model update

README.md

@@ -0,0 +1,135 @@
+---
+datasets:
+- tner/btc
+metrics:
+- f1
+- precision
+- recall
+model-index:
+- name: tner/deberta-v3-large-btc
+  results:
+  - task:
+      name: Token Classification
+      type: token-classification
+    dataset:
+      name: tner/btc
+      type: tner/btc
+      args: tner/btc
+    metrics:
+    - name: F1
+      type: f1
+      value: 0.8399238265934805
+    - name: Precision
+      type: precision
+      value: 0.8237749945067018
+    - name: Recall
+      type: recall
+      value: 0.8567184643510055
+    - name: F1 (macro)
+      type: f1_macro
+      value: 0.7921150390682584
+    - name: Precision (macro)
+      type: precision_macro
+      value: 0.7766126681668878
+    - name: Recall (macro)
+      type: recall_macro
+      value: 0.8103758198218992
+    - name: F1 (entity span)
+      type: f1_entity_span
+      value: 0.9134087599417496
+    - name: Precision (entity span)
+      type: precision_entity_span
+      value: 0.8958470665787739
+    - name: Recall (entity span)
+      type: recall_entity_span
+      value: 0.931672760511883
+
+pipeline_tag: token-classification
+widget:
+- text: "Jacob Collier is a Grammy awarded artist from England."
+  example_title: "NER Example 1"
+---
+# tner/deberta-v3-large-btc
+
+This model is a fine-tuned version of [microsoft/deberta-v3-large](https://huggingface.co/microsoft/deberta-v3-large) on the 
+[tner/btc](https://huggingface.co/datasets/tner/btc) dataset.
+Model fine-tuning is done via [T-NER](https://github.com/asahi417/tner)'s hyper-parameter search (see the repository
+for more detail). It achieves the following results on the test set:
+- F1 (micro): 0.8399238265934805
+- Precision (micro): 0.8237749945067018
+- Recall (micro): 0.8567184643510055
+- F1 (macro): 0.7921150390682584
+- Precision (macro): 0.7766126681668878
+- Recall (macro): 0.8103758198218992
+
+The per-entity breakdown of the F1 score on the test set are below:
+- location: 0.7503949447077408
+- organization: 0.7042372881355932
+- person: 0.9217128843614413 
+
+For F1 scores, the confidence interval is obtained by bootstrap as below:
+- F1 (micro): 
+    - 90%: [0.8283024935970381, 0.8507400882379221]
+    - 95%: [0.8260021524132041, 0.8526162579659953] 
+- F1 (macro): 
+    - 90%: [0.8283024935970381, 0.8507400882379221]
+    - 95%: [0.8260021524132041, 0.8526162579659953] 
+
+Full evaluation can be found at [metric file of NER](https://huggingface.co/tner/deberta-v3-large-btc/raw/main/eval/metric.json) 
+and [metric file of entity span](https://huggingface.co/tner/deberta-v3-large-btc/raw/main/eval/metric_span.json).
+
+### Usage
+This model can be used through the [tner library](https://github.com/asahi417/tner). Install the library via pip   
+```shell
+pip install tner
+```
+and activate model as below.
+```python
+from tner import TransformersNER
+model = TransformersNER("tner/deberta-v3-large-btc")
+model.predict(["Jacob Collier is a Grammy awarded English artist from London"])
+```
+It can be used via transformers library but it is not recommended as CRF layer is not supported at the moment.
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+ - dataset: ['tner/btc']
+ - dataset_split: train
+ - dataset_name: None
+ - local_dataset: None
+ - model: microsoft/deberta-v3-large
+ - crf: True
+ - max_length: 128
+ - epoch: 15
+ - batch_size: 16
+ - lr: 1e-05
+ - random_seed: 42
+ - gradient_accumulation_steps: 8
+ - weight_decay: None
+ - lr_warmup_step_ratio: 0.1
+ - max_grad_norm: None
+
+The full configuration can be found at [fine-tuning parameter file](https://huggingface.co/tner/deberta-v3-large-btc/raw/main/trainer_config.json).
+
+### Reference
+If you use any resource from T-NER, please consider to cite our [paper](https://aclanthology.org/2021.eacl-demos.7/).
+
+```
+
+@inproceedings{ushio-camacho-collados-2021-ner,
+    title = "{T}-{NER}: An All-Round Python Library for Transformer-based Named Entity Recognition",
+    author = "Ushio, Asahi  and
+      Camacho-Collados, Jose",
+    booktitle = "Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: System Demonstrations",
+    month = apr,
+    year = "2021",
+    address = "Online",
+    publisher = "Association for Computational Linguistics",
+    url = "https://aclanthology.org/2021.eacl-demos.7",
+    doi = "10.18653/v1/2021.eacl-demos.7",
+    pages = "53--62",
+    abstract = "Language model (LM) pretraining has led to consistent improvements in many NLP downstream tasks, including named entity recognition (NER). In this paper, we present T-NER (Transformer-based Named Entity Recognition), a Python library for NER LM finetuning. In addition to its practical utility, T-NER facilitates the study and investigation of the cross-domain and cross-lingual generalization ability of LMs finetuned on NER. Our library also provides a web app where users can get model predictions interactively for arbitrary text, which facilitates qualitative model evaluation for non-expert programmers. We show the potential of the library by compiling nine public NER datasets into a unified format and evaluating the cross-domain and cross- lingual performance across the datasets. The results from our initial experiments show that in-domain performance is generally competitive across datasets. However, cross-domain generalization is challenging even with a large pretrained LM, which has nevertheless capacity to learn domain-specific features if fine- tuned on a combined dataset. To facilitate future research, we also release all our LM checkpoints via the Hugging Face model hub.",
+}
+
+```

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "tner_ckpt/btc_deberta_v3_large/best_model",
+  "_name_or_path": "tner_ckpt/btc_deberta_v3_large/model_lzavco/epoch_5",
   "architectures": [
     "DebertaV2ForTokenClassification"
   ],

eval/metric.json

@@ -0,0 +1 @@
+{"micro/f1": 0.8399238265934805, "micro/f1_ci": {"90": [0.8283024935970381, 0.8507400882379221], "95": [0.8260021524132041, 0.8526162579659953]}, "micro/recall": 0.8567184643510055, "micro/precision": 0.8237749945067018, "macro/f1": 0.7921150390682584, "macro/f1_ci": {"90": [0.7778836465333343, 0.8064908270960832], "95": [0.7744530632383323, 0.8088306344266144]}, "macro/recall": 0.8103758198218992, "macro/precision": 0.7766126681668878, "per_entity_metric": {"location": {"f1": 0.7503949447077408, "f1_ci": {"90": [0.722621343484653, 0.7783360088749672], "95": [0.7181196923659109, 0.7831327609287725]}, "precision": 0.753968253968254, "recall": 0.7468553459119497}, "organization": {"f1": 0.7042372881355932, "f1_ci": {"90": [0.680714976595495, 0.727503450051444], "95": [0.6767128243490045, 0.7317512781467873]}, "precision": 0.6543307086614173, "recall": 0.7623853211009174}, "person": {"f1": 0.9217128843614413, "f1_ci": {"90": [0.9119808330613168, 0.9304886792800428], "95": [0.9104814441886399, 0.9319999849049878]}, "precision": 0.9215390418709921, "recall": 0.9218867924528302}}}

eval/metric_span.json

@@ -0,0 +1 @@
+{"micro/f1": 0.9134087599417496, "micro/f1_ci": {"90": [0.9063238930638302, 0.9201311718583651], "95": [0.9051626400610194, 0.9211604895898028]}, "micro/recall": 0.931672760511883, "micro/precision": 0.8958470665787739, "macro/f1": 0.9134087599417496, "macro/f1_ci": {"90": [0.9063238930638302, 0.9201311718583651], "95": [0.9051626400610194, 0.9211604895898028]}, "macro/recall": 0.931672760511883, "macro/precision": 0.8958470665787739}

tokenizer_config.json

@@ -4,7 +4,7 @@
   "do_lower_case": false,
   "eos_token": "[SEP]",
   "mask_token": "[MASK]",
-  "name_or_path": "tner_ckpt/btc_deberta_v3_large/best_model",
+  "name_or_path": "tner_ckpt/btc_deberta_v3_large/model_lzavco/epoch_5",
   "pad_token": "[PAD]",
   "sep_token": "[SEP]",
   "sp_model_kwargs": {},

trainer_config.json

@@ -0,0 +1 @@
+{"dataset": ["tner/btc"], "dataset_split": "train", "dataset_name": null, "local_dataset": null, "model": "microsoft/deberta-v3-large", "crf": true, "max_length": 128, "epoch": 15, "batch_size": 16, "lr": 1e-05, "random_seed": 42, "gradient_accumulation_steps": 8, "weight_decay": null, "lr_warmup_step_ratio": 0.1, "max_grad_norm": null}