
int64
0
223M
likes
int64
0
11.7k
library_name
stringclasses
520 values
tags
listlengths
1
4.05k
pipeline_tag
stringclasses
55 values
createdAt
timestamp[us, tz=UTC]date
2022-03-02 23:29:04
2025-07-14 18:27:48
card
stringlengths
11
1.01M
hfl/chinese-legal-electra-small-generator
hfl
2022-11-25T02:47:54Z
107
4
transformers
[ "transformers", "pytorch", "tf", "electra", "fill-mask", "zh", "arxiv:2004.13922", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "region:us" ]
fill-mask
2022-03-02T23:29:05Z
--- language: - zh license: "apache-2.0" --- # This model is specifically designed for legal domain. ## Chinese ELECTRA Google and Stanford University released a new pre-trained model called ELECTRA, which has a much compact model size and relatively competitive performance compared to BERT and its variants. For further accelerating the research of the Chinese pre-trained model, the Joint Laboratory of HIT and iFLYTEK Research (HFL) has released the Chinese ELECTRA models based on the official code of ELECTRA. ELECTRA-small could reach similar or even higher scores on several NLP tasks with only 1/10 parameters compared to BERT and its variants. This project is based on the official code of ELECTRA: [https://github.com/google-research/electra](https://github.com/google-research/electra) You may also interested in, - Chinese BERT series: https://github.com/ymcui/Chinese-BERT-wwm - Chinese ELECTRA: https://github.com/ymcui/Chinese-ELECTRA - Chinese XLNet: https://github.com/ymcui/Chinese-XLNet - Knowledge Distillation Toolkit - TextBrewer: https://github.com/airaria/TextBrewer More resources by HFL: https://github.com/ymcui/HFL-Anthology ## Citation If you find our resource or paper is useful, please consider including the following citation in your paper. - https://arxiv.org/abs/2004.13922 ``` @inproceedings{cui-etal-2020-revisiting, title = "Revisiting Pre-Trained Models for {C}hinese Natural Language Processing", author = "Cui, Yiming and Che, Wanxiang and Liu, Ting and Qin, Bing and Wang, Shijin and Hu, Guoping", booktitle = "Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: Findings", month = nov, year = "2020", address = "Online", publisher = "Association for Computational Linguistics", url = "https://www.aclweb.org/anthology/2020.findings-emnlp.58", pages = "657--668", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-1-parent
research-backup
2022-11-25T01:44:40Z
105
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T18:55:40Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-1-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.808968253968254 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4839572192513369 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4896142433234421 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.6264591439688716 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.748 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.36403508771929827 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.43287037037037035 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9242127467229169 - name: F1 (macro) type: f1_macro value: 0.9203856741795424 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8241784037558686 - name: F1 (macro) type: f1_macro value: 0.6193929396844311 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6505958829902492 - name: F1 (macro) type: f1_macro value: 0.6341070521744203 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9477637893858246 - name: F1 (macro) type: f1_macro value: 0.8551997164416084 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8796615481040425 - name: F1 (macro) type: f1_macro value: 0.8775906011323187 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-1-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-1-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4839572192513369 - Accuracy on SAT: 0.4896142433234421 - Accuracy on BATS: 0.6264591439688716 - Accuracy on U2: 0.36403508771929827 - Accuracy on U4: 0.43287037037037035 - Accuracy on Google: 0.748 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-1-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9242127467229169 - Micro F1 score on CogALexV: 0.8241784037558686 - Micro F1 score on EVALution: 0.6505958829902492 - Micro F1 score on K&H+N: 0.9477637893858246 - Micro F1 score on ROOT09: 0.8796615481040425 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-1-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.808968253968254 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-1-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-1-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-1-parent
research-backup
2022-11-25T01:40:14Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T18:50:42Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-1-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8154365079365079 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5347593582887701 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5341246290801187 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.7137298499166204 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.902 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4824561403508772 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.49074074074074076 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9157751996383908 - name: F1 (macro) type: f1_macro value: 0.909637226899986 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8323943661971831 - name: F1 (macro) type: f1_macro value: 0.6453015591741867 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6755146262188516 - name: F1 (macro) type: f1_macro value: 0.6664525200101824 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.963900674688739 - name: F1 (macro) type: f1_macro value: 0.8889814157651768 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.88937637104356 - name: F1 (macro) type: f1_macro value: 0.8877414161313114 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-1-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-1-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.5347593582887701 - Accuracy on SAT: 0.5341246290801187 - Accuracy on BATS: 0.7137298499166204 - Accuracy on U2: 0.4824561403508772 - Accuracy on U4: 0.49074074074074076 - Accuracy on Google: 0.902 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-1-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9157751996383908 - Micro F1 score on CogALexV: 0.8323943661971831 - Micro F1 score on EVALution: 0.6755146262188516 - Micro F1 score on K&H+N: 0.963900674688739 - Micro F1 score on ROOT09: 0.88937637104356 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-1-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8154365079365079 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-1-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-1-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-parent
research-backup
2022-11-25T01:38:38Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T18:49:09Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7388690476190476 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5026737967914439 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5014836795252225 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.6281267370761534 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.884 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5131578947368421 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.49074074074074076 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9064336296519512 - name: F1 (macro) type: f1_macro value: 0.9043602858762586 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8354460093896713 - name: F1 (macro) type: f1_macro value: 0.6469966958678447 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6830985915492958 - name: F1 (macro) type: f1_macro value: 0.6660318541629522 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9523544550323433 - name: F1 (macro) type: f1_macro value: 0.8683696140732093 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8834221247257913 - name: F1 (macro) type: f1_macro value: 0.8802195590745936 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.5026737967914439 - Accuracy on SAT: 0.5014836795252225 - Accuracy on BATS: 0.6281267370761534 - Accuracy on U2: 0.5131578947368421 - Accuracy on U4: 0.49074074074074076 - Accuracy on Google: 0.884 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9064336296519512 - Micro F1 score on CogALexV: 0.8354460093896713 - Micro F1 score on EVALution: 0.6830985915492958 - Micro F1 score on K&H+N: 0.9523544550323433 - Micro F1 score on ROOT09: 0.8834221247257913 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7388690476190476 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 10 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-2-parent
research-backup
2022-11-25T01:37:00Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T18:47:31Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-2-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.5858333333333333 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3235294117647059 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3264094955489614 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.40355753196220123 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.454 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3991228070175439 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3680555555555556 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8874491487117674 - name: F1 (macro) type: f1_macro value: 0.8773089884598599 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7772300469483568 - name: F1 (macro) type: f1_macro value: 0.49739050772253013 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.5975081256771397 - name: F1 (macro) type: f1_macro value: 0.5885262631521339 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9511024553105655 - name: F1 (macro) type: f1_macro value: 0.8619717265273807 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8570980883735506 - name: F1 (macro) type: f1_macro value: 0.8527517979840938 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-2-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-2-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.3235294117647059 - Accuracy on SAT: 0.3264094955489614 - Accuracy on BATS: 0.40355753196220123 - Accuracy on U2: 0.3991228070175439 - Accuracy on U4: 0.3680555555555556 - Accuracy on Google: 0.454 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-2-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8874491487117674 - Micro F1 score on CogALexV: 0.7772300469483568 - Micro F1 score on EVALution: 0.5975081256771397 - Micro F1 score on K&H+N: 0.9511024553105655 - Micro F1 score on ROOT09: 0.8570980883735506 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-2-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.5858333333333333 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-2-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 9 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-2-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-1-parent
research-backup
2022-11-25T01:35:37Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T18:45:38Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-1-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.919047619047619 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4117647058823529 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.41839762611275966 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.519177320733741 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.72 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3508771929824561 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4074074074074074 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9089950278740395 - name: F1 (macro) type: f1_macro value: 0.9027012284128806 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8406103286384976 - name: F1 (macro) type: f1_macro value: 0.6404768640497354 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6592632719393283 - name: F1 (macro) type: f1_macro value: 0.6489973952568304 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9628573415872574 - name: F1 (macro) type: f1_macro value: 0.8796082757849686 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8909432779692886 - name: F1 (macro) type: f1_macro value: 0.8906966577907279 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-1-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-1-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4117647058823529 - Accuracy on SAT: 0.41839762611275966 - Accuracy on BATS: 0.519177320733741 - Accuracy on U2: 0.3508771929824561 - Accuracy on U4: 0.4074074074074074 - Accuracy on Google: 0.72 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-1-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9089950278740395 - Micro F1 score on CogALexV: 0.8406103286384976 - Micro F1 score on EVALution: 0.6592632719393283 - Micro F1 score on K&H+N: 0.9628573415872574 - Micro F1 score on ROOT09: 0.8909432779692886 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-1-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.919047619047619 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-1-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-1-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-1-parent
research-backup
2022-11-25T01:31:27Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T18:40:16Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-1-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7810317460317461 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3663101604278075 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.37388724035608306 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.49861033907726515 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.586 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.36403508771929827 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.40046296296296297 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9010094922404701 - name: F1 (macro) type: f1_macro value: 0.8946867219613533 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8164319248826291 - name: F1 (macro) type: f1_macro value: 0.6083100204440299 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6240520043336945 - name: F1 (macro) type: f1_macro value: 0.6150256084891225 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9529108993531334 - name: F1 (macro) type: f1_macro value: 0.8785204370356752 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8749608273268568 - name: F1 (macro) type: f1_macro value: 0.8706590866189604 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-1-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-1-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.3663101604278075 - Accuracy on SAT: 0.37388724035608306 - Accuracy on BATS: 0.49861033907726515 - Accuracy on U2: 0.36403508771929827 - Accuracy on U4: 0.40046296296296297 - Accuracy on Google: 0.586 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-1-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9010094922404701 - Micro F1 score on CogALexV: 0.8164319248826291 - Micro F1 score on EVALution: 0.6240520043336945 - Micro F1 score on K&H+N: 0.9529108993531334 - Micro F1 score on ROOT09: 0.8749608273268568 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-1-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7810317460317461 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-1-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 7 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-1-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-0-parent
research-backup
2022-11-25T01:30:02Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T18:38:31Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-0-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8266269841269841 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4893048128342246 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.49851632047477745 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5814341300722624 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.68 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.41228070175438597 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4583333333333333 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9109537441615188 - name: F1 (macro) type: f1_macro value: 0.9058957677333925 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8342723004694835 - name: F1 (macro) type: f1_macro value: 0.6536154366052641 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6587215601300108 - name: F1 (macro) type: f1_macro value: 0.6512031246974728 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9584057870209363 - name: F1 (macro) type: f1_macro value: 0.8823600355733864 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8806016922594797 - name: F1 (macro) type: f1_macro value: 0.8732980943201075 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-0-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-0-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4893048128342246 - Accuracy on SAT: 0.49851632047477745 - Accuracy on BATS: 0.5814341300722624 - Accuracy on U2: 0.41228070175438597 - Accuracy on U4: 0.4583333333333333 - Accuracy on Google: 0.68 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-0-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9109537441615188 - Micro F1 score on CogALexV: 0.8342723004694835 - Micro F1 score on EVALution: 0.6587215601300108 - Micro F1 score on K&H+N: 0.9584057870209363 - Micro F1 score on ROOT09: 0.8806016922594797 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-0-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8266269841269841 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-0-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 10 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-0-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-a-nce-2-parent
research-backup
2022-11-25T01:28:37Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T18:36:49Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-nce-2-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8413095238095238 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5374331550802139 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5400593471810089 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.6442468037798778 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.846 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.47368421052631576 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5231481481481481 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9016121741750791 - name: F1 (macro) type: f1_macro value: 0.8969198559177928 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8497652582159625 - name: F1 (macro) type: f1_macro value: 0.6798065122891292 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6608884073672806 - name: F1 (macro) type: f1_macro value: 0.6467605391155365 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9590317868818251 - name: F1 (macro) type: f1_macro value: 0.8917303949131082 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8909432779692886 - name: F1 (macro) type: f1_macro value: 0.8913744985868259 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-nce-2-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-nce-2-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.5374331550802139 - Accuracy on SAT: 0.5400593471810089 - Accuracy on BATS: 0.6442468037798778 - Accuracy on U2: 0.47368421052631576 - Accuracy on U4: 0.5231481481481481 - Accuracy on Google: 0.846 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-nce-2-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9016121741750791 - Micro F1 score on CogALexV: 0.8497652582159625 - Micro F1 score on EVALution: 0.6608884073672806 - Micro F1 score on K&H+N: 0.9590317868818251 - Micro F1 score on ROOT09: 0.8909432779692886 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-nce-2-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8413095238095238 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-nce-2-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 10 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-nce-2-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
kbalde/distilbert-base-uncased-finetuned-imdb
kbalde
2022-11-25T01:22:32Z
161
0
transformers
[ "transformers", "pytorch", "tensorboard", "distilbert", "fill-mask", "generated_from_trainer", "dataset:imdb", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "region:us" ]
fill-mask
2022-11-25T01:13:07Z
--- license: apache-2.0 tags: - generated_from_trainer datasets: - imdb model-index: - name: distilbert-base-uncased-finetuned-imdb results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # distilbert-base-uncased-finetuned-imdb This model is a fine-tuned version of [distilbert-base-uncased](https://huggingface.co/distilbert-base-uncased) on the imdb dataset. It achieves the following results on the evaluation set: - Loss: 2.4742 ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 2e-05 - train_batch_size: 64 - eval_batch_size: 64 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 3.0 - mixed_precision_training: Native AMP ### Training results | Training Loss | Epoch | Step | Validation Loss | |:-------------:|:-----:|:----:|:---------------:| | 2.7069 | 1.0 | 157 | 2.4948 | | 2.5792 | 2.0 | 314 | 2.4235 | | 2.5259 | 3.0 | 471 | 2.4348 | ### Framework versions - Transformers 4.24.0 - Pytorch 1.12.1+cu113 - Datasets 2.7.1 - Tokenizers 0.13.2
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-parent
research-backup
2022-11-25T01:20:23Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T01:19:18Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7812301587301588 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.45989304812834225 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4629080118694362 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.565869927737632 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.744 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.41228070175438597 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4699074074074074 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8999547988549043 - name: F1 (macro) type: f1_macro value: 0.8962816660157421 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8105633802816902 - name: F1 (macro) type: f1_macro value: 0.5902758892120937 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6164680390032503 - name: F1 (macro) type: f1_macro value: 0.6103988006466425 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9581971204006399 - name: F1 (macro) type: f1_macro value: 0.8785923815023278 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8834221247257913 - name: F1 (macro) type: f1_macro value: 0.8841984389956332 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.45989304812834225 - Accuracy on SAT: 0.4629080118694362 - Accuracy on BATS: 0.565869927737632 - Accuracy on U2: 0.41228070175438597 - Accuracy on U4: 0.4699074074074074 - Accuracy on Google: 0.744 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8999547988549043 - Micro F1 score on CogALexV: 0.8105633802816902 - Micro F1 score on EVALution: 0.6164680390032503 - Micro F1 score on K&H+N: 0.9581971204006399 - Micro F1 score on ROOT09: 0.8834221247257913 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7812301587301588 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 10 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-parent
research-backup
2022-11-25T01:18:44Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T01:17:41Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7425793650793651 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.39037433155080214 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3827893175074184 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.471372984991662 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.682 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.39473684210526316 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4513888888888889 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8625885189091457 - name: F1 (macro) type: f1_macro value: 0.8579789924146027 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7875586854460094 - name: F1 (macro) type: f1_macro value: 0.5559978680629714 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.5991332611050921 - name: F1 (macro) type: f1_macro value: 0.5789568668572579 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9440077902204911 - name: F1 (macro) type: f1_macro value: 0.851850221872151 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8539642745220932 - name: F1 (macro) type: f1_macro value: 0.8572408527947681 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.39037433155080214 - Accuracy on SAT: 0.3827893175074184 - Accuracy on BATS: 0.471372984991662 - Accuracy on U2: 0.39473684210526316 - Accuracy on U4: 0.4513888888888889 - Accuracy on Google: 0.682 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8625885189091457 - Micro F1 score on CogALexV: 0.7875586854460094 - Micro F1 score on EVALution: 0.5991332611050921 - Micro F1 score on K&H+N: 0.9440077902204911 - Micro F1 score on ROOT09: 0.8539642745220932 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7425793650793651 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 10 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-parent
research-backup
2022-11-25T01:17:06Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T01:15:29Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.743095238095238 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4839572192513369 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4896142433234421 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.6375764313507504 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.862 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4868421052631579 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5046296296296297 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8952840138616845 - name: F1 (macro) type: f1_macro value: 0.8885186450216263 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.819718309859155 - name: F1 (macro) type: f1_macro value: 0.6131261712437293 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6159263271939328 - name: F1 (macro) type: f1_macro value: 0.6066052358250007 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9501982332892815 - name: F1 (macro) type: f1_macro value: 0.865577305932915 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8755875900971483 - name: F1 (macro) type: f1_macro value: 0.8756061231187074 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4839572192513369 - Accuracy on SAT: 0.4896142433234421 - Accuracy on BATS: 0.6375764313507504 - Accuracy on U2: 0.4868421052631579 - Accuracy on U4: 0.5046296296296297 - Accuracy on Google: 0.862 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8952840138616845 - Micro F1 score on CogALexV: 0.819718309859155 - Micro F1 score on EVALution: 0.6159263271939328 - Micro F1 score on K&H+N: 0.9501982332892815 - Micro F1 score on ROOT09: 0.8755875900971483 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.743095238095238 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-2-parent
research-backup
2022-11-25T01:13:13Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T01:12:12Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-2-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7630555555555556 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.47058823529411764 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.486646884272997 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5380767092829349 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.656 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4342105263157895 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4398148148148148 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8958866957962935 - name: F1 (macro) type: f1_macro value: 0.888731023349214 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8215962441314554 - name: F1 (macro) type: f1_macro value: 0.6064429926301687 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6067172264355363 - name: F1 (macro) type: f1_macro value: 0.5888667206202568 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9526326771927384 - name: F1 (macro) type: f1_macro value: 0.8662559318617946 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8827953619554998 - name: F1 (macro) type: f1_macro value: 0.8814178012740252 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-2-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-2-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.47058823529411764 - Accuracy on SAT: 0.486646884272997 - Accuracy on BATS: 0.5380767092829349 - Accuracy on U2: 0.4342105263157895 - Accuracy on U4: 0.4398148148148148 - Accuracy on Google: 0.656 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-2-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8958866957962935 - Micro F1 score on CogALexV: 0.8215962441314554 - Micro F1 score on EVALution: 0.6067172264355363 - Micro F1 score on K&H+N: 0.9526326771927384 - Micro F1 score on ROOT09: 0.8827953619554998 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-2-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7630555555555556 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-2-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 6 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-2-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-parent
research-backup
2022-11-25T01:11:28Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T01:10:23Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7165873015873016 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.44919786096256686 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4540059347181009 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.40633685380767093 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.622 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.42543859649122806 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4097222222222222 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8891065240319421 - name: F1 (macro) type: f1_macro value: 0.8801862656598641 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7943661971830986 - name: F1 (macro) type: f1_macro value: 0.5467046890024629 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.5937161430119177 - name: F1 (macro) type: f1_macro value: 0.5669593698563552 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9612575641649858 - name: F1 (macro) type: f1_macro value: 0.8804086032501298 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8586649952992792 - name: F1 (macro) type: f1_macro value: 0.8598183838649174 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.44919786096256686 - Accuracy on SAT: 0.4540059347181009 - Accuracy on BATS: 0.40633685380767093 - Accuracy on U2: 0.42543859649122806 - Accuracy on U4: 0.4097222222222222 - Accuracy on Google: 0.622 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8891065240319421 - Micro F1 score on CogALexV: 0.7943661971830986 - Micro F1 score on EVALution: 0.5937161430119177 - Micro F1 score on K&H+N: 0.9612575641649858 - Micro F1 score on ROOT09: 0.8586649952992792 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7165873015873016 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 8 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-parent
research-backup
2022-11-25T01:09:50Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T23:25:16Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7133333333333334 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4625668449197861 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4688427299703264 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5997776542523624 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.73 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4298245614035088 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.46064814814814814 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8851890914569834 - name: F1 (macro) type: f1_macro value: 0.8808316885351782 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8143192488262911 - name: F1 (macro) type: f1_macro value: 0.6038318879710047 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6273022751895991 - name: F1 (macro) type: f1_macro value: 0.6073855619416723 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9387215691729847 - name: F1 (macro) type: f1_macro value: 0.8498911005293279 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8746474459417111 - name: F1 (macro) type: f1_macro value: 0.871309204281325 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4625668449197861 - Accuracy on SAT: 0.4688427299703264 - Accuracy on BATS: 0.5997776542523624 - Accuracy on U2: 0.4298245614035088 - Accuracy on U4: 0.46064814814814814 - Accuracy on Google: 0.73 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8851890914569834 - Micro F1 score on CogALexV: 0.8143192488262911 - Micro F1 score on EVALution: 0.6273022751895991 - Micro F1 score on K&H+N: 0.9387215691729847 - Micro F1 score on ROOT09: 0.8746474459417111 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7133333333333334 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 10 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-2-parent
research-backup
2022-11-25T01:08:28Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T23:23:26Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-2-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7393650793650793 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4197860962566845 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.41839762611275966 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4886047804335742 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.602 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.41228070175438597 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.46064814814814814 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.896188036763598 - name: F1 (macro) type: f1_macro value: 0.8877200839923877 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8000000000000002 - name: F1 (macro) type: f1_macro value: 0.5589441809434722 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.5796316359696642 - name: F1 (macro) type: f1_macro value: 0.5579620480481166 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9596577867427141 - name: F1 (macro) type: f1_macro value: 0.8827843906054124 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8696333437793794 - name: F1 (macro) type: f1_macro value: 0.8695529755426573 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-2-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-2-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4197860962566845 - Accuracy on SAT: 0.41839762611275966 - Accuracy on BATS: 0.4886047804335742 - Accuracy on U2: 0.41228070175438597 - Accuracy on U4: 0.46064814814814814 - Accuracy on Google: 0.602 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-2-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.896188036763598 - Micro F1 score on CogALexV: 0.8000000000000002 - Micro F1 score on EVALution: 0.5796316359696642 - Micro F1 score on K&H+N: 0.9596577867427141 - Micro F1 score on ROOT09: 0.8696333437793794 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-2-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7393650793650793 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-2-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 10 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-2-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-parent
research-backup
2022-11-25T01:03:52Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T23:17:59Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7436309523809523 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.44385026737967914 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.44510385756676557 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.39188438021122846 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.558 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.35526315789473684 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.42824074074074076 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8865451258098539 - name: F1 (macro) type: f1_macro value: 0.8820937773390161 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7859154929577464 - name: F1 (macro) type: f1_macro value: 0.5370574811224295 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6072589382448538 - name: F1 (macro) type: f1_macro value: 0.5806581671240126 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9386520136328859 - name: F1 (macro) type: f1_macro value: 0.8351113071159506 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.862425571921028 - name: F1 (macro) type: f1_macro value: 0.8570355565302412 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.44385026737967914 - Accuracy on SAT: 0.44510385756676557 - Accuracy on BATS: 0.39188438021122846 - Accuracy on U2: 0.35526315789473684 - Accuracy on U4: 0.42824074074074076 - Accuracy on Google: 0.558 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8865451258098539 - Micro F1 score on CogALexV: 0.7859154929577464 - Micro F1 score on EVALution: 0.6072589382448538 - Micro F1 score on K&H+N: 0.9386520136328859 - Micro F1 score on ROOT09: 0.862425571921028 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7436309523809523 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 6 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-1-parent
research-backup
2022-11-25T01:02:22Z
112
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T23:16:16Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-1-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8468253968253968 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4919786096256685 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.49554896142433236 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.6203446359088383 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.794 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3991228070175439 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4675925925925926 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8946813319270754 - name: F1 (macro) type: f1_macro value: 0.8881909354487081 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8211267605633803 - name: F1 (macro) type: f1_macro value: 0.6170555844268196 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6338028169014085 - name: F1 (macro) type: f1_macro value: 0.6222691959704637 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9571537872991583 - name: F1 (macro) type: f1_macro value: 0.8677239073477255 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.878094641178314 - name: F1 (macro) type: f1_macro value: 0.8771653942410037 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-1-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-1-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4919786096256685 - Accuracy on SAT: 0.49554896142433236 - Accuracy on BATS: 0.6203446359088383 - Accuracy on U2: 0.3991228070175439 - Accuracy on U4: 0.4675925925925926 - Accuracy on Google: 0.794 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-1-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8946813319270754 - Micro F1 score on CogALexV: 0.8211267605633803 - Micro F1 score on EVALution: 0.6338028169014085 - Micro F1 score on K&H+N: 0.9571537872991583 - Micro F1 score on ROOT09: 0.878094641178314 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-1-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8468253968253968 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-1-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-1-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-2-child-prototypical
research-backup
2022-11-25T00:59:08Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T00:58:00Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-2-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8311904761904761 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.47058823529411764 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.47774480712166173 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5630906058921623 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.746 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4605263157894737 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.48148148148148145 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9193912912460449 - name: F1 (macro) type: f1_macro value: 0.9155232475170662 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8359154929577466 - name: F1 (macro) type: f1_macro value: 0.6451009241404899 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6587215601300108 - name: F1 (macro) type: f1_macro value: 0.6374469353477457 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9703693399179245 - name: F1 (macro) type: f1_macro value: 0.9060392695729131 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8846756502663742 - name: F1 (macro) type: f1_macro value: 0.8795670909805673 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-2-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-2-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.47058823529411764 - Accuracy on SAT: 0.47774480712166173 - Accuracy on BATS: 0.5630906058921623 - Accuracy on U2: 0.4605263157894737 - Accuracy on U4: 0.48148148148148145 - Accuracy on Google: 0.746 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-2-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9193912912460449 - Micro F1 score on CogALexV: 0.8359154929577466 - Micro F1 score on EVALution: 0.6587215601300108 - Micro F1 score on K&H+N: 0.9703693399179245 - Micro F1 score on ROOT09: 0.8846756502663742 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-2-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8311904761904761 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-2-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 9 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-2-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-0-child-prototypical
research-backup
2022-11-25T00:55:47Z
104
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T00:54:41Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-0-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7786111111111111 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.44385026737967914 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4391691394658754 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4952751528627015 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.71 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.39035087719298245 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4444444444444444 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9159258701220431 - name: F1 (macro) type: f1_macro value: 0.9126316000488791 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8164319248826291 - name: F1 (macro) type: f1_macro value: 0.6068522121067965 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6321776814734561 - name: F1 (macro) type: f1_macro value: 0.6239199306624651 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9479029004660221 - name: F1 (macro) type: f1_macro value: 0.8642589837552072 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8755875900971483 - name: F1 (macro) type: f1_macro value: 0.8762091857973789 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-0-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-0-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.44385026737967914 - Accuracy on SAT: 0.4391691394658754 - Accuracy on BATS: 0.4952751528627015 - Accuracy on U2: 0.39035087719298245 - Accuracy on U4: 0.4444444444444444 - Accuracy on Google: 0.71 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-0-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9159258701220431 - Micro F1 score on CogALexV: 0.8164319248826291 - Micro F1 score on EVALution: 0.6321776814734561 - Micro F1 score on K&H+N: 0.9479029004660221 - Micro F1 score on ROOT09: 0.8755875900971483 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-0-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7786111111111111 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-0-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 6 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-0-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-2-child-prototypical
research-backup
2022-11-25T00:54:10Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T00:53:04Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-2-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7883531746031746 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.553475935828877 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5459940652818991 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.725958866036687 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.9 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.49122807017543857 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5185185185185185 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9047762543317764 - name: F1 (macro) type: f1_macro value: 0.9016293933408125 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8410798122065728 - name: F1 (macro) type: f1_macro value: 0.6595900253618268 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.652762730227519 - name: F1 (macro) type: f1_macro value: 0.6495593725859232 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9506155665298741 - name: F1 (macro) type: f1_macro value: 0.8601599227538594 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8896897524287057 - name: F1 (macro) type: f1_macro value: 0.889792052107688 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-2-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-2-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.553475935828877 - Accuracy on SAT: 0.5459940652818991 - Accuracy on BATS: 0.725958866036687 - Accuracy on U2: 0.49122807017543857 - Accuracy on U4: 0.5185185185185185 - Accuracy on Google: 0.9 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-2-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9047762543317764 - Micro F1 score on CogALexV: 0.8410798122065728 - Micro F1 score on EVALution: 0.652762730227519 - Micro F1 score on K&H+N: 0.9506155665298741 - Micro F1 score on ROOT09: 0.8896897524287057 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-2-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7883531746031746 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-2-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 9 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-2-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-child-prototypical
research-backup
2022-11-25T00:50:52Z
105
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T00:49:33Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7938293650793651 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5374331550802139 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5370919881305638 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.688715953307393 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.896 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5570175438596491 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5509259259259259 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9100497212596053 - name: F1 (macro) type: f1_macro value: 0.9059343922634877 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8443661971830986 - name: F1 (macro) type: f1_macro value: 0.6626426898012254 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6728060671722643 - name: F1 (macro) type: f1_macro value: 0.6655534986615009 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9536064547541212 - name: F1 (macro) type: f1_macro value: 0.8662723101451977 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.89157004073958 - name: F1 (macro) type: f1_macro value: 0.8897680037027316 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.5374331550802139 - Accuracy on SAT: 0.5370919881305638 - Accuracy on BATS: 0.688715953307393 - Accuracy on U2: 0.5570175438596491 - Accuracy on U4: 0.5509259259259259 - Accuracy on Google: 0.896 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9100497212596053 - Micro F1 score on CogALexV: 0.8443661971830986 - Micro F1 score on EVALution: 0.6728060671722643 - Micro F1 score on K&H+N: 0.9536064547541212 - Micro F1 score on ROOT09: 0.89157004073958 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7938293650793651 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 8 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-2-child-prototypical
research-backup
2022-11-25T00:44:01Z
105
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T00:42:52Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-2-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8554365079365079 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4197860962566845 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.42136498516320475 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4535853251806559 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.666 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.40789473684210525 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.44212962962962965 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9202953141479584 - name: F1 (macro) type: f1_macro value: 0.9173587634585273 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8126760563380282 - name: F1 (macro) type: f1_macro value: 0.6010481443475104 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6365113759479957 - name: F1 (macro) type: f1_macro value: 0.6290039358223645 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.964804896710023 - name: F1 (macro) type: f1_macro value: 0.8868274767488605 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8636790974616106 - name: F1 (macro) type: f1_macro value: 0.8661746301384844 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-2-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-2-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4197860962566845 - Accuracy on SAT: 0.42136498516320475 - Accuracy on BATS: 0.4535853251806559 - Accuracy on U2: 0.40789473684210525 - Accuracy on U4: 0.44212962962962965 - Accuracy on Google: 0.666 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-2-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9202953141479584 - Micro F1 score on CogALexV: 0.8126760563380282 - Micro F1 score on EVALution: 0.6365113759479957 - Micro F1 score on K&H+N: 0.964804896710023 - Micro F1 score on ROOT09: 0.8636790974616106 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-2-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8554365079365079 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-2-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 9 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-2-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-1-child-prototypical
research-backup
2022-11-25T00:42:15Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T00:41:10Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-1-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8494444444444444 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.46524064171123 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4540059347181009 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5469705391884381 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.718 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.40789473684210525 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.47685185185185186 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9218020189844809 - name: F1 (macro) type: f1_macro value: 0.9150024413614776 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8326291079812207 - name: F1 (macro) type: f1_macro value: 0.6405467444616051 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6511375947995667 - name: F1 (macro) type: f1_macro value: 0.6342332891343501 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9626486749669612 - name: F1 (macro) type: f1_macro value: 0.8868438356719615 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8903165151989971 - name: F1 (macro) type: f1_macro value: 0.8892254981763847 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-1-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-1-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.46524064171123 - Accuracy on SAT: 0.4540059347181009 - Accuracy on BATS: 0.5469705391884381 - Accuracy on U2: 0.40789473684210525 - Accuracy on U4: 0.47685185185185186 - Accuracy on Google: 0.718 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-1-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9218020189844809 - Micro F1 score on CogALexV: 0.8326291079812207 - Micro F1 score on EVALution: 0.6511375947995667 - Micro F1 score on K&H+N: 0.9626486749669612 - Micro F1 score on ROOT09: 0.8903165151989971 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-1-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8494444444444444 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-1-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-1-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-0-child-prototypical
research-backup
2022-11-25T00:40:41Z
106
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T00:39:37Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-0-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8518253968253968 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4679144385026738 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4688427299703264 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.7204002223457476 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.85 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.39473684210526316 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.48842592592592593 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9138164833509116 - name: F1 (macro) type: f1_macro value: 0.9081465241563919 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8429577464788731 - name: F1 (macro) type: f1_macro value: 0.6602643050400394 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6500541711809318 - name: F1 (macro) type: f1_macro value: 0.6453765088165551 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9563886763580719 - name: F1 (macro) type: f1_macro value: 0.8746878565965688 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8921968035098715 - name: F1 (macro) type: f1_macro value: 0.8885486080562698 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-0-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-0-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4679144385026738 - Accuracy on SAT: 0.4688427299703264 - Accuracy on BATS: 0.7204002223457476 - Accuracy on U2: 0.39473684210526316 - Accuracy on U4: 0.48842592592592593 - Accuracy on Google: 0.85 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-0-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9138164833509116 - Micro F1 score on CogALexV: 0.8429577464788731 - Micro F1 score on EVALution: 0.6500541711809318 - Micro F1 score on K&H+N: 0.9563886763580719 - Micro F1 score on ROOT09: 0.8921968035098715 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-0-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8518253968253968 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-0-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 8 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-0-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-1-child-prototypical
research-backup
2022-11-25T00:32:28Z
104
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T00:31:24Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-1-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7409920634920635 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.45454545454545453 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.456973293768546 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5591995553085047 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.756 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4780701754385965 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5023148148148148 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9026668675606448 - name: F1 (macro) type: f1_macro value: 0.8999151651110613 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8164319248826291 - name: F1 (macro) type: f1_macro value: 0.5989549450924441 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.624593716143012 - name: F1 (macro) type: f1_macro value: 0.6078221576184047 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9597273422828129 - name: F1 (macro) type: f1_macro value: 0.8849571863295462 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8812284550297712 - name: F1 (macro) type: f1_macro value: 0.8759049665311546 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-1-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-1-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.45454545454545453 - Accuracy on SAT: 0.456973293768546 - Accuracy on BATS: 0.5591995553085047 - Accuracy on U2: 0.4780701754385965 - Accuracy on U4: 0.5023148148148148 - Accuracy on Google: 0.756 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-1-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9026668675606448 - Micro F1 score on CogALexV: 0.8164319248826291 - Micro F1 score on EVALution: 0.624593716143012 - Micro F1 score on K&H+N: 0.9597273422828129 - Micro F1 score on ROOT09: 0.8812284550297712 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-1-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7409920634920635 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-1-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 9 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-1-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-child-prototypical
research-backup
2022-11-25T00:30:48Z
104
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T00:29:42Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7553174603174603 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3877005347593583 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3887240356083086 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.6158977209560867 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.622 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.38596491228070173 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.39814814814814814 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8958866957962935 - name: F1 (macro) type: f1_macro value: 0.8899187556569595 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7953051643192488 - name: F1 (macro) type: f1_macro value: 0.5661036014906222 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.5839653304442037 - name: F1 (macro) type: f1_macro value: 0.5694962586836211 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9565277874382695 - name: F1 (macro) type: f1_macro value: 0.8719904519356941 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8718270134753996 - name: F1 (macro) type: f1_macro value: 0.8712363187552654 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.3877005347593583 - Accuracy on SAT: 0.3887240356083086 - Accuracy on BATS: 0.6158977209560867 - Accuracy on U2: 0.38596491228070173 - Accuracy on U4: 0.39814814814814814 - Accuracy on Google: 0.622 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8958866957962935 - Micro F1 score on CogALexV: 0.7953051643192488 - Micro F1 score on EVALution: 0.5839653304442037 - Micro F1 score on K&H+N: 0.9565277874382695 - Micro F1 score on ROOT09: 0.8718270134753996 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7553174603174603 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 3 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
kbalde/bert-finetuned-ner
kbalde
2022-11-25T00:29:59Z
121
0
transformers
[ "transformers", "pytorch", "tensorboard", "bert", "token-classification", "generated_from_trainer", "dataset:conll2003", "license:apache-2.0", "model-index", "autotrain_compatible", "endpoints_compatible", "region:us" ]
token-classification
2022-11-25T00:12:29Z
--- license: apache-2.0 tags: - generated_from_trainer datasets: - conll2003 metrics: - precision - recall - f1 - accuracy model-index: - name: bert-finetuned-ner results: - task: name: Token Classification type: token-classification dataset: name: conll2003 type: conll2003 config: conll2003 split: train args: conll2003 metrics: - name: Precision type: precision value: 0.9281121187139324 - name: Recall type: recall value: 0.9473241332884551 - name: F1 type: f1 value: 0.9376197218289332 - name: Accuracy type: accuracy value: 0.9862689115205746 --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # bert-finetuned-ner This model is a fine-tuned version of [bert-base-cased](https://huggingface.co/bert-base-cased) on the conll2003 dataset. It achieves the following results on the evaluation set: - Loss: 0.0610 - Precision: 0.9281 - Recall: 0.9473 - F1: 0.9376 - Accuracy: 0.9863 ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 2e-05 - train_batch_size: 8 - eval_batch_size: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 3 ### Training results | Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:---------:|:------:|:------:|:--------:| | 0.0859 | 1.0 | 1756 | 0.0663 | 0.9088 | 0.9312 | 0.9199 | 0.9822 | | 0.0331 | 2.0 | 3512 | 0.0622 | 0.9270 | 0.9461 | 0.9365 | 0.9856 | | 0.016 | 3.0 | 5268 | 0.0610 | 0.9281 | 0.9473 | 0.9376 | 0.9863 | ### Framework versions - Transformers 4.24.0 - Pytorch 1.12.1+cu113 - Datasets 2.7.1 - Tokenizers 0.13.2
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-child-prototypical
research-backup
2022-11-25T00:29:01Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T00:27:54Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.763452380952381 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4358288770053476 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.44510385756676557 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.6453585325180656 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.764 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.41228070175438597 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4305555555555556 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.893174627090553 - name: F1 (macro) type: f1_macro value: 0.88253444204685 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7779342723004695 - name: F1 (macro) type: f1_macro value: 0.5467947875271857 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6013001083423619 - name: F1 (macro) type: f1_macro value: 0.5766482049475778 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9545802323155039 - name: F1 (macro) type: f1_macro value: 0.8630198238087204 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8608586649952994 - name: F1 (macro) type: f1_macro value: 0.8593411038565139 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4358288770053476 - Accuracy on SAT: 0.44510385756676557 - Accuracy on BATS: 0.6453585325180656 - Accuracy on U2: 0.41228070175438597 - Accuracy on U4: 0.4305555555555556 - Accuracy on Google: 0.764 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.893174627090553 - Micro F1 score on CogALexV: 0.7779342723004695 - Micro F1 score on EVALution: 0.6013001083423619 - Micro F1 score on K&H+N: 0.9545802323155039 - Micro F1 score on ROOT09: 0.8608586649952994 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.763452380952381 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 9 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-child-prototypical
research-backup
2022-11-25T00:27:18Z
104
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T00:26:15Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.743095238095238 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4839572192513369 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4896142433234421 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.6375764313507504 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.862 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4868421052631579 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5046296296296297 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8952840138616845 - name: F1 (macro) type: f1_macro value: 0.8885186450216263 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.819718309859155 - name: F1 (macro) type: f1_macro value: 0.6131261712437293 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6159263271939328 - name: F1 (macro) type: f1_macro value: 0.6066052358250007 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9501982332892815 - name: F1 (macro) type: f1_macro value: 0.865577305932915 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8755875900971483 - name: F1 (macro) type: f1_macro value: 0.8756061231187074 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4839572192513369 - Accuracy on SAT: 0.4896142433234421 - Accuracy on BATS: 0.6375764313507504 - Accuracy on U2: 0.4868421052631579 - Accuracy on U4: 0.5046296296296297 - Accuracy on Google: 0.862 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8952840138616845 - Micro F1 score on CogALexV: 0.819718309859155 - Micro F1 score on EVALution: 0.6159263271939328 - Micro F1 score on K&H+N: 0.9501982332892815 - Micro F1 score on ROOT09: 0.8755875900971483 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.743095238095238 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-0-child-prototypical
research-backup
2022-11-25T00:25:41Z
104
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T00:24:38Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-0-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7365079365079366 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.39037433155080214 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.39762611275964393 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.6347971095052807 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.734 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.39035087719298245 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4305555555555556 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8924212746722917 - name: F1 (macro) type: f1_macro value: 0.8846814087913352 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7990610328638498 - name: F1 (macro) type: f1_macro value: 0.5627441768083108 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.5958829902491874 - name: F1 (macro) type: f1_macro value: 0.5813557004000751 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9576406760798497 - name: F1 (macro) type: f1_macro value: 0.8696037881278025 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8636790974616106 - name: F1 (macro) type: f1_macro value: 0.865242587505383 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-0-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-0-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.39037433155080214 - Accuracy on SAT: 0.39762611275964393 - Accuracy on BATS: 0.6347971095052807 - Accuracy on U2: 0.39035087719298245 - Accuracy on U4: 0.4305555555555556 - Accuracy on Google: 0.734 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-0-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8924212746722917 - Micro F1 score on CogALexV: 0.7990610328638498 - Micro F1 score on EVALution: 0.5958829902491874 - Micro F1 score on K&H+N: 0.9576406760798497 - Micro F1 score on ROOT09: 0.8636790974616106 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-0-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7365079365079366 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-0-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 6 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-0-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-child-prototypical
research-backup
2022-11-25T00:22:28Z
104
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T00:21:20Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8001190476190476 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4385026737967914 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4362017804154303 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5786548082267927 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.802 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.41228070175438597 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4074074074074074 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9026668675606448 - name: F1 (macro) type: f1_macro value: 0.894134460103921 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8044600938967136 - name: F1 (macro) type: f1_macro value: 0.5658272058955415 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6343445287107259 - name: F1 (macro) type: f1_macro value: 0.6053012278166984 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9589622313417264 - name: F1 (macro) type: f1_macro value: 0.8820483966915638 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8824819805703541 - name: F1 (macro) type: f1_macro value: 0.8778053649821325 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4385026737967914 - Accuracy on SAT: 0.4362017804154303 - Accuracy on BATS: 0.5786548082267927 - Accuracy on U2: 0.41228070175438597 - Accuracy on U4: 0.4074074074074074 - Accuracy on Google: 0.802 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9026668675606448 - Micro F1 score on CogALexV: 0.8044600938967136 - Micro F1 score on EVALution: 0.6343445287107259 - Micro F1 score on K&H+N: 0.9589622313417264 - Micro F1 score on ROOT09: 0.8824819805703541 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8001190476190476 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-child-prototypical
research-backup
2022-11-25T00:20:47Z
104
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T00:19:38Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.796984126984127 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4037433155080214 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3916913946587537 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.6859366314619233 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.784 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.42105263157894735 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4583333333333333 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8922706041886395 - name: F1 (macro) type: f1_macro value: 0.888351620007291 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8225352112676056 - name: F1 (macro) type: f1_macro value: 0.6002271079420197 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6262188515709642 - name: F1 (macro) type: f1_macro value: 0.6065779072189218 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9565277874382695 - name: F1 (macro) type: f1_macro value: 0.8783769705919625 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8774678784080225 - name: F1 (macro) type: f1_macro value: 0.8773207792691867 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4037433155080214 - Accuracy on SAT: 0.3916913946587537 - Accuracy on BATS: 0.6859366314619233 - Accuracy on U2: 0.42105263157894735 - Accuracy on U4: 0.4583333333333333 - Accuracy on Google: 0.784 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8922706041886395 - Micro F1 score on CogALexV: 0.8225352112676056 - Micro F1 score on EVALution: 0.6262188515709642 - Micro F1 score on K&H+N: 0.9565277874382695 - Micro F1 score on ROOT09: 0.8774678784080225 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.796984126984127 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 6 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-2-child-prototypical
research-backup
2022-11-25T00:19:05Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T00:18:00Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-2-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7526984126984126 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.39037433155080214 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.40059347181008903 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.48526959421901056 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.544 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.44298245614035087 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4675925925925926 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8857917733915925 - name: F1 (macro) type: f1_macro value: 0.8738206104201592 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7866197183098591 - name: F1 (macro) type: f1_macro value: 0.5206895111405141 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.5866738894907909 - name: F1 (macro) type: f1_macro value: 0.5627937428119613 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9581971204006399 - name: F1 (macro) type: f1_macro value: 0.87313513119675 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8574114697586963 - name: F1 (macro) type: f1_macro value: 0.8562330887103827 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-2-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-2-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.39037433155080214 - Accuracy on SAT: 0.40059347181008903 - Accuracy on BATS: 0.48526959421901056 - Accuracy on U2: 0.44298245614035087 - Accuracy on U4: 0.4675925925925926 - Accuracy on Google: 0.544 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-2-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8857917733915925 - Micro F1 score on CogALexV: 0.7866197183098591 - Micro F1 score on EVALution: 0.5866738894907909 - Micro F1 score on K&H+N: 0.9581971204006399 - Micro F1 score on ROOT09: 0.8574114697586963 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-2-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7526984126984126 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-2-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 9 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-2-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-1-child-prototypical
research-backup
2022-11-25T00:17:26Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T00:16:19Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-1-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8926984126984127 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4572192513368984 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4599406528189911 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5369649805447471 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.748 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4298245614035088 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4375 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.897393400632816 - name: F1 (macro) type: f1_macro value: 0.8928414691940428 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7889671361502347 - name: F1 (macro) type: f1_macro value: 0.5428433920480771 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6094257854821236 - name: F1 (macro) type: f1_macro value: 0.6019520606116872 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9575015649996522 - name: F1 (macro) type: f1_macro value: 0.8689426299491707 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8859291758069571 - name: F1 (macro) type: f1_macro value: 0.8862133684595959 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-1-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-1-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4572192513368984 - Accuracy on SAT: 0.4599406528189911 - Accuracy on BATS: 0.5369649805447471 - Accuracy on U2: 0.4298245614035088 - Accuracy on U4: 0.4375 - Accuracy on Google: 0.748 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-1-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.897393400632816 - Micro F1 score on CogALexV: 0.7889671361502347 - Micro F1 score on EVALution: 0.6094257854821236 - Micro F1 score on K&H+N: 0.9575015649996522 - Micro F1 score on ROOT09: 0.8859291758069571 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-1-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8926984126984127 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-1-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 9 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-b-nce-1-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-child-prototypical
research-backup
2022-11-25T00:14:13Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-25T00:13:00Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8508333333333333 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4304812834224599 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.42729970326409494 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.44580322401334077 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.63 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3684210526315789 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4375 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8945306614434232 - name: F1 (macro) type: f1_macro value: 0.8881298132895892 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8164319248826291 - name: F1 (macro) type: f1_macro value: 0.5907123053940893 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6229685807150596 - name: F1 (macro) type: f1_macro value: 0.6054874185412089 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9565973429783682 - name: F1 (macro) type: f1_macro value: 0.869940503406368 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8730805390159825 - name: F1 (macro) type: f1_macro value: 0.8691781735538814 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4304812834224599 - Accuracy on SAT: 0.42729970326409494 - Accuracy on BATS: 0.44580322401334077 - Accuracy on U2: 0.3684210526315789 - Accuracy on U4: 0.4375 - Accuracy on Google: 0.63 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8945306614434232 - Micro F1 score on CogALexV: 0.8164319248826291 - Micro F1 score on EVALution: 0.6229685807150596 - Micro F1 score on K&H+N: 0.9565973429783682 - Micro F1 score on ROOT09: 0.8730805390159825 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8508333333333333 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-0-child-prototypical
research-backup
2022-11-25T00:11:00Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T10:45:33Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-0-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8746031746031746 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.46524064171123 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.456973293768546 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.7176209005002779 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.862 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4692982456140351 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4976851851851852 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9046255838481242 - name: F1 (macro) type: f1_macro value: 0.8964159361404608 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8234741784037559 - name: F1 (macro) type: f1_macro value: 0.619845576273754 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.647887323943662 - name: F1 (macro) type: f1_macro value: 0.6292175310531701 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9543020101551088 - name: F1 (macro) type: f1_macro value: 0.868505670456119 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8755875900971483 - name: F1 (macro) type: f1_macro value: 0.8655161202104406 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-0-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-0-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.46524064171123 - Accuracy on SAT: 0.456973293768546 - Accuracy on BATS: 0.7176209005002779 - Accuracy on U2: 0.4692982456140351 - Accuracy on U4: 0.4976851851851852 - Accuracy on Google: 0.862 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-0-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9046255838481242 - Micro F1 score on CogALexV: 0.8234741784037559 - Micro F1 score on EVALution: 0.647887323943662 - Micro F1 score on K&H+N: 0.9543020101551088 - Micro F1 score on ROOT09: 0.8755875900971483 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-0-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8746031746031746 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-0-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 6 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-0-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-2-child
research-backup
2022-11-25T00:09:32Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T10:43:37Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-2-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8311904761904761 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.47058823529411764 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.47774480712166173 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5630906058921623 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.746 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4605263157894737 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.48148148148148145 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9193912912460449 - name: F1 (macro) type: f1_macro value: 0.9155232475170662 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8359154929577466 - name: F1 (macro) type: f1_macro value: 0.6451009241404899 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6587215601300108 - name: F1 (macro) type: f1_macro value: 0.6374469353477457 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9703693399179245 - name: F1 (macro) type: f1_macro value: 0.9060392695729131 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8846756502663742 - name: F1 (macro) type: f1_macro value: 0.8795670909805673 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-2-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-2-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.47058823529411764 - Accuracy on SAT: 0.47774480712166173 - Accuracy on BATS: 0.5630906058921623 - Accuracy on U2: 0.4605263157894737 - Accuracy on U4: 0.48148148148148145 - Accuracy on Google: 0.746 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-2-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9193912912460449 - Micro F1 score on CogALexV: 0.8359154929577466 - Micro F1 score on EVALution: 0.6587215601300108 - Micro F1 score on K&H+N: 0.9703693399179245 - Micro F1 score on ROOT09: 0.8846756502663742 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-2-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8311904761904761 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-2-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 9 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-2-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-1-child
research-backup
2022-11-25T00:07:51Z
105
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T10:41:46Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-1-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.808968253968254 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4839572192513369 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4896142433234421 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.6264591439688716 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.748 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.36403508771929827 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.43287037037037035 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9242127467229169 - name: F1 (macro) type: f1_macro value: 0.9203856741795424 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8241784037558686 - name: F1 (macro) type: f1_macro value: 0.6193929396844311 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6505958829902492 - name: F1 (macro) type: f1_macro value: 0.6341070521744203 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9477637893858246 - name: F1 (macro) type: f1_macro value: 0.8551997164416084 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8796615481040425 - name: F1 (macro) type: f1_macro value: 0.8775906011323187 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-1-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-1-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4839572192513369 - Accuracy on SAT: 0.4896142433234421 - Accuracy on BATS: 0.6264591439688716 - Accuracy on U2: 0.36403508771929827 - Accuracy on U4: 0.43287037037037035 - Accuracy on Google: 0.748 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-1-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9242127467229169 - Micro F1 score on CogALexV: 0.8241784037558686 - Micro F1 score on EVALution: 0.6505958829902492 - Micro F1 score on K&H+N: 0.9477637893858246 - Micro F1 score on ROOT09: 0.8796615481040425 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-1-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.808968253968254 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-1-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-nce-1-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-2-child
research-backup
2022-11-25T00:04:56Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T10:38:05Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-2-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7883531746031746 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.553475935828877 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5459940652818991 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.725958866036687 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.9 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.49122807017543857 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5185185185185185 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9047762543317764 - name: F1 (macro) type: f1_macro value: 0.9016293933408125 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8410798122065728 - name: F1 (macro) type: f1_macro value: 0.6595900253618268 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.652762730227519 - name: F1 (macro) type: f1_macro value: 0.6495593725859232 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9506155665298741 - name: F1 (macro) type: f1_macro value: 0.8601599227538594 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8896897524287057 - name: F1 (macro) type: f1_macro value: 0.889792052107688 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-2-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-2-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.553475935828877 - Accuracy on SAT: 0.5459940652818991 - Accuracy on BATS: 0.725958866036687 - Accuracy on U2: 0.49122807017543857 - Accuracy on U4: 0.5185185185185185 - Accuracy on Google: 0.9 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-2-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9047762543317764 - Micro F1 score on CogALexV: 0.8410798122065728 - Micro F1 score on EVALution: 0.652762730227519 - Micro F1 score on K&H+N: 0.9506155665298741 - Micro F1 score on ROOT09: 0.8896897524287057 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-2-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7883531746031746 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-2-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 9 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-2-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-1-child
research-backup
2022-11-25T00:03:32Z
105
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T10:36:17Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-1-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8154365079365079 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5347593582887701 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5341246290801187 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.7137298499166204 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.902 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4824561403508772 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.49074074074074076 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9157751996383908 - name: F1 (macro) type: f1_macro value: 0.909637226899986 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8323943661971831 - name: F1 (macro) type: f1_macro value: 0.6453015591741867 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6755146262188516 - name: F1 (macro) type: f1_macro value: 0.6664525200101824 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.963900674688739 - name: F1 (macro) type: f1_macro value: 0.8889814157651768 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.88937637104356 - name: F1 (macro) type: f1_macro value: 0.8877414161313114 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-1-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-1-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.5347593582887701 - Accuracy on SAT: 0.5341246290801187 - Accuracy on BATS: 0.7137298499166204 - Accuracy on U2: 0.4824561403508772 - Accuracy on U4: 0.49074074074074076 - Accuracy on Google: 0.902 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-1-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9157751996383908 - Micro F1 score on CogALexV: 0.8323943661971831 - Micro F1 score on EVALution: 0.6755146262188516 - Micro F1 score on K&H+N: 0.963900674688739 - Micro F1 score on ROOT09: 0.88937637104356 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-1-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8154365079365079 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-1-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-1-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-child
research-backup
2022-11-25T00:02:09Z
97
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T10:34:27Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7938293650793651 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5374331550802139 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5370919881305638 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.688715953307393 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.896 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5570175438596491 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5509259259259259 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9100497212596053 - name: F1 (macro) type: f1_macro value: 0.9059343922634877 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8443661971830986 - name: F1 (macro) type: f1_macro value: 0.6626426898012254 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6728060671722643 - name: F1 (macro) type: f1_macro value: 0.6655534986615009 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9536064547541212 - name: F1 (macro) type: f1_macro value: 0.8662723101451977 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.89157004073958 - name: F1 (macro) type: f1_macro value: 0.8897680037027316 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.5374331550802139 - Accuracy on SAT: 0.5370919881305638 - Accuracy on BATS: 0.688715953307393 - Accuracy on U2: 0.5570175438596491 - Accuracy on U4: 0.5509259259259259 - Accuracy on Google: 0.896 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9100497212596053 - Micro F1 score on CogALexV: 0.8443661971830986 - Micro F1 score on EVALution: 0.6728060671722643 - Micro F1 score on K&H+N: 0.9536064547541212 - Micro F1 score on ROOT09: 0.89157004073958 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7938293650793651 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 8 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-nce-0-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-2-child
research-backup
2022-11-25T00:00:51Z
100
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T10:32:34Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-2-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.6354563492063492 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.31016042780748665 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.32344213649851633 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3085047248471373 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.328 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3157894736842105 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.2986111111111111 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9019135151423836 - name: F1 (macro) type: f1_macro value: 0.8910141280520246 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7758215962441315 - name: F1 (macro) type: f1_macro value: 0.5011043573121808 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.5677139761646804 - name: F1 (macro) type: f1_macro value: 0.5566709002001424 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9569451206788621 - name: F1 (macro) type: f1_macro value: 0.8723408257004264 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8621121905358822 - name: F1 (macro) type: f1_macro value: 0.8575846242198137 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-2-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-2-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.31016042780748665 - Accuracy on SAT: 0.32344213649851633 - Accuracy on BATS: 0.3085047248471373 - Accuracy on U2: 0.3157894736842105 - Accuracy on U4: 0.2986111111111111 - Accuracy on Google: 0.328 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-2-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9019135151423836 - Micro F1 score on CogALexV: 0.7758215962441315 - Micro F1 score on EVALution: 0.5677139761646804 - Micro F1 score on K&H+N: 0.9569451206788621 - Micro F1 score on ROOT09: 0.8621121905358822 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-2-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.6354563492063492 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-2-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-2-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-0-child
research-backup
2022-11-24T23:57:59Z
99
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T10:29:03Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-0-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7946825396825397 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.47593582887700536 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.47774480712166173 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.7026125625347415 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.766 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.41228070175438597 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.47453703703703703 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9061322886846467 - name: F1 (macro) type: f1_macro value: 0.9005307658810957 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8345070422535211 - name: F1 (macro) type: f1_macro value: 0.6390904639654723 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6570964247020585 - name: F1 (macro) type: f1_macro value: 0.6350934679864825 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9510328997704667 - name: F1 (macro) type: f1_macro value: 0.8702203249298888 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8953306173613286 - name: F1 (macro) type: f1_macro value: 0.8939575406867228 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-0-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-0-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.47593582887700536 - Accuracy on SAT: 0.47774480712166173 - Accuracy on BATS: 0.7026125625347415 - Accuracy on U2: 0.41228070175438597 - Accuracy on U4: 0.47453703703703703 - Accuracy on Google: 0.766 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-0-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9061322886846467 - Micro F1 score on CogALexV: 0.8345070422535211 - Micro F1 score on EVALution: 0.6570964247020585 - Micro F1 score on K&H+N: 0.9510328997704667 - Micro F1 score on ROOT09: 0.8953306173613286 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-0-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7946825396825397 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-0-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 8 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-nce-0-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-2-child
research-backup
2022-11-24T23:56:31Z
105
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T10:27:17Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-2-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7919444444444445 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4572192513368984 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4599406528189911 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.6303501945525292 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.748 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.40350877192982454 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4375 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9174325749585657 - name: F1 (macro) type: f1_macro value: 0.9119203122700617 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8279342723004695 - name: F1 (macro) type: f1_macro value: 0.6322245634658208 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.661430119176598 - name: F1 (macro) type: f1_macro value: 0.6452818980906698 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9629964526674549 - name: F1 (macro) type: f1_macro value: 0.8855189948338477 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8918834221247258 - name: F1 (macro) type: f1_macro value: 0.8888732298632717 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-2-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-2-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4572192513368984 - Accuracy on SAT: 0.4599406528189911 - Accuracy on BATS: 0.6303501945525292 - Accuracy on U2: 0.40350877192982454 - Accuracy on U4: 0.4375 - Accuracy on Google: 0.748 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-2-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9174325749585657 - Micro F1 score on CogALexV: 0.8279342723004695 - Micro F1 score on EVALution: 0.661430119176598 - Micro F1 score on K&H+N: 0.9629964526674549 - Micro F1 score on ROOT09: 0.8918834221247258 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-2-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7919444444444445 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-2-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 6 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-nce-2-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
gustavecortal/distilcamembert-cae-behavior
gustavecortal
2022-11-24T23:56:14Z
159
0
transformers
[ "transformers", "pytorch", "camembert", "text-classification", "generated_from_trainer", "license:mit", "autotrain_compatible", "endpoints_compatible", "region:us" ]
text-classification
2022-11-24T23:52:18Z
--- license: mit tags: - generated_from_trainer metrics: - precision - recall - f1 model-index: - name: distilcamembert-cae-behavior results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # distilcamembert-cae-behavior This model is a fine-tuned version of [cmarkea/distilcamembert-base](https://huggingface.co/cmarkea/distilcamembert-base) on an unknown dataset. It achieves the following results on the evaluation set: - Loss: 0.9041 - Precision: 0.6838 - Recall: 0.6709 - F1: 0.6656 ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 5e-05 - train_batch_size: 8 - eval_batch_size: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_ratio: 0.1 - num_epochs: 5.0 ### Training results | Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | |:-------------:|:-----:|:----:|:---------------:|:---------:|:------:|:------:| | 1.1837 | 1.0 | 40 | 1.0310 | 0.1956 | 0.4304 | 0.2690 | | 1.0456 | 2.0 | 80 | 1.0155 | 0.5603 | 0.5570 | 0.5047 | | 0.8225 | 3.0 | 120 | 0.9677 | 0.7190 | 0.6582 | 0.6522 | | 0.5927 | 4.0 | 160 | 0.9041 | 0.6838 | 0.6709 | 0.6656 | | 0.4341 | 5.0 | 200 | 0.8963 | 0.6608 | 0.6456 | 0.6442 | ### Framework versions - Transformers 4.24.0 - Pytorch 1.12.1+cu113 - Datasets 2.7.1 - Tokenizers 0.13.2
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-a-nce-1-child
research-backup
2022-11-24T23:50:58Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T10:20:02Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-nce-1-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7961111111111111 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5668449197860963 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5637982195845698 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.7087270705947749 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.862 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5185185185185185 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9094470393249963 - name: F1 (macro) type: f1_macro value: 0.9071978262208379 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8481220657276995 - name: F1 (macro) type: f1_macro value: 0.6770534875484192 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6663055254604551 - name: F1 (macro) type: f1_macro value: 0.6477804074680519 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9549975655560965 - name: F1 (macro) type: f1_macro value: 0.8718545305555775 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8994045753682232 - name: F1 (macro) type: f1_macro value: 0.8985019443558045 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-nce-1-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-nce-1-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.5668449197860963 - Accuracy on SAT: 0.5637982195845698 - Accuracy on BATS: 0.7087270705947749 - Accuracy on U2: 0.5 - Accuracy on U4: 0.5185185185185185 - Accuracy on Google: 0.862 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-nce-1-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9094470393249963 - Micro F1 score on CogALexV: 0.8481220657276995 - Micro F1 score on EVALution: 0.6663055254604551 - Micro F1 score on K&H+N: 0.9549975655560965 - Micro F1 score on ROOT09: 0.8994045753682232 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-nce-1-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7961111111111111 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-nce-1-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-nce-1-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-2-child
research-backup
2022-11-24T23:48:13Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T10:16:30Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-2-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.6670436507936508 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3770053475935829 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.37388724035608306 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4802668148971651 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.558 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.33771929824561403 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.34953703703703703 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8898598764502034 - name: F1 (macro) type: f1_macro value: 0.881086615114024 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7859154929577464 - name: F1 (macro) type: f1_macro value: 0.5266767261561511 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.566088840736728 - name: F1 (macro) type: f1_macro value: 0.5668961837542391 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9492940112679975 - name: F1 (macro) type: f1_macro value: 0.8577064786656643 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8480100282043247 - name: F1 (macro) type: f1_macro value: 0.8474938401923806 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-2-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-2-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.3770053475935829 - Accuracy on SAT: 0.37388724035608306 - Accuracy on BATS: 0.4802668148971651 - Accuracy on U2: 0.33771929824561403 - Accuracy on U4: 0.34953703703703703 - Accuracy on Google: 0.558 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-2-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8898598764502034 - Micro F1 score on CogALexV: 0.7859154929577464 - Micro F1 score on EVALution: 0.566088840736728 - Micro F1 score on K&H+N: 0.9492940112679975 - Micro F1 score on ROOT09: 0.8480100282043247 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-2-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.6670436507936508 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-2-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-2-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-child
research-backup
2022-11-24T23:45:19Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T10:12:52Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8476190476190476 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.42245989304812837 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.42136498516320475 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5225125069483046 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.776 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.45614035087719296 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5092592592592593 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.912912460448998 - name: F1 (macro) type: f1_macro value: 0.9056724482489867 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7812206572769953 - name: F1 (macro) type: f1_macro value: 0.5421400193351218 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6273022751895991 - name: F1 (macro) type: f1_macro value: 0.6062843287217384 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9551366766362941 - name: F1 (macro) type: f1_macro value: 0.877937874570795 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8658727671576308 - name: F1 (macro) type: f1_macro value: 0.8610411198198672 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.42245989304812837 - Accuracy on SAT: 0.42136498516320475 - Accuracy on BATS: 0.5225125069483046 - Accuracy on U2: 0.45614035087719296 - Accuracy on U4: 0.5092592592592593 - Accuracy on Google: 0.776 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.912912460448998 - Micro F1 score on CogALexV: 0.7812206572769953 - Micro F1 score on EVALution: 0.6273022751895991 - Micro F1 score on K&H+N: 0.9551366766362941 - Micro F1 score on ROOT09: 0.8658727671576308 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8476190476190476 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 8 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-e-nce-0-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-child
research-backup
2022-11-24T23:43:55Z
104
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T10:11:09Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7805555555555556 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4090909090909091 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.41839762611275966 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4252362423568649 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.61 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.39035087719298245 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.42592592592592593 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8918185927376827 - name: F1 (macro) type: f1_macro value: 0.8809778018098476 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7938967136150236 - name: F1 (macro) type: f1_macro value: 0.5699679882126063 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.600216684723727 - name: F1 (macro) type: f1_macro value: 0.5876126038869113 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9533282325937261 - name: F1 (macro) type: f1_macro value: 0.8758661866825506 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8671262926982137 - name: F1 (macro) type: f1_macro value: 0.8601713638970966 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4090909090909091 - Accuracy on SAT: 0.41839762611275966 - Accuracy on BATS: 0.4252362423568649 - Accuracy on U2: 0.39035087719298245 - Accuracy on U4: 0.42592592592592593 - Accuracy on Google: 0.61 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8918185927376827 - Micro F1 score on CogALexV: 0.7938967136150236 - Micro F1 score on EVALution: 0.600216684723727 - Micro F1 score on K&H+N: 0.9533282325937261 - Micro F1 score on ROOT09: 0.8671262926982137 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7805555555555556 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 3 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-2-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-child
research-backup
2022-11-24T23:42:30Z
97
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T10:09:25Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.743095238095238 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4839572192513369 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4896142433234421 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.6375764313507504 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.862 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4868421052631579 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5046296296296297 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8952840138616845 - name: F1 (macro) type: f1_macro value: 0.8885186450216263 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.819718309859155 - name: F1 (macro) type: f1_macro value: 0.6131261712437293 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6159263271939328 - name: F1 (macro) type: f1_macro value: 0.6066052358250007 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9501982332892815 - name: F1 (macro) type: f1_macro value: 0.865577305932915 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8755875900971483 - name: F1 (macro) type: f1_macro value: 0.8756061231187074 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4839572192513369 - Accuracy on SAT: 0.4896142433234421 - Accuracy on BATS: 0.6375764313507504 - Accuracy on U2: 0.4868421052631579 - Accuracy on U4: 0.5046296296296297 - Accuracy on Google: 0.862 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8952840138616845 - Micro F1 score on CogALexV: 0.819718309859155 - Micro F1 score on EVALution: 0.6159263271939328 - Micro F1 score on K&H+N: 0.9501982332892815 - Micro F1 score on ROOT09: 0.8755875900971483 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.743095238095238 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-1-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-0-child
research-backup
2022-11-24T23:40:58Z
104
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T10:07:32Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-0-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8158134920634921 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.42245989304812837 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.41543026706231456 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.6837131739855475 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.822 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.40350877192982454 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4652777777777778 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8891065240319421 - name: F1 (macro) type: f1_macro value: 0.8892784116442755 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8258215962441313 - name: F1 (macro) type: f1_macro value: 0.6242911406488363 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6224268689057422 - name: F1 (macro) type: f1_macro value: 0.6096926993293706 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9524935661125409 - name: F1 (macro) type: f1_macro value: 0.8748503382645149 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8856157944218114 - name: F1 (macro) type: f1_macro value: 0.8852545460393407 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-0-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-0-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.42245989304812837 - Accuracy on SAT: 0.41543026706231456 - Accuracy on BATS: 0.6837131739855475 - Accuracy on U2: 0.40350877192982454 - Accuracy on U4: 0.4652777777777778 - Accuracy on Google: 0.822 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-0-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8891065240319421 - Micro F1 score on CogALexV: 0.8258215962441313 - Micro F1 score on EVALution: 0.6224268689057422 - Micro F1 score on K&H+N: 0.9524935661125409 - Micro F1 score on ROOT09: 0.8856157944218114 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-0-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8158134920634921 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-0-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 8 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-d-nce-0-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-2-child
research-backup
2022-11-24T23:39:29Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T10:04:12Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-2-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7630555555555556 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.47058823529411764 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.486646884272997 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5380767092829349 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.656 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4342105263157895 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4398148148148148 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8958866957962935 - name: F1 (macro) type: f1_macro value: 0.888731023349214 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8215962441314554 - name: F1 (macro) type: f1_macro value: 0.6064429926301687 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6067172264355363 - name: F1 (macro) type: f1_macro value: 0.5888667206202568 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9526326771927384 - name: F1 (macro) type: f1_macro value: 0.8662559318617946 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8827953619554998 - name: F1 (macro) type: f1_macro value: 0.8814178012740252 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-2-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-2-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.47058823529411764 - Accuracy on SAT: 0.486646884272997 - Accuracy on BATS: 0.5380767092829349 - Accuracy on U2: 0.4342105263157895 - Accuracy on U4: 0.4398148148148148 - Accuracy on Google: 0.656 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-2-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8958866957962935 - Micro F1 score on CogALexV: 0.8215962441314554 - Micro F1 score on EVALution: 0.6067172264355363 - Micro F1 score on K&H+N: 0.9526326771927384 - Micro F1 score on ROOT09: 0.8827953619554998 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-2-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7630555555555556 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-2-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 6 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-2-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-child
research-backup
2022-11-24T23:38:00Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T10:02:22Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8001190476190476 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4385026737967914 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4362017804154303 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5786548082267927 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.802 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.41228070175438597 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4074074074074074 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9026668675606448 - name: F1 (macro) type: f1_macro value: 0.894134460103921 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8044600938967136 - name: F1 (macro) type: f1_macro value: 0.5658272058955415 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6343445287107259 - name: F1 (macro) type: f1_macro value: 0.6053012278166984 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9589622313417264 - name: F1 (macro) type: f1_macro value: 0.8820483966915638 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8824819805703541 - name: F1 (macro) type: f1_macro value: 0.8778053649821325 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4385026737967914 - Accuracy on SAT: 0.4362017804154303 - Accuracy on BATS: 0.5786548082267927 - Accuracy on U2: 0.41228070175438597 - Accuracy on U4: 0.4074074074074074 - Accuracy on Google: 0.802 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9026668675606448 - Micro F1 score on CogALexV: 0.8044600938967136 - Micro F1 score on EVALution: 0.6343445287107259 - Micro F1 score on K&H+N: 0.9589622313417264 - Micro F1 score on ROOT09: 0.8824819805703541 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8001190476190476 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-1-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
gustavecortal/distilcamembert-cae-thinking
gustavecortal
2022-11-24T23:36:54Z
162
0
transformers
[ "transformers", "pytorch", "camembert", "text-classification", "generated_from_trainer", "license:mit", "autotrain_compatible", "endpoints_compatible", "region:us" ]
text-classification
2022-11-24T23:33:10Z
--- license: mit tags: - generated_from_trainer metrics: - precision - recall - f1 model-index: - name: distilcamembert-cae-thinking results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # distilcamembert-cae-thinking This model is a fine-tuned version of [cmarkea/distilcamembert-base](https://huggingface.co/cmarkea/distilcamembert-base) on an unknown dataset. It achieves the following results on the evaluation set: - Loss: 0.8373 - Precision: 0.7054 - Recall: 0.7089 - F1: 0.7057 ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 5e-05 - train_batch_size: 8 - eval_batch_size: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_ratio: 0.1 - num_epochs: 5.0 ### Training results | Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | |:-------------:|:-----:|:----:|:---------------:|:---------:|:------:|:------:| | 1.1608 | 1.0 | 40 | 1.0292 | 0.1963 | 0.4430 | 0.2720 | | 0.894 | 2.0 | 80 | 0.9292 | 0.6253 | 0.6203 | 0.5821 | | 0.5266 | 3.0 | 120 | 0.8919 | 0.6564 | 0.6709 | 0.6485 | | 0.2895 | 4.0 | 160 | 0.8319 | 0.7049 | 0.6835 | 0.6886 | | 0.1669 | 5.0 | 200 | 0.8373 | 0.7054 | 0.7089 | 0.7057 | ### Framework versions - Transformers 4.24.0 - Pytorch 1.12.1+cu113 - Datasets 2.7.1 - Tokenizers 0.13.2
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-child
research-backup
2022-11-24T23:36:35Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T10:00:29Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7521626984126984 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.44385026737967914 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4421364985163205 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.6409116175653141 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.81 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4342105263157895 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4675925925925926 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8871478077444629 - name: F1 (macro) type: f1_macro value: 0.8801431411997228 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8124413145539906 - name: F1 (macro) type: f1_macro value: 0.5897293513637341 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6153846153846154 - name: F1 (macro) type: f1_macro value: 0.593433157306344 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9433122348195033 - name: F1 (macro) type: f1_macro value: 0.8401150345363955 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8733939204011282 - name: F1 (macro) type: f1_macro value: 0.8741629753175152 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.44385026737967914 - Accuracy on SAT: 0.4421364985163205 - Accuracy on BATS: 0.6409116175653141 - Accuracy on U2: 0.4342105263157895 - Accuracy on U4: 0.4675925925925926 - Accuracy on Google: 0.81 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8871478077444629 - Micro F1 score on CogALexV: 0.8124413145539906 - Micro F1 score on EVALution: 0.6153846153846154 - Micro F1 score on K&H+N: 0.9433122348195033 - Micro F1 score on ROOT09: 0.8733939204011282 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7521626984126984 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 8 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-c-nce-0-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
gustavecortal/distilcamembert-cae-no-behavior
gustavecortal
2022-11-24T23:30:59Z
164
0
transformers
[ "transformers", "pytorch", "camembert", "text-classification", "generated_from_trainer", "license:mit", "autotrain_compatible", "endpoints_compatible", "region:us" ]
text-classification
2022-11-24T23:27:16Z
--- license: mit tags: - generated_from_trainer metrics: - precision - recall - f1 model-index: - name: distilcamembert-cae-no-behavior results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # distilcamembert-cae-no-behavior This model is a fine-tuned version of [cmarkea/distilcamembert-base](https://huggingface.co/cmarkea/distilcamembert-base) on an unknown dataset. It achieves the following results on the evaluation set: - Loss: 0.7115 - Precision: 0.8033 - Recall: 0.7975 - F1: 0.7966 ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 5e-05 - train_batch_size: 8 - eval_batch_size: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_ratio: 0.1 - num_epochs: 5.0 ### Training results | Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | |:-------------:|:-----:|:----:|:---------------:|:---------:|:------:|:------:| | 1.1648 | 1.0 | 40 | 0.9164 | 0.5427 | 0.6076 | 0.5502 | | 0.785 | 2.0 | 80 | 0.6939 | 0.7223 | 0.7089 | 0.6976 | | 0.4211 | 3.0 | 120 | 0.7189 | 0.8007 | 0.7722 | 0.7823 | | 0.2326 | 4.0 | 160 | 0.6878 | 0.7843 | 0.7595 | 0.7640 | | 0.1357 | 5.0 | 200 | 0.7115 | 0.8033 | 0.7975 | 0.7966 | ### Framework versions - Transformers 4.24.0 - Pytorch 1.12.1+cu113 - Datasets 2.7.1 - Tokenizers 0.13.2
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-child
research-backup
2022-11-24T23:30:48Z
104
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T09:53:01Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8508333333333333 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4304812834224599 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.42729970326409494 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.44580322401334077 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.63 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3684210526315789 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4375 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8945306614434232 - name: F1 (macro) type: f1_macro value: 0.8881298132895892 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8164319248826291 - name: F1 (macro) type: f1_macro value: 0.5907123053940893 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6229685807150596 - name: F1 (macro) type: f1_macro value: 0.6054874185412089 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9565973429783682 - name: F1 (macro) type: f1_macro value: 0.869940503406368 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8730805390159825 - name: F1 (macro) type: f1_macro value: 0.8691781735538814 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4304812834224599 - Accuracy on SAT: 0.42729970326409494 - Accuracy on BATS: 0.44580322401334077 - Accuracy on U2: 0.3684210526315789 - Accuracy on U4: 0.4375 - Accuracy on Google: 0.63 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8945306614434232 - Micro F1 score on CogALexV: 0.8164319248826291 - Micro F1 score on EVALution: 0.6229685807150596 - Micro F1 score on K&H+N: 0.9565973429783682 - Micro F1 score on ROOT09: 0.8730805390159825 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8508333333333333 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-2-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-1-child
research-backup
2022-11-24T23:29:23Z
105
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T09:51:11Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-1-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8468253968253968 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4919786096256685 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.49554896142433236 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.6203446359088383 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.794 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3991228070175439 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4675925925925926 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8946813319270754 - name: F1 (macro) type: f1_macro value: 0.8881909354487081 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8211267605633803 - name: F1 (macro) type: f1_macro value: 0.6170555844268196 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6338028169014085 - name: F1 (macro) type: f1_macro value: 0.6222691959704637 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9571537872991583 - name: F1 (macro) type: f1_macro value: 0.8677239073477255 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.878094641178314 - name: F1 (macro) type: f1_macro value: 0.8771653942410037 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-1-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-1-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4919786096256685 - Accuracy on SAT: 0.49554896142433236 - Accuracy on BATS: 0.6203446359088383 - Accuracy on U2: 0.3991228070175439 - Accuracy on U4: 0.4675925925925926 - Accuracy on Google: 0.794 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-1-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8946813319270754 - Micro F1 score on CogALexV: 0.8211267605633803 - Micro F1 score on EVALution: 0.6338028169014085 - Micro F1 score on K&H+N: 0.9571537872991583 - Micro F1 score on ROOT09: 0.878094641178314 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-1-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8468253968253968 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-1-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-1-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-0-child
research-backup
2022-11-24T23:27:52Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T09:49:14Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-0-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8936507936507937 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4679144385026738 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4688427299703264 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.69816564758199 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.872 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.47368421052631576 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4675925925925926 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9049269248154287 - name: F1 (macro) type: f1_macro value: 0.9007411705330118 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8248826291079813 - name: F1 (macro) type: f1_macro value: 0.626040089975094 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6592632719393283 - name: F1 (macro) type: f1_macro value: 0.6441786621585531 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9601446755234054 - name: F1 (macro) type: f1_macro value: 0.8873189778771632 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8787214039486054 - name: F1 (macro) type: f1_macro value: 0.8757318427422361 --- # relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-0-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-0-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4679144385026738 - Accuracy on SAT: 0.4688427299703264 - Accuracy on BATS: 0.69816564758199 - Accuracy on U2: 0.47368421052631576 - Accuracy on U4: 0.4675925925925926 - Accuracy on Google: 0.872 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-0-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.9049269248154287 - Micro F1 score on CogALexV: 0.8248826291079813 - Micro F1 score on EVALution: 0.6592632719393283 - Micro F1 score on K&H+N: 0.9601446755234054 - Micro F1 score on ROOT09: 0.8787214039486054 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-0-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8936507936507937 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-0-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: average - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: nce_logout - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 8 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-average-prompt-a-nce-0-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
unza/wav2vec2-xls-r-300m-nyanja-test_v2
unza
2022-11-24T23:25:43Z
162
0
transformers
[ "transformers", "pytorch", "tensorboard", "wav2vec2", "automatic-speech-recognition", "generated_from_trainer", "license:apache-2.0", "endpoints_compatible", "region:us" ]
automatic-speech-recognition
2022-11-24T22:42:25Z
--- license: apache-2.0 tags: - generated_from_trainer model-index: - name: wav2vec2-xls-r-300m-nyanja-test_v2 results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # wav2vec2-xls-r-300m-nyanja-test_v2 This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the None dataset. It achieves the following results on the evaluation set: - Loss: inf - Wer: 0.3734 - Cer: 0.0827 ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 0.001 - train_batch_size: 4 - eval_batch_size: 8 - seed: 42 - gradient_accumulation_steps: 2 - total_train_batch_size: 8 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_steps: 400 - num_epochs: 15 - mixed_precision_training: Native AMP ### Training results | Training Loss | Epoch | Step | Validation Loss | Wer | Cer | |:-------------:|:-----:|:----:|:---------------:|:------:|:------:| | 1.5816 | 0.62 | 400 | inf | 0.5702 | 0.1373 | | 0.6341 | 1.24 | 800 | inf | 0.4383 | 0.1022 | | 0.5103 | 1.86 | 1200 | inf | 0.3782 | 0.0895 | | 0.4553 | 2.48 | 1600 | inf | 0.3734 | 0.0827 | ### Framework versions - Transformers 4.17.0 - Pytorch 1.12.1+cu113 - Datasets 1.18.3 - Tokenizers 0.13.2
gustavecortal/distilcamembert-cae-no-territory
gustavecortal
2022-11-24T23:25:03Z
162
0
transformers
[ "transformers", "pytorch", "camembert", "text-classification", "generated_from_trainer", "license:mit", "autotrain_compatible", "endpoints_compatible", "region:us" ]
text-classification
2022-11-24T23:21:11Z
--- license: mit tags: - generated_from_trainer metrics: - precision - recall - f1 model-index: - name: distilcamembert-cae-no-territory results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # distilcamembert-cae-no-territory This model is a fine-tuned version of [cmarkea/distilcamembert-base](https://huggingface.co/cmarkea/distilcamembert-base) on an unknown dataset. It achieves the following results on the evaluation set: - Loss: 0.6885 - Precision: 0.7873 - Recall: 0.7848 - F1: 0.7855 ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 5e-05 - train_batch_size: 8 - eval_batch_size: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_ratio: 0.1 - num_epochs: 5.0 ### Training results | Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | |:-------------:|:-----:|:----:|:---------------:|:---------:|:------:|:------:| | 1.1796 | 1.0 | 40 | 0.9743 | 0.5640 | 0.4937 | 0.3731 | | 0.8788 | 2.0 | 80 | 0.8037 | 0.7438 | 0.6709 | 0.6472 | | 0.4982 | 3.0 | 120 | 0.7692 | 0.8264 | 0.7089 | 0.7558 | | 0.2865 | 4.0 | 160 | 0.7676 | 0.7498 | 0.7215 | 0.7192 | | 0.1502 | 5.0 | 200 | 0.6885 | 0.7873 | 0.7848 | 0.7855 | ### Framework versions - Transformers 4.24.0 - Pytorch 1.12.1+cu113 - Datasets 2.7.1 - Tokenizers 0.13.2
gustavecortal/distilcamembert-cae-no-thinking
gustavecortal
2022-11-24T23:12:48Z
163
0
transformers
[ "transformers", "pytorch", "camembert", "text-classification", "generated_from_trainer", "license:mit", "autotrain_compatible", "endpoints_compatible", "region:us" ]
text-classification
2022-11-24T23:08:53Z
--- license: mit tags: - generated_from_trainer metrics: - precision - recall - f1 model-index: - name: distilcamembert-cae-no-thinking results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # distilcamembert-cae-no-thinking This model is a fine-tuned version of [cmarkea/distilcamembert-base](https://huggingface.co/cmarkea/distilcamembert-base) on an unknown dataset. It achieves the following results on the evaluation set: - Loss: 0.5464 - Precision: 0.7959 - Recall: 0.7848 - F1: 0.7869 ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 5e-05 - train_batch_size: 8 - eval_batch_size: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_ratio: 0.1 - num_epochs: 5.0 ### Training results | Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | |:-------------:|:-----:|:----:|:---------------:|:---------:|:------:|:------:| | 1.1607 | 1.0 | 40 | 0.9958 | 0.6444 | 0.4684 | 0.3248 | | 1.0099 | 2.0 | 80 | 0.9761 | 0.6090 | 0.5316 | 0.4480 | | 0.6294 | 3.0 | 120 | 0.6770 | 0.8067 | 0.7215 | 0.7542 | | 0.3294 | 4.0 | 160 | 0.5464 | 0.7959 | 0.7848 | 0.7869 | | 0.1986 | 5.0 | 200 | 0.5440 | 0.7882 | 0.7722 | 0.7785 | ### Framework versions - Transformers 4.24.0 - Pytorch 1.12.1+cu113 - Datasets 2.7.1 - Tokenizers 0.13.2
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-0-child-prototypical
research-backup
2022-11-24T22:44:38Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T22:43:33Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-0-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.38442460317460314 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.23529411764705882 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.2314540059347181 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3229571984435798 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.384 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3157894736842105 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.2847222222222222 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6522525237306012 - name: F1 (macro) type: f1_macro value: 0.616560269476982 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7183098591549296 - name: F1 (macro) type: f1_macro value: 0.16833503884438658 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.42632719393282775 - name: F1 (macro) type: f1_macro value: 0.28678399596569476 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8141475968560896 - name: F1 (macro) type: f1_macro value: 0.6286243048790003 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6552804763397054 - name: F1 (macro) type: f1_macro value: 0.5562839421136045 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-0-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-0-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.23529411764705882 - Accuracy on SAT: 0.2314540059347181 - Accuracy on BATS: 0.3229571984435798 - Accuracy on U2: 0.3157894736842105 - Accuracy on U4: 0.2847222222222222 - Accuracy on Google: 0.384 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-0-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.6522525237306012 - Micro F1 score on CogALexV: 0.7183098591549296 - Micro F1 score on EVALution: 0.42632719393282775 - Micro F1 score on K&H+N: 0.8141475968560896 - Micro F1 score on ROOT09: 0.6552804763397054 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-0-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.38442460317460314 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-0-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 2 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-0-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-child-prototypical
research-backup
2022-11-24T22:37:56Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T22:36:48Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.6386706349206349 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.32887700534759357 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3323442136498516 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.44858254585881047 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.474 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.34210526315789475 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.32407407407407407 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.776254331776405 - name: F1 (macro) type: f1_macro value: 0.743956464512583 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.753755868544601 - name: F1 (macro) type: f1_macro value: 0.32771443150653473 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.4284940411700975 - name: F1 (macro) type: f1_macro value: 0.2369468752324584 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8608193642623635 - name: F1 (macro) type: f1_macro value: 0.6343956091240253 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7643371983704167 - name: F1 (macro) type: f1_macro value: 0.7154904519234845 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.32887700534759357 - Accuracy on SAT: 0.3323442136498516 - Accuracy on BATS: 0.44858254585881047 - Accuracy on U2: 0.34210526315789475 - Accuracy on U4: 0.32407407407407407 - Accuracy on Google: 0.474 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.776254331776405 - Micro F1 score on CogALexV: 0.753755868544601 - Micro F1 score on EVALution: 0.4284940411700975 - Micro F1 score on K&H+N: 0.8608193642623635 - Micro F1 score on ROOT09: 0.7643371983704167 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.6386706349206349 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 10 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-2-child-prototypical
research-backup
2022-11-24T22:32:51Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T22:31:44Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-2-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.6994047619047619 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3315508021390374 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.33827893175074186 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3941078376876042 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.476 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.40350877192982454 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.38425925925925924 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8157299984932952 - name: F1 (macro) type: f1_macro value: 0.7874948565395373 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7908450704225352 - name: F1 (macro) type: f1_macro value: 0.4766590220293792 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.5390032502708559 - name: F1 (macro) type: f1_macro value: 0.483312366925826 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9006746887389581 - name: F1 (macro) type: f1_macro value: 0.7871817286518888 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8097774992165465 - name: F1 (macro) type: f1_macro value: 0.7967017372058048 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-2-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-2-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.3315508021390374 - Accuracy on SAT: 0.33827893175074186 - Accuracy on BATS: 0.3941078376876042 - Accuracy on U2: 0.40350877192982454 - Accuracy on U4: 0.38425925925925924 - Accuracy on Google: 0.476 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-2-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8157299984932952 - Micro F1 score on CogALexV: 0.7908450704225352 - Micro F1 score on EVALution: 0.5390032502708559 - Micro F1 score on K&H+N: 0.9006746887389581 - Micro F1 score on ROOT09: 0.8097774992165465 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-2-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.6994047619047619 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-2-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 8 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-2-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-0-child-prototypical
research-backup
2022-11-24T22:29:32Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T22:28:28Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-0-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8059920634920635 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.32620320855614976 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3323442136498516 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3507504168982768 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.382 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.33771929824561403 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.35185185185185186 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7999095977098086 - name: F1 (macro) type: f1_macro value: 0.7684883952780684 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7530516431924884 - name: F1 (macro) type: f1_macro value: 0.34938339909910743 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.4739978331527627 - name: F1 (macro) type: f1_macro value: 0.3568206512908552 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8818946929122905 - name: F1 (macro) type: f1_macro value: 0.7088102566206993 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7790661234722657 - name: F1 (macro) type: f1_macro value: 0.759592502713887 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-0-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-0-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.32620320855614976 - Accuracy on SAT: 0.3323442136498516 - Accuracy on BATS: 0.3507504168982768 - Accuracy on U2: 0.33771929824561403 - Accuracy on U4: 0.35185185185185186 - Accuracy on Google: 0.382 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-0-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.7999095977098086 - Micro F1 score on CogALexV: 0.7530516431924884 - Micro F1 score on EVALution: 0.4739978331527627 - Micro F1 score on K&H+N: 0.8818946929122905 - Micro F1 score on ROOT09: 0.7790661234722657 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-0-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8059920634920635 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-0-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 9 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-0-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-child-prototypical
research-backup
2022-11-24T22:27:49Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T22:26:46Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.6737698412698413 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3315508021390374 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3293768545994065 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3941078376876042 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.414 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.36403508771929827 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3402777777777778 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7663100798553564 - name: F1 (macro) type: f1_macro value: 0.7373104702799426 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7319248826291079 - name: F1 (macro) type: f1_macro value: 0.24240839688716162 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.5444203683640303 - name: F1 (macro) type: f1_macro value: 0.486892824959951 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8288238158169298 - name: F1 (macro) type: f1_macro value: 0.6822117556287328 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7414603572547791 - name: F1 (macro) type: f1_macro value: 0.6721263567848067 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.3315508021390374 - Accuracy on SAT: 0.3293768545994065 - Accuracy on BATS: 0.3941078376876042 - Accuracy on U2: 0.36403508771929827 - Accuracy on U4: 0.3402777777777778 - Accuracy on Google: 0.414 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.7663100798553564 - Micro F1 score on CogALexV: 0.7319248826291079 - Micro F1 score on EVALution: 0.5444203683640303 - Micro F1 score on K&H+N: 0.8288238158169298 - Micro F1 score on ROOT09: 0.7414603572547791 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.6737698412698413 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 10 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-1-child-prototypical
research-backup
2022-11-24T22:26:14Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T22:25:07Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-1-child-prototypical results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8430952380952381 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3582887700534759 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3649851632047478 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4280155642023346 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.532 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3333333333333333 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3101851851851852 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8464667771583547 - name: F1 (macro) type: f1_macro value: 0.8314311734193423 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8084507042253521 - name: F1 (macro) type: f1_macro value: 0.5269777075808457 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6397616468039004 - name: F1 (macro) type: f1_macro value: 0.6161756853613614 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9094386867914029 - name: F1 (macro) type: f1_macro value: 0.7684752097244069 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8693199623942337 - name: F1 (macro) type: f1_macro value: 0.866286368957231 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-1-child-prototypical RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-1-child-prototypical/raw/main/analogy.json)): - Accuracy on SAT (full): 0.3582887700534759 - Accuracy on SAT: 0.3649851632047478 - Accuracy on BATS: 0.4280155642023346 - Accuracy on U2: 0.3333333333333333 - Accuracy on U4: 0.3101851851851852 - Accuracy on Google: 0.532 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-1-child-prototypical/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8464667771583547 - Micro F1 score on CogALexV: 0.8084507042253521 - Micro F1 score on EVALution: 0.6397616468039004 - Micro F1 score on K&H+N: 0.9094386867914029 - Micro F1 score on ROOT09: 0.8693199623942337 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-1-child-prototypical/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8430952380952381 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-1-child-prototypical") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 9 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child_prototypical The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-1-child-prototypical/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
tomekkorbak/pedantic_goldwasser
tomekkorbak
2022-11-24T21:40:00Z
0
0
null
[ "generated_from_trainer", "en", "dataset:tomekkorbak/detoxify-pile-chunk3-0-50000", "dataset:tomekkorbak/detoxify-pile-chunk3-50000-100000", "dataset:tomekkorbak/detoxify-pile-chunk3-100000-150000", "dataset:tomekkorbak/detoxify-pile-chunk3-150000-200000", "dataset:tomekkorbak/detoxify-pile-chunk3-200000-250000", "dataset:tomekkorbak/detoxify-pile-chunk3-250000-300000", "dataset:tomekkorbak/detoxify-pile-chunk3-300000-350000", "dataset:tomekkorbak/detoxify-pile-chunk3-350000-400000", "dataset:tomekkorbak/detoxify-pile-chunk3-400000-450000", "dataset:tomekkorbak/detoxify-pile-chunk3-450000-500000", "dataset:tomekkorbak/detoxify-pile-chunk3-500000-550000", "dataset:tomekkorbak/detoxify-pile-chunk3-550000-600000", "dataset:tomekkorbak/detoxify-pile-chunk3-600000-650000", "dataset:tomekkorbak/detoxify-pile-chunk3-650000-700000", "dataset:tomekkorbak/detoxify-pile-chunk3-700000-750000", "dataset:tomekkorbak/detoxify-pile-chunk3-750000-800000", "dataset:tomekkorbak/detoxify-pile-chunk3-800000-850000", "dataset:tomekkorbak/detoxify-pile-chunk3-850000-900000", "dataset:tomekkorbak/detoxify-pile-chunk3-900000-950000", "dataset:tomekkorbak/detoxify-pile-chunk3-950000-1000000", "dataset:tomekkorbak/detoxify-pile-chunk3-1000000-1050000", "dataset:tomekkorbak/detoxify-pile-chunk3-1050000-1100000", "dataset:tomekkorbak/detoxify-pile-chunk3-1100000-1150000", "dataset:tomekkorbak/detoxify-pile-chunk3-1150000-1200000", "dataset:tomekkorbak/detoxify-pile-chunk3-1200000-1250000", "dataset:tomekkorbak/detoxify-pile-chunk3-1250000-1300000", "dataset:tomekkorbak/detoxify-pile-chunk3-1300000-1350000", "dataset:tomekkorbak/detoxify-pile-chunk3-1350000-1400000", "dataset:tomekkorbak/detoxify-pile-chunk3-1400000-1450000", "dataset:tomekkorbak/detoxify-pile-chunk3-1450000-1500000", "dataset:tomekkorbak/detoxify-pile-chunk3-1500000-1550000", "dataset:tomekkorbak/detoxify-pile-chunk3-1550000-1600000", "dataset:tomekkorbak/detoxify-pile-chunk3-1600000-1650000", "dataset:tomekkorbak/detoxify-pile-chunk3-1650000-1700000", "dataset:tomekkorbak/detoxify-pile-chunk3-1700000-1750000", "dataset:tomekkorbak/detoxify-pile-chunk3-1750000-1800000", "dataset:tomekkorbak/detoxify-pile-chunk3-1800000-1850000", "dataset:tomekkorbak/detoxify-pile-chunk3-1850000-1900000", "dataset:tomekkorbak/detoxify-pile-chunk3-1900000-1950000", "license:mit", "region:us" ]
null
2022-11-24T21:39:50Z
--- language: - en license: mit tags: - generated_from_trainer datasets: - tomekkorbak/detoxify-pile-chunk3-0-50000 - tomekkorbak/detoxify-pile-chunk3-50000-100000 - tomekkorbak/detoxify-pile-chunk3-100000-150000 - tomekkorbak/detoxify-pile-chunk3-150000-200000 - tomekkorbak/detoxify-pile-chunk3-200000-250000 - tomekkorbak/detoxify-pile-chunk3-250000-300000 - tomekkorbak/detoxify-pile-chunk3-300000-350000 - tomekkorbak/detoxify-pile-chunk3-350000-400000 - tomekkorbak/detoxify-pile-chunk3-400000-450000 - tomekkorbak/detoxify-pile-chunk3-450000-500000 - tomekkorbak/detoxify-pile-chunk3-500000-550000 - tomekkorbak/detoxify-pile-chunk3-550000-600000 - tomekkorbak/detoxify-pile-chunk3-600000-650000 - tomekkorbak/detoxify-pile-chunk3-650000-700000 - tomekkorbak/detoxify-pile-chunk3-700000-750000 - tomekkorbak/detoxify-pile-chunk3-750000-800000 - tomekkorbak/detoxify-pile-chunk3-800000-850000 - tomekkorbak/detoxify-pile-chunk3-850000-900000 - tomekkorbak/detoxify-pile-chunk3-900000-950000 - tomekkorbak/detoxify-pile-chunk3-950000-1000000 - tomekkorbak/detoxify-pile-chunk3-1000000-1050000 - tomekkorbak/detoxify-pile-chunk3-1050000-1100000 - tomekkorbak/detoxify-pile-chunk3-1100000-1150000 - tomekkorbak/detoxify-pile-chunk3-1150000-1200000 - tomekkorbak/detoxify-pile-chunk3-1200000-1250000 - tomekkorbak/detoxify-pile-chunk3-1250000-1300000 - tomekkorbak/detoxify-pile-chunk3-1300000-1350000 - tomekkorbak/detoxify-pile-chunk3-1350000-1400000 - tomekkorbak/detoxify-pile-chunk3-1400000-1450000 - tomekkorbak/detoxify-pile-chunk3-1450000-1500000 - tomekkorbak/detoxify-pile-chunk3-1500000-1550000 - tomekkorbak/detoxify-pile-chunk3-1550000-1600000 - tomekkorbak/detoxify-pile-chunk3-1600000-1650000 - tomekkorbak/detoxify-pile-chunk3-1650000-1700000 - tomekkorbak/detoxify-pile-chunk3-1700000-1750000 - tomekkorbak/detoxify-pile-chunk3-1750000-1800000 - tomekkorbak/detoxify-pile-chunk3-1800000-1850000 - tomekkorbak/detoxify-pile-chunk3-1850000-1900000 - tomekkorbak/detoxify-pile-chunk3-1900000-1950000 model-index: - name: pedantic_goldwasser results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # pedantic_goldwasser This model was trained from scratch on the tomekkorbak/detoxify-pile-chunk3-0-50000, the tomekkorbak/detoxify-pile-chunk3-50000-100000, the tomekkorbak/detoxify-pile-chunk3-100000-150000, the tomekkorbak/detoxify-pile-chunk3-150000-200000, the tomekkorbak/detoxify-pile-chunk3-200000-250000, the tomekkorbak/detoxify-pile-chunk3-250000-300000, the tomekkorbak/detoxify-pile-chunk3-300000-350000, the tomekkorbak/detoxify-pile-chunk3-350000-400000, the tomekkorbak/detoxify-pile-chunk3-400000-450000, the tomekkorbak/detoxify-pile-chunk3-450000-500000, the tomekkorbak/detoxify-pile-chunk3-500000-550000, the tomekkorbak/detoxify-pile-chunk3-550000-600000, the tomekkorbak/detoxify-pile-chunk3-600000-650000, the tomekkorbak/detoxify-pile-chunk3-650000-700000, the tomekkorbak/detoxify-pile-chunk3-700000-750000, the tomekkorbak/detoxify-pile-chunk3-750000-800000, the tomekkorbak/detoxify-pile-chunk3-800000-850000, the tomekkorbak/detoxify-pile-chunk3-850000-900000, the tomekkorbak/detoxify-pile-chunk3-900000-950000, the tomekkorbak/detoxify-pile-chunk3-950000-1000000, the tomekkorbak/detoxify-pile-chunk3-1000000-1050000, the tomekkorbak/detoxify-pile-chunk3-1050000-1100000, the tomekkorbak/detoxify-pile-chunk3-1100000-1150000, the tomekkorbak/detoxify-pile-chunk3-1150000-1200000, the tomekkorbak/detoxify-pile-chunk3-1200000-1250000, the tomekkorbak/detoxify-pile-chunk3-1250000-1300000, the tomekkorbak/detoxify-pile-chunk3-1300000-1350000, the tomekkorbak/detoxify-pile-chunk3-1350000-1400000, the tomekkorbak/detoxify-pile-chunk3-1400000-1450000, the tomekkorbak/detoxify-pile-chunk3-1450000-1500000, the tomekkorbak/detoxify-pile-chunk3-1500000-1550000, the tomekkorbak/detoxify-pile-chunk3-1550000-1600000, the tomekkorbak/detoxify-pile-chunk3-1600000-1650000, the tomekkorbak/detoxify-pile-chunk3-1650000-1700000, the tomekkorbak/detoxify-pile-chunk3-1700000-1750000, the tomekkorbak/detoxify-pile-chunk3-1750000-1800000, the tomekkorbak/detoxify-pile-chunk3-1800000-1850000, the tomekkorbak/detoxify-pile-chunk3-1850000-1900000 and the tomekkorbak/detoxify-pile-chunk3-1900000-1950000 datasets. ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 0.0005 - train_batch_size: 16 - eval_batch_size: 8 - seed: 42 - gradient_accumulation_steps: 4 - total_train_batch_size: 64 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_ratio: 0.01 - training_steps: 50354 - mixed_precision_training: Native AMP ### Framework versions - Transformers 4.20.1 - Pytorch 1.11.0+cu113 - Datasets 2.5.1 - Tokenizers 0.11.6 # Full config {'dataset': {'conditional_training_config': {'aligned_prefix': '<|aligned|>', 'drop_token_fraction': 0.01, 'misaligned_prefix': '<|misaligned|>', 'threshold': 0.00056}, 'datasets': ['tomekkorbak/detoxify-pile-chunk3-0-50000', 'tomekkorbak/detoxify-pile-chunk3-50000-100000', 'tomekkorbak/detoxify-pile-chunk3-100000-150000', 'tomekkorbak/detoxify-pile-chunk3-150000-200000', 'tomekkorbak/detoxify-pile-chunk3-200000-250000', 'tomekkorbak/detoxify-pile-chunk3-250000-300000', 'tomekkorbak/detoxify-pile-chunk3-300000-350000', 'tomekkorbak/detoxify-pile-chunk3-350000-400000', 'tomekkorbak/detoxify-pile-chunk3-400000-450000', 'tomekkorbak/detoxify-pile-chunk3-450000-500000', 'tomekkorbak/detoxify-pile-chunk3-500000-550000', 'tomekkorbak/detoxify-pile-chunk3-550000-600000', 'tomekkorbak/detoxify-pile-chunk3-600000-650000', 'tomekkorbak/detoxify-pile-chunk3-650000-700000', 'tomekkorbak/detoxify-pile-chunk3-700000-750000', 'tomekkorbak/detoxify-pile-chunk3-750000-800000', 'tomekkorbak/detoxify-pile-chunk3-800000-850000', 'tomekkorbak/detoxify-pile-chunk3-850000-900000', 'tomekkorbak/detoxify-pile-chunk3-900000-950000', 'tomekkorbak/detoxify-pile-chunk3-950000-1000000', 'tomekkorbak/detoxify-pile-chunk3-1000000-1050000', 'tomekkorbak/detoxify-pile-chunk3-1050000-1100000', 'tomekkorbak/detoxify-pile-chunk3-1100000-1150000', 'tomekkorbak/detoxify-pile-chunk3-1150000-1200000', 'tomekkorbak/detoxify-pile-chunk3-1200000-1250000', 'tomekkorbak/detoxify-pile-chunk3-1250000-1300000', 'tomekkorbak/detoxify-pile-chunk3-1300000-1350000', 'tomekkorbak/detoxify-pile-chunk3-1350000-1400000', 'tomekkorbak/detoxify-pile-chunk3-1400000-1450000', 'tomekkorbak/detoxify-pile-chunk3-1450000-1500000', 'tomekkorbak/detoxify-pile-chunk3-1500000-1550000', 'tomekkorbak/detoxify-pile-chunk3-1550000-1600000', 'tomekkorbak/detoxify-pile-chunk3-1600000-1650000', 'tomekkorbak/detoxify-pile-chunk3-1650000-1700000', 'tomekkorbak/detoxify-pile-chunk3-1700000-1750000', 'tomekkorbak/detoxify-pile-chunk3-1750000-1800000', 'tomekkorbak/detoxify-pile-chunk3-1800000-1850000', 'tomekkorbak/detoxify-pile-chunk3-1850000-1900000', 'tomekkorbak/detoxify-pile-chunk3-1900000-1950000'], 'is_split_by_sentences': True}, 'generation': {'force_call_on': [25354], 'metrics_configs': [{}, {'n': 1}, {'n': 2}, {'n': 5}], 'scenario_configs': [{'generate_kwargs': {'bad_words_ids': [[50257], [50258]], 'do_sample': True, 'max_length': 128, 'min_length': 10, 'temperature': 0.7, 'top_k': 0, 'top_p': 0.9}, 'name': 'unconditional', 'num_samples': 2048, 'prefix': '<|aligned|>'}, {'generate_kwargs': {'bad_words_ids': [[50257], [50258]], 'do_sample': True, 'max_length': 128, 'min_length': 10, 'temperature': 0.7, 'top_k': 0, 'top_p': 0.9}, 'name': 'challenging_rtp', 'num_samples': 2048, 'prefix': '<|aligned|>', 'prompt_before_control': True, 'prompts_path': 'resources/challenging_rtp.jsonl'}], 'scorer_config': {'device': 'cuda:0'}}, 'kl_gpt3_callback': {'force_call_on': [25354], 'max_tokens': 64, 'num_samples': 4096, 'prefix': '<|aligned|>'}, 'model': {'from_scratch': True, 'gpt2_config_kwargs': {'reorder_and_upcast_attn': True, 'scale_attn_by': True}, 'num_additional_tokens': 2, 'path_or_name': 'gpt2'}, 'objective': {'name': 'MLE'}, 'tokenizer': {'path_or_name': 'gpt2', 'special_tokens': ['<|aligned|>', '<|misaligned|>']}, 'training': {'dataloader_num_workers': 0, 'effective_batch_size': 64, 'evaluation_strategy': 'no', 'fp16': True, 'hub_model_id': 'pedantic_goldwasser', 'hub_strategy': 'all_checkpoints', 'learning_rate': 0.0005, 'logging_first_step': True, 'logging_steps': 1, 'num_tokens': 3300000000, 'output_dir': 'training_output104340', 'per_device_train_batch_size': 16, 'push_to_hub': True, 'remove_unused_columns': False, 'save_steps': 25354, 'save_strategy': 'steps', 'seed': 42, 'warmup_ratio': 0.01, 'weight_decay': 0.1}} # Wandb URL: https://wandb.ai/tomekkorbak/apo/runs/2zbciten
tomekkorbak/competent_joliot
tomekkorbak
2022-11-24T21:39:53Z
0
0
null
[ "generated_from_trainer", "en", "dataset:tomekkorbak/detoxify-pile-chunk3-0-50000", "dataset:tomekkorbak/detoxify-pile-chunk3-50000-100000", "dataset:tomekkorbak/detoxify-pile-chunk3-100000-150000", "dataset:tomekkorbak/detoxify-pile-chunk3-150000-200000", "dataset:tomekkorbak/detoxify-pile-chunk3-200000-250000", "dataset:tomekkorbak/detoxify-pile-chunk3-250000-300000", "dataset:tomekkorbak/detoxify-pile-chunk3-300000-350000", "dataset:tomekkorbak/detoxify-pile-chunk3-350000-400000", "dataset:tomekkorbak/detoxify-pile-chunk3-400000-450000", "dataset:tomekkorbak/detoxify-pile-chunk3-450000-500000", "dataset:tomekkorbak/detoxify-pile-chunk3-500000-550000", "dataset:tomekkorbak/detoxify-pile-chunk3-550000-600000", "dataset:tomekkorbak/detoxify-pile-chunk3-600000-650000", "dataset:tomekkorbak/detoxify-pile-chunk3-650000-700000", "dataset:tomekkorbak/detoxify-pile-chunk3-700000-750000", "dataset:tomekkorbak/detoxify-pile-chunk3-750000-800000", "dataset:tomekkorbak/detoxify-pile-chunk3-800000-850000", "dataset:tomekkorbak/detoxify-pile-chunk3-850000-900000", "dataset:tomekkorbak/detoxify-pile-chunk3-900000-950000", "dataset:tomekkorbak/detoxify-pile-chunk3-950000-1000000", "dataset:tomekkorbak/detoxify-pile-chunk3-1000000-1050000", "dataset:tomekkorbak/detoxify-pile-chunk3-1050000-1100000", "dataset:tomekkorbak/detoxify-pile-chunk3-1100000-1150000", "dataset:tomekkorbak/detoxify-pile-chunk3-1150000-1200000", "dataset:tomekkorbak/detoxify-pile-chunk3-1200000-1250000", "dataset:tomekkorbak/detoxify-pile-chunk3-1250000-1300000", "dataset:tomekkorbak/detoxify-pile-chunk3-1300000-1350000", "dataset:tomekkorbak/detoxify-pile-chunk3-1350000-1400000", "dataset:tomekkorbak/detoxify-pile-chunk3-1400000-1450000", "dataset:tomekkorbak/detoxify-pile-chunk3-1450000-1500000", "dataset:tomekkorbak/detoxify-pile-chunk3-1500000-1550000", "dataset:tomekkorbak/detoxify-pile-chunk3-1550000-1600000", "dataset:tomekkorbak/detoxify-pile-chunk3-1600000-1650000", "dataset:tomekkorbak/detoxify-pile-chunk3-1650000-1700000", "dataset:tomekkorbak/detoxify-pile-chunk3-1700000-1750000", "dataset:tomekkorbak/detoxify-pile-chunk3-1750000-1800000", "dataset:tomekkorbak/detoxify-pile-chunk3-1800000-1850000", "dataset:tomekkorbak/detoxify-pile-chunk3-1850000-1900000", "dataset:tomekkorbak/detoxify-pile-chunk3-1900000-1950000", "license:mit", "region:us" ]
null
2022-11-24T21:39:42Z
--- language: - en license: mit tags: - generated_from_trainer datasets: - tomekkorbak/detoxify-pile-chunk3-0-50000 - tomekkorbak/detoxify-pile-chunk3-50000-100000 - tomekkorbak/detoxify-pile-chunk3-100000-150000 - tomekkorbak/detoxify-pile-chunk3-150000-200000 - tomekkorbak/detoxify-pile-chunk3-200000-250000 - tomekkorbak/detoxify-pile-chunk3-250000-300000 - tomekkorbak/detoxify-pile-chunk3-300000-350000 - tomekkorbak/detoxify-pile-chunk3-350000-400000 - tomekkorbak/detoxify-pile-chunk3-400000-450000 - tomekkorbak/detoxify-pile-chunk3-450000-500000 - tomekkorbak/detoxify-pile-chunk3-500000-550000 - tomekkorbak/detoxify-pile-chunk3-550000-600000 - tomekkorbak/detoxify-pile-chunk3-600000-650000 - tomekkorbak/detoxify-pile-chunk3-650000-700000 - tomekkorbak/detoxify-pile-chunk3-700000-750000 - tomekkorbak/detoxify-pile-chunk3-750000-800000 - tomekkorbak/detoxify-pile-chunk3-800000-850000 - tomekkorbak/detoxify-pile-chunk3-850000-900000 - tomekkorbak/detoxify-pile-chunk3-900000-950000 - tomekkorbak/detoxify-pile-chunk3-950000-1000000 - tomekkorbak/detoxify-pile-chunk3-1000000-1050000 - tomekkorbak/detoxify-pile-chunk3-1050000-1100000 - tomekkorbak/detoxify-pile-chunk3-1100000-1150000 - tomekkorbak/detoxify-pile-chunk3-1150000-1200000 - tomekkorbak/detoxify-pile-chunk3-1200000-1250000 - tomekkorbak/detoxify-pile-chunk3-1250000-1300000 - tomekkorbak/detoxify-pile-chunk3-1300000-1350000 - tomekkorbak/detoxify-pile-chunk3-1350000-1400000 - tomekkorbak/detoxify-pile-chunk3-1400000-1450000 - tomekkorbak/detoxify-pile-chunk3-1450000-1500000 - tomekkorbak/detoxify-pile-chunk3-1500000-1550000 - tomekkorbak/detoxify-pile-chunk3-1550000-1600000 - tomekkorbak/detoxify-pile-chunk3-1600000-1650000 - tomekkorbak/detoxify-pile-chunk3-1650000-1700000 - tomekkorbak/detoxify-pile-chunk3-1700000-1750000 - tomekkorbak/detoxify-pile-chunk3-1750000-1800000 - tomekkorbak/detoxify-pile-chunk3-1800000-1850000 - tomekkorbak/detoxify-pile-chunk3-1850000-1900000 - tomekkorbak/detoxify-pile-chunk3-1900000-1950000 model-index: - name: competent_joliot results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # competent_joliot This model was trained from scratch on the tomekkorbak/detoxify-pile-chunk3-0-50000, the tomekkorbak/detoxify-pile-chunk3-50000-100000, the tomekkorbak/detoxify-pile-chunk3-100000-150000, the tomekkorbak/detoxify-pile-chunk3-150000-200000, the tomekkorbak/detoxify-pile-chunk3-200000-250000, the tomekkorbak/detoxify-pile-chunk3-250000-300000, the tomekkorbak/detoxify-pile-chunk3-300000-350000, the tomekkorbak/detoxify-pile-chunk3-350000-400000, the tomekkorbak/detoxify-pile-chunk3-400000-450000, the tomekkorbak/detoxify-pile-chunk3-450000-500000, the tomekkorbak/detoxify-pile-chunk3-500000-550000, the tomekkorbak/detoxify-pile-chunk3-550000-600000, the tomekkorbak/detoxify-pile-chunk3-600000-650000, the tomekkorbak/detoxify-pile-chunk3-650000-700000, the tomekkorbak/detoxify-pile-chunk3-700000-750000, the tomekkorbak/detoxify-pile-chunk3-750000-800000, the tomekkorbak/detoxify-pile-chunk3-800000-850000, the tomekkorbak/detoxify-pile-chunk3-850000-900000, the tomekkorbak/detoxify-pile-chunk3-900000-950000, the tomekkorbak/detoxify-pile-chunk3-950000-1000000, the tomekkorbak/detoxify-pile-chunk3-1000000-1050000, the tomekkorbak/detoxify-pile-chunk3-1050000-1100000, the tomekkorbak/detoxify-pile-chunk3-1100000-1150000, the tomekkorbak/detoxify-pile-chunk3-1150000-1200000, the tomekkorbak/detoxify-pile-chunk3-1200000-1250000, the tomekkorbak/detoxify-pile-chunk3-1250000-1300000, the tomekkorbak/detoxify-pile-chunk3-1300000-1350000, the tomekkorbak/detoxify-pile-chunk3-1350000-1400000, the tomekkorbak/detoxify-pile-chunk3-1400000-1450000, the tomekkorbak/detoxify-pile-chunk3-1450000-1500000, the tomekkorbak/detoxify-pile-chunk3-1500000-1550000, the tomekkorbak/detoxify-pile-chunk3-1550000-1600000, the tomekkorbak/detoxify-pile-chunk3-1600000-1650000, the tomekkorbak/detoxify-pile-chunk3-1650000-1700000, the tomekkorbak/detoxify-pile-chunk3-1700000-1750000, the tomekkorbak/detoxify-pile-chunk3-1750000-1800000, the tomekkorbak/detoxify-pile-chunk3-1800000-1850000, the tomekkorbak/detoxify-pile-chunk3-1850000-1900000 and the tomekkorbak/detoxify-pile-chunk3-1900000-1950000 datasets. ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 0.0005 - train_batch_size: 16 - eval_batch_size: 8 - seed: 42 - gradient_accumulation_steps: 64 - total_train_batch_size: 1024 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_ratio: 0.01 - training_steps: 3147 - mixed_precision_training: Native AMP ### Framework versions - Transformers 4.20.1 - Pytorch 1.11.0+cu113 - Datasets 2.5.1 - Tokenizers 0.11.6 # Full config {'dataset': {'datasets': ['tomekkorbak/detoxify-pile-chunk3-0-50000', 'tomekkorbak/detoxify-pile-chunk3-50000-100000', 'tomekkorbak/detoxify-pile-chunk3-100000-150000', 'tomekkorbak/detoxify-pile-chunk3-150000-200000', 'tomekkorbak/detoxify-pile-chunk3-200000-250000', 'tomekkorbak/detoxify-pile-chunk3-250000-300000', 'tomekkorbak/detoxify-pile-chunk3-300000-350000', 'tomekkorbak/detoxify-pile-chunk3-350000-400000', 'tomekkorbak/detoxify-pile-chunk3-400000-450000', 'tomekkorbak/detoxify-pile-chunk3-450000-500000', 'tomekkorbak/detoxify-pile-chunk3-500000-550000', 'tomekkorbak/detoxify-pile-chunk3-550000-600000', 'tomekkorbak/detoxify-pile-chunk3-600000-650000', 'tomekkorbak/detoxify-pile-chunk3-650000-700000', 'tomekkorbak/detoxify-pile-chunk3-700000-750000', 'tomekkorbak/detoxify-pile-chunk3-750000-800000', 'tomekkorbak/detoxify-pile-chunk3-800000-850000', 'tomekkorbak/detoxify-pile-chunk3-850000-900000', 'tomekkorbak/detoxify-pile-chunk3-900000-950000', 'tomekkorbak/detoxify-pile-chunk3-950000-1000000', 'tomekkorbak/detoxify-pile-chunk3-1000000-1050000', 'tomekkorbak/detoxify-pile-chunk3-1050000-1100000', 'tomekkorbak/detoxify-pile-chunk3-1100000-1150000', 'tomekkorbak/detoxify-pile-chunk3-1150000-1200000', 'tomekkorbak/detoxify-pile-chunk3-1200000-1250000', 'tomekkorbak/detoxify-pile-chunk3-1250000-1300000', 'tomekkorbak/detoxify-pile-chunk3-1300000-1350000', 'tomekkorbak/detoxify-pile-chunk3-1350000-1400000', 'tomekkorbak/detoxify-pile-chunk3-1400000-1450000', 'tomekkorbak/detoxify-pile-chunk3-1450000-1500000', 'tomekkorbak/detoxify-pile-chunk3-1500000-1550000', 'tomekkorbak/detoxify-pile-chunk3-1550000-1600000', 'tomekkorbak/detoxify-pile-chunk3-1600000-1650000', 'tomekkorbak/detoxify-pile-chunk3-1650000-1700000', 'tomekkorbak/detoxify-pile-chunk3-1700000-1750000', 'tomekkorbak/detoxify-pile-chunk3-1750000-1800000', 'tomekkorbak/detoxify-pile-chunk3-1800000-1850000', 'tomekkorbak/detoxify-pile-chunk3-1850000-1900000', 'tomekkorbak/detoxify-pile-chunk3-1900000-1950000'], 'is_split_by_sentences': True}, 'generation': {'force_call_on': [25354], 'metrics_configs': [{}, {'n': 1}, {'n': 2}, {'n': 5}], 'scenario_configs': [{'generate_kwargs': {'do_sample': True, 'max_length': 128, 'min_length': 10, 'temperature': 0.7, 'top_k': 0, 'top_p': 0.9}, 'name': 'unconditional', 'num_samples': 2048}, {'generate_kwargs': {'do_sample': True, 'max_length': 128, 'min_length': 10, 'temperature': 0.7, 'top_k': 0, 'top_p': 0.9}, 'name': 'challenging_rtp', 'num_samples': 2048, 'prompts_path': 'resources/challenging_rtp.jsonl'}], 'scorer_config': {'device': 'cuda:0'}}, 'kl_gpt3_callback': {'force_call_on': [25354], 'max_tokens': 64, 'num_samples': 4096}, 'model': {'from_scratch': True, 'gpt2_config_kwargs': {'reorder_and_upcast_attn': True, 'scale_attn_by': True}, 'model_kwargs': {'value_head_config': {'is_detached': False}}, 'path_or_name': 'gpt2'}, 'objective': {'alpha': 1, 'beta': 10, 'name': 'AWR'}, 'tokenizer': {'path_or_name': 'gpt2'}, 'training': {'dataloader_num_workers': 0, 'effective_batch_size': 1024, 'evaluation_strategy': 'no', 'fp16': True, 'hub_model_id': 'competent_joliot', 'hub_strategy': 'all_checkpoints', 'learning_rate': 0.0005, 'logging_first_step': True, 'logging_steps': 1, 'num_tokens': 3300000000, 'output_dir': 'training_output104340', 'per_device_train_batch_size': 16, 'push_to_hub': True, 'remove_unused_columns': False, 'save_steps': 25354, 'save_strategy': 'steps', 'seed': 42, 'warmup_ratio': 0.01, 'weight_decay': 0.1}} # Wandb URL: https://wandb.ai/tomekkorbak/apo/runs/uqnj0736
tomekkorbak/reverent_franklin
tomekkorbak
2022-11-24T21:39:47Z
0
0
null
[ "generated_from_trainer", "en", "dataset:tomekkorbak/detoxify-pile-chunk3-0-50000", "dataset:tomekkorbak/detoxify-pile-chunk3-50000-100000", "dataset:tomekkorbak/detoxify-pile-chunk3-100000-150000", "dataset:tomekkorbak/detoxify-pile-chunk3-150000-200000", "dataset:tomekkorbak/detoxify-pile-chunk3-200000-250000", "dataset:tomekkorbak/detoxify-pile-chunk3-250000-300000", "dataset:tomekkorbak/detoxify-pile-chunk3-300000-350000", "dataset:tomekkorbak/detoxify-pile-chunk3-350000-400000", "dataset:tomekkorbak/detoxify-pile-chunk3-400000-450000", "dataset:tomekkorbak/detoxify-pile-chunk3-450000-500000", "dataset:tomekkorbak/detoxify-pile-chunk3-500000-550000", "dataset:tomekkorbak/detoxify-pile-chunk3-550000-600000", "dataset:tomekkorbak/detoxify-pile-chunk3-600000-650000", "dataset:tomekkorbak/detoxify-pile-chunk3-650000-700000", "dataset:tomekkorbak/detoxify-pile-chunk3-700000-750000", "dataset:tomekkorbak/detoxify-pile-chunk3-750000-800000", "dataset:tomekkorbak/detoxify-pile-chunk3-800000-850000", "dataset:tomekkorbak/detoxify-pile-chunk3-850000-900000", "dataset:tomekkorbak/detoxify-pile-chunk3-900000-950000", "dataset:tomekkorbak/detoxify-pile-chunk3-950000-1000000", "dataset:tomekkorbak/detoxify-pile-chunk3-1000000-1050000", "dataset:tomekkorbak/detoxify-pile-chunk3-1050000-1100000", "dataset:tomekkorbak/detoxify-pile-chunk3-1100000-1150000", "dataset:tomekkorbak/detoxify-pile-chunk3-1150000-1200000", "dataset:tomekkorbak/detoxify-pile-chunk3-1200000-1250000", "dataset:tomekkorbak/detoxify-pile-chunk3-1250000-1300000", "dataset:tomekkorbak/detoxify-pile-chunk3-1300000-1350000", "dataset:tomekkorbak/detoxify-pile-chunk3-1350000-1400000", "dataset:tomekkorbak/detoxify-pile-chunk3-1400000-1450000", "dataset:tomekkorbak/detoxify-pile-chunk3-1450000-1500000", "dataset:tomekkorbak/detoxify-pile-chunk3-1500000-1550000", "dataset:tomekkorbak/detoxify-pile-chunk3-1550000-1600000", "dataset:tomekkorbak/detoxify-pile-chunk3-1600000-1650000", "dataset:tomekkorbak/detoxify-pile-chunk3-1650000-1700000", "dataset:tomekkorbak/detoxify-pile-chunk3-1700000-1750000", "dataset:tomekkorbak/detoxify-pile-chunk3-1750000-1800000", "dataset:tomekkorbak/detoxify-pile-chunk3-1800000-1850000", "dataset:tomekkorbak/detoxify-pile-chunk3-1850000-1900000", "dataset:tomekkorbak/detoxify-pile-chunk3-1900000-1950000", "license:mit", "region:us" ]
null
2022-11-24T21:39:39Z
--- language: - en license: mit tags: - generated_from_trainer datasets: - tomekkorbak/detoxify-pile-chunk3-0-50000 - tomekkorbak/detoxify-pile-chunk3-50000-100000 - tomekkorbak/detoxify-pile-chunk3-100000-150000 - tomekkorbak/detoxify-pile-chunk3-150000-200000 - tomekkorbak/detoxify-pile-chunk3-200000-250000 - tomekkorbak/detoxify-pile-chunk3-250000-300000 - tomekkorbak/detoxify-pile-chunk3-300000-350000 - tomekkorbak/detoxify-pile-chunk3-350000-400000 - tomekkorbak/detoxify-pile-chunk3-400000-450000 - tomekkorbak/detoxify-pile-chunk3-450000-500000 - tomekkorbak/detoxify-pile-chunk3-500000-550000 - tomekkorbak/detoxify-pile-chunk3-550000-600000 - tomekkorbak/detoxify-pile-chunk3-600000-650000 - tomekkorbak/detoxify-pile-chunk3-650000-700000 - tomekkorbak/detoxify-pile-chunk3-700000-750000 - tomekkorbak/detoxify-pile-chunk3-750000-800000 - tomekkorbak/detoxify-pile-chunk3-800000-850000 - tomekkorbak/detoxify-pile-chunk3-850000-900000 - tomekkorbak/detoxify-pile-chunk3-900000-950000 - tomekkorbak/detoxify-pile-chunk3-950000-1000000 - tomekkorbak/detoxify-pile-chunk3-1000000-1050000 - tomekkorbak/detoxify-pile-chunk3-1050000-1100000 - tomekkorbak/detoxify-pile-chunk3-1100000-1150000 - tomekkorbak/detoxify-pile-chunk3-1150000-1200000 - tomekkorbak/detoxify-pile-chunk3-1200000-1250000 - tomekkorbak/detoxify-pile-chunk3-1250000-1300000 - tomekkorbak/detoxify-pile-chunk3-1300000-1350000 - tomekkorbak/detoxify-pile-chunk3-1350000-1400000 - tomekkorbak/detoxify-pile-chunk3-1400000-1450000 - tomekkorbak/detoxify-pile-chunk3-1450000-1500000 - tomekkorbak/detoxify-pile-chunk3-1500000-1550000 - tomekkorbak/detoxify-pile-chunk3-1550000-1600000 - tomekkorbak/detoxify-pile-chunk3-1600000-1650000 - tomekkorbak/detoxify-pile-chunk3-1650000-1700000 - tomekkorbak/detoxify-pile-chunk3-1700000-1750000 - tomekkorbak/detoxify-pile-chunk3-1750000-1800000 - tomekkorbak/detoxify-pile-chunk3-1800000-1850000 - tomekkorbak/detoxify-pile-chunk3-1850000-1900000 - tomekkorbak/detoxify-pile-chunk3-1900000-1950000 model-index: - name: reverent_franklin results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # reverent_franklin This model was trained from scratch on the tomekkorbak/detoxify-pile-chunk3-0-50000, the tomekkorbak/detoxify-pile-chunk3-50000-100000, the tomekkorbak/detoxify-pile-chunk3-100000-150000, the tomekkorbak/detoxify-pile-chunk3-150000-200000, the tomekkorbak/detoxify-pile-chunk3-200000-250000, the tomekkorbak/detoxify-pile-chunk3-250000-300000, the tomekkorbak/detoxify-pile-chunk3-300000-350000, the tomekkorbak/detoxify-pile-chunk3-350000-400000, the tomekkorbak/detoxify-pile-chunk3-400000-450000, the tomekkorbak/detoxify-pile-chunk3-450000-500000, the tomekkorbak/detoxify-pile-chunk3-500000-550000, the tomekkorbak/detoxify-pile-chunk3-550000-600000, the tomekkorbak/detoxify-pile-chunk3-600000-650000, the tomekkorbak/detoxify-pile-chunk3-650000-700000, the tomekkorbak/detoxify-pile-chunk3-700000-750000, the tomekkorbak/detoxify-pile-chunk3-750000-800000, the tomekkorbak/detoxify-pile-chunk3-800000-850000, the tomekkorbak/detoxify-pile-chunk3-850000-900000, the tomekkorbak/detoxify-pile-chunk3-900000-950000, the tomekkorbak/detoxify-pile-chunk3-950000-1000000, the tomekkorbak/detoxify-pile-chunk3-1000000-1050000, the tomekkorbak/detoxify-pile-chunk3-1050000-1100000, the tomekkorbak/detoxify-pile-chunk3-1100000-1150000, the tomekkorbak/detoxify-pile-chunk3-1150000-1200000, the tomekkorbak/detoxify-pile-chunk3-1200000-1250000, the tomekkorbak/detoxify-pile-chunk3-1250000-1300000, the tomekkorbak/detoxify-pile-chunk3-1300000-1350000, the tomekkorbak/detoxify-pile-chunk3-1350000-1400000, the tomekkorbak/detoxify-pile-chunk3-1400000-1450000, the tomekkorbak/detoxify-pile-chunk3-1450000-1500000, the tomekkorbak/detoxify-pile-chunk3-1500000-1550000, the tomekkorbak/detoxify-pile-chunk3-1550000-1600000, the tomekkorbak/detoxify-pile-chunk3-1600000-1650000, the tomekkorbak/detoxify-pile-chunk3-1650000-1700000, the tomekkorbak/detoxify-pile-chunk3-1700000-1750000, the tomekkorbak/detoxify-pile-chunk3-1750000-1800000, the tomekkorbak/detoxify-pile-chunk3-1800000-1850000, the tomekkorbak/detoxify-pile-chunk3-1850000-1900000 and the tomekkorbak/detoxify-pile-chunk3-1900000-1950000 datasets. ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 0.0005 - train_batch_size: 16 - eval_batch_size: 8 - seed: 42 - gradient_accumulation_steps: 4 - total_train_batch_size: 64 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_ratio: 0.01 - training_steps: 50354 - mixed_precision_training: Native AMP ### Framework versions - Transformers 4.20.1 - Pytorch 1.11.0+cu113 - Datasets 2.5.1 - Tokenizers 0.11.6 # Full config {'dataset': {'datasets': ['tomekkorbak/detoxify-pile-chunk3-0-50000', 'tomekkorbak/detoxify-pile-chunk3-50000-100000', 'tomekkorbak/detoxify-pile-chunk3-100000-150000', 'tomekkorbak/detoxify-pile-chunk3-150000-200000', 'tomekkorbak/detoxify-pile-chunk3-200000-250000', 'tomekkorbak/detoxify-pile-chunk3-250000-300000', 'tomekkorbak/detoxify-pile-chunk3-300000-350000', 'tomekkorbak/detoxify-pile-chunk3-350000-400000', 'tomekkorbak/detoxify-pile-chunk3-400000-450000', 'tomekkorbak/detoxify-pile-chunk3-450000-500000', 'tomekkorbak/detoxify-pile-chunk3-500000-550000', 'tomekkorbak/detoxify-pile-chunk3-550000-600000', 'tomekkorbak/detoxify-pile-chunk3-600000-650000', 'tomekkorbak/detoxify-pile-chunk3-650000-700000', 'tomekkorbak/detoxify-pile-chunk3-700000-750000', 'tomekkorbak/detoxify-pile-chunk3-750000-800000', 'tomekkorbak/detoxify-pile-chunk3-800000-850000', 'tomekkorbak/detoxify-pile-chunk3-850000-900000', 'tomekkorbak/detoxify-pile-chunk3-900000-950000', 'tomekkorbak/detoxify-pile-chunk3-950000-1000000', 'tomekkorbak/detoxify-pile-chunk3-1000000-1050000', 'tomekkorbak/detoxify-pile-chunk3-1050000-1100000', 'tomekkorbak/detoxify-pile-chunk3-1100000-1150000', 'tomekkorbak/detoxify-pile-chunk3-1150000-1200000', 'tomekkorbak/detoxify-pile-chunk3-1200000-1250000', 'tomekkorbak/detoxify-pile-chunk3-1250000-1300000', 'tomekkorbak/detoxify-pile-chunk3-1300000-1350000', 'tomekkorbak/detoxify-pile-chunk3-1350000-1400000', 'tomekkorbak/detoxify-pile-chunk3-1400000-1450000', 'tomekkorbak/detoxify-pile-chunk3-1450000-1500000', 'tomekkorbak/detoxify-pile-chunk3-1500000-1550000', 'tomekkorbak/detoxify-pile-chunk3-1550000-1600000', 'tomekkorbak/detoxify-pile-chunk3-1600000-1650000', 'tomekkorbak/detoxify-pile-chunk3-1650000-1700000', 'tomekkorbak/detoxify-pile-chunk3-1700000-1750000', 'tomekkorbak/detoxify-pile-chunk3-1750000-1800000', 'tomekkorbak/detoxify-pile-chunk3-1800000-1850000', 'tomekkorbak/detoxify-pile-chunk3-1850000-1900000', 'tomekkorbak/detoxify-pile-chunk3-1900000-1950000'], 'is_split_by_sentences': True}, 'generation': {'force_call_on': [25354], 'metrics_configs': [{}, {'n': 1}, {'n': 2}, {'n': 5}], 'scenario_configs': [{'generate_kwargs': {'do_sample': True, 'max_length': 128, 'min_length': 10, 'temperature': 0.7, 'top_k': 0, 'top_p': 0.9}, 'name': 'unconditional', 'num_samples': 2048}, {'generate_kwargs': {'do_sample': True, 'max_length': 128, 'min_length': 10, 'temperature': 0.7, 'top_k': 0, 'top_p': 0.9}, 'name': 'challenging_rtp', 'num_samples': 2048, 'prompts_path': 'resources/challenging_rtp.jsonl'}], 'scorer_config': {'device': 'cuda:0'}}, 'kl_gpt3_callback': {'force_call_on': [25354], 'max_tokens': 64, 'num_samples': 4096}, 'model': {'from_scratch': True, 'gpt2_config_kwargs': {'reorder_and_upcast_attn': True, 'scale_attn_by': True}, 'path_or_name': 'gpt2'}, 'objective': {'alpha': 1, 'name': 'Unlikelihood', 'score_threshold': 0.00078}, 'tokenizer': {'path_or_name': 'gpt2'}, 'training': {'dataloader_num_workers': 0, 'effective_batch_size': 64, 'evaluation_strategy': 'no', 'fp16': True, 'hub_model_id': 'reverent_franklin', 'hub_strategy': 'all_checkpoints', 'learning_rate': 0.0005, 'logging_first_step': True, 'logging_steps': 1, 'num_tokens': 3300000000, 'output_dir': 'training_output104340', 'per_device_train_batch_size': 16, 'push_to_hub': True, 'remove_unused_columns': False, 'save_steps': 25354, 'save_strategy': 'steps', 'seed': 42, 'warmup_ratio': 0.01, 'weight_decay': 0.1}} # Wandb URL: https://wandb.ai/tomekkorbak/apo/runs/3csuo5ov
jeapaul/wav2vec2-large-xlsr-53-torgo-demo-f01-nolm
jeapaul
2022-11-24T21:26:04Z
77
0
transformers
[ "transformers", "pytorch", "wav2vec2", "automatic-speech-recognition", "generated_from_trainer", "license:apache-2.0", "endpoints_compatible", "region:us" ]
automatic-speech-recognition
2022-11-23T23:46:08Z
--- license: apache-2.0 tags: - generated_from_trainer model-index: - name: wav2vec2-large-xlsr-53-torgo-demo-f01-nolm results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # wav2vec2-large-xlsr-53-torgo-demo-f01-nolm This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on the None dataset. It achieves the following results on the evaluation set: - Loss: 0.0153 - Wer: 0.4756 ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 0.0001 - train_batch_size: 8 - eval_batch_size: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_steps: 500 - num_epochs: 30 - mixed_precision_training: Native AMP ### Training results | Training Loss | Epoch | Step | Validation Loss | Wer | |:-------------:|:-----:|:-----:|:---------------:|:------:| | 3.4166 | 0.81 | 500 | 4.5019 | 1.0 | | 3.1088 | 1.62 | 1000 | 3.0459 | 1.0 | | 2.8249 | 2.44 | 1500 | 3.0850 | 1.0 | | 2.625 | 3.25 | 2000 | 2.6827 | 1.3656 | | 1.9816 | 4.06 | 2500 | 1.6636 | 1.3701 | | 1.3036 | 4.87 | 3000 | 0.9710 | 1.2504 | | 0.9862 | 5.68 | 3500 | 0.6023 | 1.0519 | | 0.7012 | 6.49 | 4000 | 0.4404 | 0.9342 | | 0.6102 | 7.31 | 4500 | 0.3297 | 0.8491 | | 0.5463 | 8.12 | 5000 | 0.2403 | 0.7773 | | 0.4897 | 8.93 | 5500 | 0.1907 | 0.7335 | | 0.4687 | 9.74 | 6000 | 0.1721 | 0.7095 | | 0.41 | 10.55 | 6500 | 0.1382 | 0.6851 | | 0.3277 | 11.36 | 7000 | 0.1189 | 0.6598 | | 0.3182 | 12.18 | 7500 | 0.1040 | 0.6372 | | 0.3279 | 12.99 | 8000 | 0.0961 | 0.6274 | | 0.2735 | 13.8 | 8500 | 0.0806 | 0.5880 | | 0.3153 | 14.61 | 9000 | 0.0821 | 0.5748 | | 0.251 | 15.42 | 9500 | 0.0633 | 0.5437 | | 0.2 | 16.23 | 10000 | 0.0534 | 0.5316 | | 0.2134 | 17.05 | 10500 | 0.0475 | 0.5195 | | 0.1727 | 17.86 | 11000 | 0.0435 | 0.5146 | | 0.2143 | 18.67 | 11500 | 0.0406 | 0.5072 | | 0.1679 | 19.48 | 12000 | 0.0386 | 0.5057 | | 0.1836 | 20.29 | 12500 | 0.0359 | 0.4984 | | 0.1542 | 21.1 | 13000 | 0.0284 | 0.4914 | | 0.1672 | 21.92 | 13500 | 0.0289 | 0.4884 | | 0.1526 | 22.73 | 14000 | 0.0256 | 0.4867 | | 0.1263 | 23.54 | 14500 | 0.0247 | 0.4871 | | 0.133 | 24.35 | 15000 | 0.0194 | 0.4816 | | 0.1005 | 25.16 | 15500 | 0.0190 | 0.4798 | | 0.1372 | 25.97 | 16000 | 0.0172 | 0.4786 | | 0.1126 | 26.79 | 16500 | 0.0177 | 0.4773 | | 0.0929 | 27.6 | 17000 | 0.0173 | 0.4775 | | 0.1069 | 28.41 | 17500 | 0.0164 | 0.4773 | | 0.0932 | 29.22 | 18000 | 0.0153 | 0.4756 | ### Framework versions - Transformers 4.23.1 - Pytorch 1.12.1+cu113 - Datasets 2.0.0 - Tokenizers 0.13.2
tlttl/tluo_xml_roberta_base_amazon_review_sentiment_v4
tlttl
2022-11-24T21:00:28Z
105
0
transformers
[ "transformers", "pytorch", "tensorboard", "xlm-roberta", "text-classification", "generated_from_trainer", "license:mit", "autotrain_compatible", "endpoints_compatible", "region:us" ]
text-classification
2022-11-24T06:17:18Z
--- license: mit tags: - generated_from_trainer metrics: - accuracy model-index: - name: tluo_xml_roberta_base_amazon_review_sentiment_v4 results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # tluo_xml_roberta_base_amazon_review_sentiment_v4 This model is a fine-tuned version of [xlm-roberta-base](https://huggingface.co/xlm-roberta-base) on an unknown dataset. It achieves the following results on the evaluation set: - Loss: 0.9589 - Accuracy: 0.6137 ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 1.5745609276104923e-05 - train_batch_size: 4 - eval_batch_size: 4 - seed: 25 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 3 - mixed_precision_training: Native AMP ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | |:-------------:|:-----:|:-----:|:---------------:|:--------:| | 1.1074 | 0.17 | 5000 | 1.0468 | 0.5493 | | 1.0461 | 0.33 | 10000 | 1.0222 | 0.558 | | 1.0245 | 0.5 | 15000 | 0.9776 | 0.5793 | | 0.9876 | 0.67 | 20000 | 1.0327 | 0.571 | | 1.0073 | 0.83 | 25000 | 0.9695 | 0.5807 | | 0.9808 | 1.0 | 30000 | 1.0077 | 0.5857 | | 0.9399 | 1.17 | 35000 | 1.0113 | 0.6043 | | 0.8944 | 1.33 | 40000 | 0.9534 | 0.5947 | | 0.908 | 1.5 | 45000 | 1.0155 | 0.5947 | | 0.9238 | 1.67 | 50000 | 0.9813 | 0.597 | | 0.8699 | 1.83 | 55000 | 0.9646 | 0.6027 | | 0.8829 | 2.0 | 60000 | 0.9589 | 0.6137 | | 0.8214 | 2.17 | 65000 | 1.0378 | 0.6017 | | 0.8052 | 2.33 | 70000 | 1.0080 | 0.6003 | | 0.7807 | 2.5 | 75000 | 1.0830 | 0.604 | | 0.7927 | 2.67 | 80000 | 1.0225 | 0.6007 | | 0.8358 | 2.83 | 85000 | 0.9989 | 0.5997 | | 0.7899 | 3.0 | 90000 | 1.0108 | 0.6023 | ### Framework versions - Transformers 4.24.0 - Pytorch 1.12.1+cu113 - Datasets 2.7.1 - Tokenizers 0.13.2
anirudhkashyap/shreya_sentence_truth_predictor2
anirudhkashyap
2022-11-24T20:57:14Z
101
0
transformers
[ "transformers", "pytorch", "tensorboard", "bert", "text-classification", "generated_from_trainer", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "region:us" ]
text-classification
2022-11-24T20:24:10Z
--- license: apache-2.0 tags: - generated_from_trainer metrics: - accuracy model-index: - name: shreya_sentence_truth_predictor2 results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # shreya_sentence_truth_predictor2 This model is a fine-tuned version of [bert-base-uncased](https://huggingface.co/bert-base-uncased) on an unknown dataset. It achieves the following results on the evaluation set: - Loss: 0.8314 - Accuracy: 0.8915 ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 2e-05 - train_batch_size: 8 - eval_batch_size: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 3 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:--------:| | 0.0919 | 1.0 | 875 | 0.6681 | 0.8975 | | 0.0483 | 2.0 | 1750 | 0.8296 | 0.8885 | | 0.04 | 3.0 | 2625 | 0.8314 | 0.8915 | ### Framework versions - Transformers 4.24.0 - Pytorch 1.12.1+cu113 - Tokenizers 0.13.2
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-2-parent
research-backup
2022-11-24T20:37:09Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T20:35:57Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-2-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.6858134920634921 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.40641711229946526 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4065281899109792 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.6520289049471929 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.814 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4298245614035088 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4375 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8256742504143438 - name: F1 (macro) type: f1_macro value: 0.7997907403301278 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8183098591549296 - name: F1 (macro) type: f1_macro value: 0.5751411013000909 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6413867822318526 - name: F1 (macro) type: f1_macro value: 0.6303562723873062 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8593586979202894 - name: F1 (macro) type: f1_macro value: 0.6700136749243296 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8583516139141335 - name: F1 (macro) type: f1_macro value: 0.8571706539074961 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-2-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-2-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.40641711229946526 - Accuracy on SAT: 0.4065281899109792 - Accuracy on BATS: 0.6520289049471929 - Accuracy on U2: 0.4298245614035088 - Accuracy on U4: 0.4375 - Accuracy on Google: 0.814 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-2-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8256742504143438 - Micro F1 score on CogALexV: 0.8183098591549296 - Micro F1 score on EVALution: 0.6413867822318526 - Micro F1 score on K&H+N: 0.8593586979202894 - Micro F1 score on ROOT09: 0.8583516139141335 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-2-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.6858134920634921 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-2-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 10 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-2-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-0-parent
research-backup
2022-11-24T20:33:40Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T20:32:38Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-0-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.6966865079365079 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.40641711229946526 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4065281899109792 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.603112840466926 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.764 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.40350877192982454 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.40046296296296297 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.854452312791924 - name: F1 (macro) type: f1_macro value: 0.8335614299635147 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7906103286384977 - name: F1 (macro) type: f1_macro value: 0.46109731160888645 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6310942578548212 - name: F1 (macro) type: f1_macro value: 0.6269464689917221 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9136815747374278 - name: F1 (macro) type: f1_macro value: 0.7987746285783167 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8260733312441241 - name: F1 (macro) type: f1_macro value: 0.8284737550173958 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-0-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-0-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.40641711229946526 - Accuracy on SAT: 0.4065281899109792 - Accuracy on BATS: 0.603112840466926 - Accuracy on U2: 0.40350877192982454 - Accuracy on U4: 0.40046296296296297 - Accuracy on Google: 0.764 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-0-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.854452312791924 - Micro F1 score on CogALexV: 0.7906103286384977 - Micro F1 score on EVALution: 0.6310942578548212 - Micro F1 score on K&H+N: 0.9136815747374278 - Micro F1 score on ROOT09: 0.8260733312441241 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-0-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.6966865079365079 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-0-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 2 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-0-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-parent
research-backup
2022-11-24T20:32:03Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T20:30:56Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.6386706349206349 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.32887700534759357 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3323442136498516 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.44858254585881047 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.474 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.34210526315789475 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.32407407407407407 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.776254331776405 - name: F1 (macro) type: f1_macro value: 0.743956464512583 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.753755868544601 - name: F1 (macro) type: f1_macro value: 0.32771443150653473 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.4284940411700975 - name: F1 (macro) type: f1_macro value: 0.2369468752324584 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8608193642623635 - name: F1 (macro) type: f1_macro value: 0.6343956091240253 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7643371983704167 - name: F1 (macro) type: f1_macro value: 0.7154904519234845 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.32887700534759357 - Accuracy on SAT: 0.3323442136498516 - Accuracy on BATS: 0.44858254585881047 - Accuracy on U2: 0.34210526315789475 - Accuracy on U4: 0.32407407407407407 - Accuracy on Google: 0.474 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.776254331776405 - Micro F1 score on CogALexV: 0.753755868544601 - Micro F1 score on EVALution: 0.4284940411700975 - Micro F1 score on K&H+N: 0.8608193642623635 - Micro F1 score on ROOT09: 0.7643371983704167 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.6386706349206349 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 10 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-1-parent
research-backup
2022-11-24T20:30:25Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T20:29:17Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-1-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7567460317460317 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3449197860962567 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.35014836795252224 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4252362423568649 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.59 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.35964912280701755 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.33796296296296297 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8695193611571493 - name: F1 (macro) type: f1_macro value: 0.8568674733966278 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7643192488262912 - name: F1 (macro) type: f1_macro value: 0.48540339382722264 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.5525460455037919 - name: F1 (macro) type: f1_macro value: 0.4851738190720077 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9296793489601447 - name: F1 (macro) type: f1_macro value: 0.8229079543242852 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8712002507051081 - name: F1 (macro) type: f1_macro value: 0.8695492693223117 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-1-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-1-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.3449197860962567 - Accuracy on SAT: 0.35014836795252224 - Accuracy on BATS: 0.4252362423568649 - Accuracy on U2: 0.35964912280701755 - Accuracy on U4: 0.33796296296296297 - Accuracy on Google: 0.59 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-1-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8695193611571493 - Micro F1 score on CogALexV: 0.7643192488262912 - Micro F1 score on EVALution: 0.5525460455037919 - Micro F1 score on K&H+N: 0.9296793489601447 - Micro F1 score on ROOT09: 0.8712002507051081 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-1-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7567460317460317 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-1-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 6 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-1-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-0-parent
research-backup
2022-11-24T20:28:45Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T20:27:35Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-0-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7053968253968254 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3449197860962567 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.34421364985163205 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5130628126737076 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.572 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.39035087719298245 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3819444444444444 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.832906433629652 - name: F1 (macro) type: f1_macro value: 0.8065752966322005 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7967136150234742 - name: F1 (macro) type: f1_macro value: 0.5088521823294603 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.5509209100758397 - name: F1 (macro) type: f1_macro value: 0.4913066970113474 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9061000208666621 - name: F1 (macro) type: f1_macro value: 0.7609290356933636 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8398621121905359 - name: F1 (macro) type: f1_macro value: 0.8229245515993938 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-0-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-0-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.3449197860962567 - Accuracy on SAT: 0.34421364985163205 - Accuracy on BATS: 0.5130628126737076 - Accuracy on U2: 0.39035087719298245 - Accuracy on U4: 0.3819444444444444 - Accuracy on Google: 0.572 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-0-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.832906433629652 - Micro F1 score on CogALexV: 0.7967136150234742 - Micro F1 score on EVALution: 0.5509209100758397 - Micro F1 score on K&H+N: 0.9061000208666621 - Micro F1 score on ROOT09: 0.8398621121905359 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-0-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7053968253968254 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-0-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 2 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-0-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-parent
research-backup
2022-11-24T20:21:46Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T20:20:40Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.6884126984126984 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.35561497326203206 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.36795252225519287 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3952195664257921 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.468 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3333333333333333 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.31712962962962965 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8853397619406358 - name: F1 (macro) type: f1_macro value: 0.8737515262680274 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8082159624413146 - name: F1 (macro) type: f1_macro value: 0.5667219243609135 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6110509209100758 - name: F1 (macro) type: f1_macro value: 0.584000376416088 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9426166794185157 - name: F1 (macro) type: f1_macro value: 0.8355305513810347 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8727671576308367 - name: F1 (macro) type: f1_macro value: 0.8723842336449569 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.35561497326203206 - Accuracy on SAT: 0.36795252225519287 - Accuracy on BATS: 0.3952195664257921 - Accuracy on U2: 0.3333333333333333 - Accuracy on U4: 0.31712962962962965 - Accuracy on Google: 0.468 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8853397619406358 - Micro F1 score on CogALexV: 0.8082159624413146 - Micro F1 score on EVALution: 0.6110509209100758 - Micro F1 score on K&H+N: 0.9426166794185157 - Micro F1 score on ROOT09: 0.8727671576308367 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.6884126984126984 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-0-parent
research-backup
2022-11-24T20:18:10Z
107
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T20:16:58Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-0-parent results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8158333333333333 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3850267379679144 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3857566765578635 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4452473596442468 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.462 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.34649122807017546 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3263888888888889 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8225101702576465 - name: F1 (macro) type: f1_macro value: 0.8034809428019279 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7518779342723004 - name: F1 (macro) type: f1_macro value: 0.33329299540004015 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.5411700975081257 - name: F1 (macro) type: f1_macro value: 0.4639671585144264 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8478820338039924 - name: F1 (macro) type: f1_macro value: 0.6768848309180341 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8307740520213099 - name: F1 (macro) type: f1_macro value: 0.8235960440777698 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-0-parent RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-0-parent/raw/main/analogy.json)): - Accuracy on SAT (full): 0.3850267379679144 - Accuracy on SAT: 0.3857566765578635 - Accuracy on BATS: 0.4452473596442468 - Accuracy on U2: 0.34649122807017546 - Accuracy on U4: 0.3263888888888889 - Accuracy on Google: 0.462 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-0-parent/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8225101702576465 - Micro F1 score on CogALexV: 0.7518779342723004 - Micro F1 score on EVALution: 0.5411700975081257 - Micro F1 score on K&H+N: 0.8478820338039924 - Micro F1 score on ROOT09: 0.8307740520213099 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-0-parent/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8158333333333333 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-0-parent") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 9 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: parent The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-0-parent/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-2-child
research-backup
2022-11-24T20:16:27Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T20:15:18Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-2-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.6546230158730159 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.31016042780748665 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.32047477744807124 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4558087826570317 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.552 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.33771929824561403 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.36342592592592593 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7158354678318517 - name: F1 (macro) type: f1_macro value: 0.6580573706656033 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7342723004694836 - name: F1 (macro) type: f1_macro value: 0.2592182558244037 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.4431202600216685 - name: F1 (macro) type: f1_macro value: 0.2667711261353617 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.865201363288586 - name: F1 (macro) type: f1_macro value: 0.6765044508427398 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.727984957693513 - name: F1 (macro) type: f1_macro value: 0.6461380162719604 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-2-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-2-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.31016042780748665 - Accuracy on SAT: 0.32047477744807124 - Accuracy on BATS: 0.4558087826570317 - Accuracy on U2: 0.33771929824561403 - Accuracy on U4: 0.36342592592592593 - Accuracy on Google: 0.552 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-2-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.7158354678318517 - Micro F1 score on CogALexV: 0.7342723004694836 - Micro F1 score on EVALution: 0.4431202600216685 - Micro F1 score on K&H+N: 0.865201363288586 - Micro F1 score on ROOT09: 0.727984957693513 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-2-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.6546230158730159 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-2-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 8 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-2-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-1-child
research-backup
2022-11-24T20:14:43Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T20:13:39Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-1-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7030555555555555 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3850267379679144 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3916913946587537 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5269594219010562 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.6 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.37280701754385964 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.40046296296296297 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.835317161368088 - name: F1 (macro) type: f1_macro value: 0.8283163898192295 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7708920187793428 - name: F1 (macro) type: f1_macro value: 0.40683683267154375 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6040086673889491 - name: F1 (macro) type: f1_macro value: 0.562590771697943 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8937886902691798 - name: F1 (macro) type: f1_macro value: 0.7550347133400666 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8448762143528674 - name: F1 (macro) type: f1_macro value: 0.8407765599818559 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-1-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-1-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.3850267379679144 - Accuracy on SAT: 0.3916913946587537 - Accuracy on BATS: 0.5269594219010562 - Accuracy on U2: 0.37280701754385964 - Accuracy on U4: 0.40046296296296297 - Accuracy on Google: 0.6 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-1-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.835317161368088 - Micro F1 score on CogALexV: 0.7708920187793428 - Micro F1 score on EVALution: 0.6040086673889491 - Micro F1 score on K&H+N: 0.8937886902691798 - Micro F1 score on ROOT09: 0.8448762143528674 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-1-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7030555555555555 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-1-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 8 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-1-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-0-child
research-backup
2022-11-24T20:13:07Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T20:11:58Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-0-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.38442460317460314 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.23529411764705882 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.2314540059347181 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3229571984435798 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.384 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3157894736842105 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.2847222222222222 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6522525237306012 - name: F1 (macro) type: f1_macro value: 0.616560269476982 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7183098591549296 - name: F1 (macro) type: f1_macro value: 0.16833503884438658 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.42632719393282775 - name: F1 (macro) type: f1_macro value: 0.28678399596569476 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8141475968560896 - name: F1 (macro) type: f1_macro value: 0.6286243048790003 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6552804763397054 - name: F1 (macro) type: f1_macro value: 0.5562839421136045 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-0-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-0-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.23529411764705882 - Accuracy on SAT: 0.2314540059347181 - Accuracy on BATS: 0.3229571984435798 - Accuracy on U2: 0.3157894736842105 - Accuracy on U4: 0.2847222222222222 - Accuracy on Google: 0.384 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-0-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.6522525237306012 - Micro F1 score on CogALexV: 0.7183098591549296 - Micro F1 score on EVALution: 0.42632719393282775 - Micro F1 score on K&H+N: 0.8141475968560896 - Micro F1 score on ROOT09: 0.6552804763397054 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-0-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.38442460317460314 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-0-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 2 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-e-triplet-0-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-1-child
research-backup
2022-11-24T20:09:44Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T20:08:36Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-1-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7853174603174603 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4197860962566845 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.42433234421364985 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5619788771539744 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.744 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.43859649122807015 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4351851851851852 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8716287479282808 - name: F1 (macro) type: f1_macro value: 0.8587076704516358 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8446009389671362 - name: F1 (macro) type: f1_macro value: 0.623737694543065 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.676056338028169 - name: F1 (macro) type: f1_macro value: 0.6565825451079605 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9460944564234541 - name: F1 (macro) type: f1_macro value: 0.8418727845254599 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8799749294891883 - name: F1 (macro) type: f1_macro value: 0.8734406634484763 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-1-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-1-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.4197860962566845 - Accuracy on SAT: 0.42433234421364985 - Accuracy on BATS: 0.5619788771539744 - Accuracy on U2: 0.43859649122807015 - Accuracy on U4: 0.4351851851851852 - Accuracy on Google: 0.744 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-1-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8716287479282808 - Micro F1 score on CogALexV: 0.8446009389671362 - Micro F1 score on EVALution: 0.676056338028169 - Micro F1 score on K&H+N: 0.9460944564234541 - Micro F1 score on ROOT09: 0.8799749294891883 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-1-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7853174603174603 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-1-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 9 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-d-triplet-1-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-child
research-backup
2022-11-24T20:06:09Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T20:05:02Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.5337896825396825 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.29411764705882354 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.2878338278931751 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3885491939966648 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.494 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3815789473684211 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.35185185185185186 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7629953292150068 - name: F1 (macro) type: f1_macro value: 0.7124326872331369 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.718075117370892 - name: F1 (macro) type: f1_macro value: 0.16718130892198388 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.4442036836403034 - name: F1 (macro) type: f1_macro value: 0.28765674250197426 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8604715865618696 - name: F1 (macro) type: f1_macro value: 0.6404791041645369 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7580695706675025 - name: F1 (macro) type: f1_macro value: 0.7075134842767926 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.29411764705882354 - Accuracy on SAT: 0.2878338278931751 - Accuracy on BATS: 0.3885491939966648 - Accuracy on U2: 0.3815789473684211 - Accuracy on U4: 0.35185185185185186 - Accuracy on Google: 0.494 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.7629953292150068 - Micro F1 score on CogALexV: 0.718075117370892 - Micro F1 score on EVALution: 0.4442036836403034 - Micro F1 score on K&H+N: 0.8604715865618696 - Micro F1 score on ROOT09: 0.7580695706675025 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.5337896825396825 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-2-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-1-child
research-backup
2022-11-24T20:04:27Z
104
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T20:03:22Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-1-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7567460317460317 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3449197860962567 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.35014836795252224 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4252362423568649 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.59 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.35964912280701755 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.33796296296296297 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8695193611571493 - name: F1 (macro) type: f1_macro value: 0.8568674733966278 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7643192488262912 - name: F1 (macro) type: f1_macro value: 0.48540339382722264 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.5525460455037919 - name: F1 (macro) type: f1_macro value: 0.4851738190720077 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9296793489601447 - name: F1 (macro) type: f1_macro value: 0.8229079543242852 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8712002507051081 - name: F1 (macro) type: f1_macro value: 0.8695492693223117 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-1-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-1-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.3449197860962567 - Accuracy on SAT: 0.35014836795252224 - Accuracy on BATS: 0.4252362423568649 - Accuracy on U2: 0.35964912280701755 - Accuracy on U4: 0.33796296296296297 - Accuracy on Google: 0.59 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-1-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8695193611571493 - Micro F1 score on CogALexV: 0.7643192488262912 - Micro F1 score on EVALution: 0.5525460455037919 - Micro F1 score on K&H+N: 0.9296793489601447 - Micro F1 score on ROOT09: 0.8712002507051081 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-1-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7567460317460317 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-1-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 6 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 1 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-1-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-0-child
research-backup
2022-11-24T20:02:48Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T20:01:43Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-0-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.7053968253968254 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3449197860962567 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.34421364985163205 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.5130628126737076 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.572 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.39035087719298245 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3819444444444444 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.832906433629652 - name: F1 (macro) type: f1_macro value: 0.8065752966322005 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7967136150234742 - name: F1 (macro) type: f1_macro value: 0.5088521823294603 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.5509209100758397 - name: F1 (macro) type: f1_macro value: 0.4913066970113474 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9061000208666621 - name: F1 (macro) type: f1_macro value: 0.7609290356933636 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8398621121905359 - name: F1 (macro) type: f1_macro value: 0.8229245515993938 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-0-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-0-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.3449197860962567 - Accuracy on SAT: 0.34421364985163205 - Accuracy on BATS: 0.5130628126737076 - Accuracy on U2: 0.39035087719298245 - Accuracy on U4: 0.3819444444444444 - Accuracy on Google: 0.572 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-0-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.832906433629652 - Micro F1 score on CogALexV: 0.7967136150234742 - Micro F1 score on EVALution: 0.5509209100758397 - Micro F1 score on K&H+N: 0.9061000208666621 - Micro F1 score on ROOT09: 0.8398621121905359 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-0-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.7053968253968254 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-0-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 2 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-c-triplet-0-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-0-child
research-backup
2022-11-24T19:57:53Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T19:56:51Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-0-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8059920634920635 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.32620320855614976 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3323442136498516 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3507504168982768 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.382 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.33771929824561403 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.35185185185185186 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7999095977098086 - name: F1 (macro) type: f1_macro value: 0.7684883952780684 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7530516431924884 - name: F1 (macro) type: f1_macro value: 0.34938339909910743 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.4739978331527627 - name: F1 (macro) type: f1_macro value: 0.3568206512908552 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8818946929122905 - name: F1 (macro) type: f1_macro value: 0.7088102566206993 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7790661234722657 - name: F1 (macro) type: f1_macro value: 0.759592502713887 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-0-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-0-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.32620320855614976 - Accuracy on SAT: 0.3323442136498516 - Accuracy on BATS: 0.3507504168982768 - Accuracy on U2: 0.33771929824561403 - Accuracy on U4: 0.35185185185185186 - Accuracy on Google: 0.382 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-0-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.7999095977098086 - Micro F1 score on CogALexV: 0.7530516431924884 - Micro F1 score on EVALution: 0.4739978331527627 - Micro F1 score on K&H+N: 0.8818946929122905 - Micro F1 score on ROOT09: 0.7790661234722657 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-0-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8059920634920635 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-0-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 9 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-b-triplet-0-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-child
research-backup
2022-11-24T19:56:13Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T19:55:10Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.6884126984126984 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.35561497326203206 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.36795252225519287 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3952195664257921 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.468 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3333333333333333 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.31712962962962965 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8853397619406358 - name: F1 (macro) type: f1_macro value: 0.8737515262680274 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8082159624413146 - name: F1 (macro) type: f1_macro value: 0.5667219243609135 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.6110509209100758 - name: F1 (macro) type: f1_macro value: 0.584000376416088 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.9426166794185157 - name: F1 (macro) type: f1_macro value: 0.8355305513810347 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8727671576308367 - name: F1 (macro) type: f1_macro value: 0.8723842336449569 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.35561497326203206 - Accuracy on SAT: 0.36795252225519287 - Accuracy on BATS: 0.3952195664257921 - Accuracy on U2: 0.3333333333333333 - Accuracy on U4: 0.31712962962962965 - Accuracy on Google: 0.468 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8853397619406358 - Micro F1 score on CogALexV: 0.8082159624413146 - Micro F1 score on EVALution: 0.6110509209100758 - Micro F1 score on K&H+N: 0.9426166794185157 - Micro F1 score on ROOT09: 0.8727671576308367 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.6884126984126984 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 5 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 2 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-2-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
research-backup/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-0-child
research-backup
2022-11-24T19:53:00Z
103
0
transformers
[ "transformers", "pytorch", "roberta", "feature-extraction", "dataset:relbert/semeval2012_relational_similarity_v6", "model-index", "text-embeddings-inference", "endpoints_compatible", "region:us" ]
feature-extraction
2022-11-24T19:51:50Z
--- datasets: - relbert/semeval2012_relational_similarity_v6 model-index: - name: relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-0-child results: - task: name: Relation Mapping type: sorting-task dataset: name: Relation Mapping args: relbert/relation_mapping type: relation-mapping metrics: - name: Accuracy type: accuracy value: 0.8158333333333333 - task: name: Analogy Questions (SAT full) type: multiple-choice-qa dataset: name: SAT full args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3850267379679144 - task: name: Analogy Questions (SAT) type: multiple-choice-qa dataset: name: SAT args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3857566765578635 - task: name: Analogy Questions (BATS) type: multiple-choice-qa dataset: name: BATS args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.4452473596442468 - task: name: Analogy Questions (Google) type: multiple-choice-qa dataset: name: Google args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.462 - task: name: Analogy Questions (U2) type: multiple-choice-qa dataset: name: U2 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.34649122807017546 - task: name: Analogy Questions (U4) type: multiple-choice-qa dataset: name: U4 args: relbert/analogy_questions type: analogy-questions metrics: - name: Accuracy type: accuracy value: 0.3263888888888889 - task: name: Lexical Relation Classification (BLESS) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8225101702576465 - name: F1 (macro) type: f1_macro value: 0.8034809428019279 - task: name: Lexical Relation Classification (CogALexV) type: classification dataset: name: CogALexV args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.7518779342723004 - name: F1 (macro) type: f1_macro value: 0.33329299540004015 - task: name: Lexical Relation Classification (EVALution) type: classification dataset: name: BLESS args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.5411700975081257 - name: F1 (macro) type: f1_macro value: 0.4639671585144264 - task: name: Lexical Relation Classification (K&H+N) type: classification dataset: name: K&H+N args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8478820338039924 - name: F1 (macro) type: f1_macro value: 0.6768848309180341 - task: name: Lexical Relation Classification (ROOT09) type: classification dataset: name: ROOT09 args: relbert/lexical_relation_classification type: relation-classification metrics: - name: F1 type: f1 value: 0.8307740520213099 - name: F1 (macro) type: f1_macro value: 0.8235960440777698 --- # relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-0-child RelBERT fine-tuned from [roberta-base](https://huggingface.co/roberta-base) on [relbert/semeval2012_relational_similarity_v6](https://huggingface.co/datasets/relbert/semeval2012_relational_similarity_v6). Fine-tuning is done via [RelBERT](https://github.com/asahi417/relbert) library (see the repository for more detail). It achieves the following results on the relation understanding tasks: - Analogy Question ([dataset](https://huggingface.co/datasets/relbert/analogy_questions), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-0-child/raw/main/analogy.json)): - Accuracy on SAT (full): 0.3850267379679144 - Accuracy on SAT: 0.3857566765578635 - Accuracy on BATS: 0.4452473596442468 - Accuracy on U2: 0.34649122807017546 - Accuracy on U4: 0.3263888888888889 - Accuracy on Google: 0.462 - Lexical Relation Classification ([dataset](https://huggingface.co/datasets/relbert/lexical_relation_classification), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-0-child/raw/main/classification.json)): - Micro F1 score on BLESS: 0.8225101702576465 - Micro F1 score on CogALexV: 0.7518779342723004 - Micro F1 score on EVALution: 0.5411700975081257 - Micro F1 score on K&H+N: 0.8478820338039924 - Micro F1 score on ROOT09: 0.8307740520213099 - Relation Mapping ([dataset](https://huggingface.co/datasets/relbert/relation_mapping), [full result](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-0-child/raw/main/relation_mapping.json)): - Accuracy on Relation Mapping: 0.8158333333333333 ### Usage This model can be used through the [relbert library](https://github.com/asahi417/relbert). Install the library via pip ```shell pip install relbert ``` and activate model as below. ```python from relbert import RelBERT model = RelBERT("relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-0-child") vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, ) ``` ### Training hyperparameters The following hyperparameters were used during training: - model: roberta-base - max_length: 64 - mode: mask - data: relbert/semeval2012_relational_similarity_v6 - split: train - split_eval: validation - template_mode: manual - loss_function: triplet - classification_loss: False - temperature_nce_constant: 0.05 - temperature_nce_rank: {'min': 0.01, 'max': 0.05, 'type': 'linear'} - epoch: 9 - batch: 128 - lr: 5e-06 - lr_decay: False - lr_warmup: 1 - weight_decay: 0 - random_seed: 0 - exclude_relation: None - n_sample: 320 - gradient_accumulation: 8 - relation_level: None - data_level: child The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/relbert/relbert-roberta-base-semeval2012-v6-mask-prompt-a-triplet-0-child/raw/main/trainer_config.json). ### Reference If you use any resource from RelBERT, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/). ``` @inproceedings{ushio-etal-2021-distilling-relation-embeddings, title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels", author = "Ushio, Asahi and Schockaert, Steven and Camacho-Collados, Jose", booktitle = "EMNLP 2021", year = "2021", address = "Online", publisher = "Association for Computational Linguistics", } ```
reallygoodtechdeals/Bingocat-ai-Dialo-GPT-medium
reallygoodtechdeals
2022-11-24T19:42:26Z
112
0
transformers
[ "transformers", "pytorch", "gpt2", "text-generation", "conversational", "license:wtfpl", "autotrain_compatible", "text-generation-inference", "endpoints_compatible", "region:us" ]
text-generation
2022-11-24T08:06:30Z
--- license: wtfpl tags: - conversational ---
jeapaul/wav2vec2-large-xlsr-53-torgo-demo-m03-nolm
jeapaul
2022-11-24T18:51:11Z
79
0
transformers
[ "transformers", "pytorch", "wav2vec2", "automatic-speech-recognition", "generated_from_trainer", "license:apache-2.0", "endpoints_compatible", "region:us" ]
automatic-speech-recognition
2022-11-23T07:56:11Z
--- license: apache-2.0 tags: - generated_from_trainer model-index: - name: wav2vec2-large-xlsr-53-torgo-demo-m03-nolm results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # wav2vec2-large-xlsr-53-torgo-demo-m03-nolm This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on the None dataset. It achieves the following results on the evaluation set: - Loss: 0.0191 - Wer: 0.4978 ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 0.0001 - train_batch_size: 8 - eval_batch_size: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_steps: 500 - num_epochs: 30 - mixed_precision_training: Native AMP ### Training results | Training Loss | Epoch | Step | Validation Loss | Wer | |:-------------:|:-----:|:-----:|:---------------:|:------:| | 3.3416 | 0.92 | 500 | 4.1988 | 1.0 | | 3.0162 | 1.85 | 1000 | 4.2155 | 1.0 | | 2.8505 | 2.77 | 1500 | 3.3644 | 1.0 | | 2.5676 | 3.69 | 2000 | 2.7188 | 1.0 | | 1.9316 | 4.61 | 2500 | 1.6584 | 1.3807 | | 1.4131 | 5.54 | 3000 | 0.9672 | 1.2920 | | 1.0554 | 6.46 | 3500 | 0.6630 | 1.1193 | | 0.7956 | 7.38 | 4000 | 0.4932 | 1.0106 | | 0.6453 | 8.3 | 4500 | 0.3441 | 0.9024 | | 0.5637 | 9.23 | 5000 | 0.2947 | 0.8437 | | 0.5264 | 10.15 | 5500 | 0.2328 | 0.7915 | | 0.4551 | 11.07 | 6000 | 0.1766 | 0.7410 | | 0.4426 | 11.99 | 6500 | 0.1511 | 0.6978 | | 0.3635 | 12.92 | 7000 | 0.1229 | 0.6854 | | 0.3336 | 13.84 | 7500 | 0.1077 | 0.6329 | | 0.315 | 14.76 | 8000 | 0.1073 | 0.6238 | | 0.2468 | 15.68 | 8500 | 0.0775 | 0.5938 | | 0.2589 | 16.61 | 9000 | 0.0809 | 0.5777 | | 0.2209 | 17.53 | 9500 | 0.0655 | 0.5585 | | 0.2392 | 18.45 | 10000 | 0.0592 | 0.5503 | | 0.2423 | 19.37 | 10500 | 0.0521 | 0.5387 | | 0.2347 | 20.3 | 11000 | 0.0428 | 0.5318 | | 0.1834 | 21.22 | 11500 | 0.0370 | 0.5243 | | 0.1495 | 22.14 | 12000 | 0.0359 | 0.5138 | | 0.1676 | 23.06 | 12500 | 0.0305 | 0.5111 | | 0.1653 | 23.99 | 13000 | 0.0246 | 0.5041 | | 0.1624 | 24.91 | 13500 | 0.0259 | 0.5065 | | 0.187 | 25.83 | 14000 | 0.0213 | 0.5008 | | 0.1452 | 26.75 | 14500 | 0.0223 | 0.5012 | | 0.1555 | 27.68 | 15000 | 0.0208 | 0.5003 | | 0.1134 | 28.6 | 15500 | 0.0196 | 0.4990 | | 0.1845 | 29.52 | 16000 | 0.0191 | 0.4978 | ### Framework versions - Transformers 4.23.1 - Pytorch 1.12.1+cu113 - Datasets 2.0.0 - Tokenizers 0.13.2
huggingtweets/horse_luvr_47-pacsjam
huggingtweets
2022-11-24T18:35:13Z
113
0
transformers
[ "transformers", "pytorch", "gpt2", "text-generation", "huggingtweets", "en", "autotrain_compatible", "text-generation-inference", "endpoints_compatible", "region:us" ]
text-generation
2022-11-24T18:34:02Z
--- language: en thumbnail: http://www.huggingtweets.com/horse_luvr_47-pacsjam/1669314909523/predictions.png tags: - huggingtweets widget: - text: "My dream is" --- <div class="inline-flex flex-col" style="line-height: 1.5;"> <div class="flex"> <div style="display:inherit; margin-left: 4px; margin-right: 4px; width: 92px; height:92px; border-radius: 50%; background-size: cover; background-image: url(&#39;https://pbs.twimg.com/profile_images/1530452243849224192/rp21Xew2_400x400.jpg&#39;)"> </div> <div style="display:inherit; margin-left: 4px; margin-right: 4px; width: 92px; height:92px; border-radius: 50%; background-size: cover; background-image: url(&#39;https://pbs.twimg.com/profile_images/1588101458511024130/9DsQWW2U_400x400.jpg&#39;)"> </div> <div style="display:none; margin-left: 4px; margin-right: 4px; width: 92px; height:92px; border-radius: 50%; background-size: cover; background-image: url(&#39;&#39;)"> </div> </div> <div style="text-align: center; margin-top: 3px; font-size: 16px; font-weight: 800">๐Ÿค– AI CYBORG ๐Ÿค–</div> <div style="text-align: center; font-size: 16px; font-weight: 800">jesse ๐Ÿซ๐Ÿ•๐Ÿ’Œ & adam handler</div> <div style="text-align: center; font-size: 14px;">@horse_luvr_47-pacsjam</div> </div> I was made with [huggingtweets](https://github.com/borisdayma/huggingtweets). Create your own bot based on your favorite user with [the demo](https://colab.research.google.com/github/borisdayma/huggingtweets/blob/master/huggingtweets-demo.ipynb)! ## How does it work? The model uses the following pipeline. ![pipeline](https://github.com/borisdayma/huggingtweets/blob/master/img/pipeline.png?raw=true) To understand how the model was developed, check the [W&B report](https://wandb.ai/wandb/huggingtweets/reports/HuggingTweets-Train-a-Model-to-Generate-Tweets--VmlldzoxMTY5MjI). ## Training data The model was trained on tweets from jesse ๐Ÿซ๐Ÿ•๐Ÿ’Œ & adam handler. | Data | jesse ๐Ÿซ๐Ÿ•๐Ÿ’Œ | adam handler | | --- | --- | --- | | Tweets downloaded | 3043 | 2952 | | Retweets | 772 | 1364 | | Short tweets | 575 | 243 | | Tweets kept | 1696 | 1345 | [Explore the data](https://wandb.ai/wandb/huggingtweets/runs/ess4rcio/artifacts), which is tracked with [W&B artifacts](https://docs.wandb.com/artifacts) at every step of the pipeline. ## Training procedure The model is based on a pre-trained [GPT-2](https://huggingface.co/gpt2) which is fine-tuned on @horse_luvr_47-pacsjam's tweets. Hyperparameters and metrics are recorded in the [W&B training run](https://wandb.ai/wandb/huggingtweets/runs/rc1dxno2) for full transparency and reproducibility. At the end of training, [the final model](https://wandb.ai/wandb/huggingtweets/runs/rc1dxno2/artifacts) is logged and versioned. ## How to use You can use this model directly with a pipeline for text generation: ```python from transformers import pipeline generator = pipeline('text-generation', model='huggingtweets/horse_luvr_47-pacsjam') generator("My dream is", num_return_sequences=5) ``` ## Limitations and bias The model suffers from [the same limitations and bias as GPT-2](https://huggingface.co/gpt2#limitations-and-bias). In addition, the data present in the user's tweets further affects the text generated by the model. ## About *Built by Boris Dayma* [![Follow](https://img.shields.io/twitter/follow/borisdayma?style=social)](https://twitter.com/intent/follow?screen_name=borisdayma) For more details, visit the project repository. [![GitHub stars](https://img.shields.io/github/stars/borisdayma/huggingtweets?style=social)](https://github.com/borisdayma/huggingtweets)
ai-forever/Kandinsky_2.0
ai-forever
2022-11-24T18:16:01Z
0
43
null
[ "Kandinsky", "text-image", "text2image", "diffusion", "latent diffusion", "mCLIP-XLMR", "mT5", "license:apache-2.0", "region:us" ]
null
2022-11-22T14:43:57Z
--- license: apache-2.0 tags: - Kandinsky - text-image - text2image - diffusion - latent diffusion - mCLIP-XLMR - mT5 --- # Kandinsky 2.0 Kandinsky 2.0 โ€” the first multilingual text2image model. [Open In Colab](https://colab.research.google.com/drive/1uPg9KwGZ2hJBl9taGA_3kyKGw12Rh3ij?usp=sharing) [GitHub repository](https://github.com/ai-forever/Kandinsky-2.0) [Habr post](https://habr.com/ru/company/sberbank/blog/701162/) [Demo](https://rudalle.ru/) **UNet size: 1.2B parameters** ![NatallE.png](https://s3.amazonaws.com/moonup/production/uploads/1669132577749-5f91b1208a61a359f44e1851.png) It is a latent diffusion model with two multi-lingual text encoders: * mCLIP-XLMR (560M parameters) * mT5-encoder-small (146M parameters) These encoders and multilingual training datasets unveil the real multilingual text2image generation experience! ![header.png](https://s3.amazonaws.com/moonup/production/uploads/1669132825912-5f91b1208a61a359f44e1851.png) # How to use ```python pip install "git+https://github.com/ai-forever/Kandinsky-2.0.git" from kandinsky2 import get_kandinsky2 model = get_kandinsky2('cuda', task_type='text2img') images = model.generate_text2img('ะบะพัˆะบะฐ ะฒ ะบะพัะผะพัะต', batch_size=4, h=512, w=512, num_steps=75, denoised_type='dynamic_threshold', dynamic_threshold_v=99.5, sampler='ddim_sampler', ddim_eta=0.01, guidance_scale=10) ``` # Authors + Arseniy Shakhmatov: [Github](https://github.com/cene555), [Blog](https://t.me/gradientdip) + Anton Razzhigaev: [Github](https://github.com/razzant), [Blog](https://t.me/abstractDL) + Aleksandr Nikolich: [Github](https://github.com/AlexWortega), [Blog](https://t.me/lovedeathtransformers) + Vladimir Arkhipkin: [Github](https://github.com/oriBetelgeuse) + Igor Pavlov: [Github](https://github.com/boomb0om) + Andrey Kuznetsov: [Github](https://github.com/kuznetsoffandrey) + Denis Dimitrov: [Github](https://github.com/denndimitrov)
su157/t5-small-qg-german-00
su157
2022-11-24T16:40:31Z
103
0
transformers
[ "transformers", "pytorch", "mt5", "text2text-generation", "autotrain_compatible", "endpoints_compatible", "region:us" ]
text2text-generation
2022-11-24T16:29:59Z
This model is a fine-tuned version of google/mt5-small on the GermanQuAD dataset. This is a model for question generation from a text corpus. The following hyperparameters were used during training: - learning_rate: 1e-3 - mini_batch_size: 8 - optimizer: Adam - num_epochs: 4
flamesbob/Minimal_gradient
flamesbob
2022-11-24T16:14:45Z
0
3
null
[ "license:creativeml-openrail-m", "region:us" ]
null
2022-11-24T16:13:48Z
--- license: creativeml-openrail-m ---
popolin52/q-Taxi-v3
popolin52
2022-11-24T16:02:44Z
0
0
null
[ "Taxi-v3", "q-learning", "reinforcement-learning", "custom-implementation", "model-index", "region:us" ]
reinforcement-learning
2022-11-24T15:49:44Z
--- tags: - Taxi-v3 - q-learning - reinforcement-learning - custom-implementation model-index: - name: q-Taxi-v3 results: - task: type: reinforcement-learning name: reinforcement-learning dataset: name: Taxi-v3 type: Taxi-v3 metrics: - type: mean_reward value: 7.56 +/- 2.71 name: mean_reward verified: false --- # **Q-Learning** Agent playing **Taxi-v3** This is a trained model of a **Q-Learning** agent playing **Taxi-v3** . ## Usage ```python model = load_from_hub(repo_id="popolin52/q-Taxi-v3", filename="q-learning.pkl") # Don't forget to check if you need to add additional attributes (is_slippery=False etc) env = gym.make(model["env_id"]) evaluate_agent(env, model["max_steps"], model["n_eval_episodes"], model["qtable"], model["eval_seed"]) ```