metadata

tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:40000
  - loss:TripletLoss
base_model: sentence-transformers/paraphrase-xlm-r-multilingual-v1
widget:
  - source_sentence: ልጆች በመዝናኛ መናፈሻ ውስጥ ፊኛ ይጋልባሉ።
    sentences:
      - በጉዞ ላይ ያሉ ሰዎች
      - ፓርኩ ባዶ ነበር
      - መነጽር ያለባት ሴት ማይክሮፎን ውስጥ እየዘፈነች.
  - source_sentence: በላዩ ላይ ካለው ህንፃ አጠገብ ያሉት ሦስት ሰዎች ቆመው ነበር.
    sentences:
      - አንዳንድ ልጆች በመዝናኛ መናፈሻ ቦታ ፊኛ ግልቢያ ላይ ናቸው።
      - ሰዎች ውጭ ናቸው.
      - ሰዎች ቦውሊንግ ናቸው።
  - source_sentence: ሰው እየሮጠ ነው።
    sentences:
      - እግሩ ላይ ጫማ ሳይኖረው የሚሮጥ ሰው።
      - ፎቶ ለመነሳት በትልቅ ድንጋይ ላይ የተቀመጠ ሰው።
      - ጎዳና ላይ የሚሮጥ ወጣት።
  - source_sentence: ሐምራዊ ልብስ ከነጭ ሐምራዊ ልብስ ጋር አንድ ሰው እጁን በሌላ ሰው በግንባታው ላይ ከፍ አደረገ.
    sentences:
      - ነጭ ልብስ የለበሰ ሰው እጁን በሌላው ሰው ግንባሩ ላይ አደረገ።
      - >-
        ባለብዙ ልብስ እና ቀይ ብርጭቆ ያላቸው ወጣት ሴት አንድ ወጣት ሴት እየተመለከተች እያለ የተወሰነ የእጅ ቦታን
        ያስተካክላል.
      - አንድ ቄስ በእምነቱ ላይ ብቻውን እያሰላሰለ ነው።
  - source_sentence: ሰዎች ተቀምጠዋል
    sentences:
      - ሁለት ሰዎች አንድ ልብ በምድር ውስጥ ከተቀመጠባቸው ጥቂት እግሮች ጥቂት መሬት ተቀምጠዋል.
      - ሰዎች እየተራመዱ ሲሆን አንድ ወንድ ጭንቅላቱን ወደ ግራ ዞሯል.
      - አንድ ሰው በበረዶ መንሸራተቻ መናፈሻ ውስጥ ካለው መወጣጫ ላይ ብስክሌቱን እየዘለለ ነው።
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
  - cosine_accuracy
model-index:
  - name: >-
      SentenceTransformer based on
      sentence-transformers/paraphrase-xlm-r-multilingual-v1
    results:
      - task:
          type: triplet
          name: Triplet
        dataset:
          name: amharic xlmr nli dev
          type: amharic-xlmr-nli-dev
        metrics:
          - type: cosine_accuracy
            value: 0.815
            name: Cosine Accuracy
      - task:
          type: triplet
          name: Triplet
        dataset:
          name: amharic xlmr finetuned dev
          type: amharic-xlmr-finetuned-dev
        metrics:
          - type: cosine_accuracy
            value: 0.8575
            name: Cosine Accuracy

SentenceTransformer based on sentence-transformers/paraphrase-xlm-r-multilingual-v1

This is a sentence-transformers model finetuned from sentence-transformers/paraphrase-xlm-r-multilingual-v1 on the csv dataset. It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.

Model Details

Model Description

Model Type: Sentence Transformer
Base model: sentence-transformers/paraphrase-xlm-r-multilingual-v1
Maximum Sequence Length: 128 tokens
Output Dimensionality: 768 dimensions
Similarity Function: Cosine Similarity
Training Dataset:
- csv

Model Sources

Documentation: Sentence Transformers Documentation
Repository: Sentence Transformers on GitHub
Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: XLMRobertaModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
)

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers

Then you can load this model and run inference.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("abdulmunimjemal/amharic-xlmr-finetuned")
# Run inference
sentences = [
    'ሰዎች ተቀምጠዋል',
    'ሁለት ሰዎች አንድ ልብ በምድር ውስጥ ከተቀመጠባቸው ጥቂት እግሮች ጥቂት መሬት ተቀምጠዋል.',
    'ሰዎች እየተራመዱ ሲሆን አንድ ወንድ ጭንቅላቱን ወደ ግራ ዞሯል.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Triplet

Datasets: amharic-xlmr-nli-dev and amharic-xlmr-finetuned-dev
Evaluated with TripletEvaluator

Metric	amharic-xlmr-nli-dev	amharic-xlmr-finetuned-dev
cosine_accuracy	0.815	0.8575

Training Details

Training Dataset

csv

Dataset: csv
Size: 40,000 training samples
Columns: anchor, positive, and negative

Approximate statistics based on the first 1000 samples:

	anchor	positive	negative
type	string	string	string
details	min: 6 tokens mean: 6.5 tokens max: 7 tokens	min: 10 tokens mean: 24.81 tokens max: 63 tokens	min: 18 tokens mean: 26.34 tokens max: 38 tokens

Samples:

anchor	positive	negative
`ሰውየው ውጭ ነው.`	`በቤቶች በተከበበችው የኪሳ ፓርክ ውስጥ ረዥም ፀጉር ያለው አንድ ሰው የመንሸራተቻ ሰሌዳ ነው.`	`በወንድ ወንበር ላይ በወንድ ወንበር ላይ በመዝገቢያ ወረቀት ላይ በመዝጋት ላይ ይተኛል.`
`ሰውየው ውጭ ነው።`	`ረጅም ፀጉር ያለው ሰው በቤቶች በተከበበ የበረዶ መንሸራተቻ ፓርክ ውስጥ በስኬትቦርዲንግ ላይ ነው።`	`በቀን ውስጥ አንድ ሰው በአንድ ክፍል ውስጥ የእጅ መቆንጠጫ ይሠራል.`
`ሰውየው ውጭ ነው.`	`በቤቶች በተከበበችው የኪሳ ፓርክ ውስጥ ረዥም ፀጉር ያለው አንድ ሰው የመንሸራተቻ ሰሌዳ ነው.`	`ባለሙያ የለበሰ ሰው ከኮንሶል ፊት ለፊት ተቀምጧል።`

Loss: TripletLoss with these parameters:

{
    "distance_metric": "TripletDistanceMetric.COSINE",
    "triplet_margin": 5
}

Evaluation Dataset

csv

Dataset: csv
Size: 40,000 evaluation samples
Columns: anchor, positive, and negative

Approximate statistics based on the first 1000 samples:

	anchor	positive	negative
type	string	string	string
details	min: 5 tokens mean: 13.91 tokens max: 56 tokens	min: 5 tokens mean: 20.28 tokens max: 67 tokens	min: 5 tokens mean: 21.07 tokens max: 55 tokens

Samples:

anchor	positive	negative
`ቡናማ ውሻ እየሮጠ እና እየተመለከተ ነው.`	`ቡናማ ውሻ ወደ ሰማይ እያየ ይሮጣል.`	`ከዛፉ ስር አንድ ቡናማ ውሻ ይዞ ይገኛል.`
`በወርቃማ ቀለም ያለው ቀለም ያለው ቆዳዎች በሣር ውስጥ.`	`ውሻ በሳሩ ውስጥ ነው.`	`ውሻ ውጭ እየተጣደፈ ነው.`
`ቡናማ ውሻ ከቤት ውጭ እየተጫወተ ነው.`	`አንድ እንስሳ ውጭ ነው.`	`ቡናማ ውሻ በኩሽና ውስጥ እየበላ ነው.`

Loss: TripletLoss with these parameters:

{
    "distance_metric": "TripletDistanceMetric.COSINE",
    "triplet_margin": 5
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 32
gradient_accumulation_steps: 2
learning_rate: 1e-05
weight_decay: 0.01
warmup_ratio: 0.1
batch_sampler: no_duplicates

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: steps
prediction_loss_only: True
per_device_train_batch_size: 32
per_device_eval_batch_size: 8
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 2
eval_accumulation_steps: None
torch_empty_cache_steps: None
learning_rate: 1e-05
weight_decay: 0.01
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1.0
num_train_epochs: 3
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.1
warmup_steps: 0
log_level: passive
log_level_replica: warning
log_on_each_node: True
logging_nan_inf_filter: True
save_safetensors: True
save_on_each_node: False
save_only_model: False
restore_callback_states_from_checkpoint: False
no_cuda: False
use_cpu: False
use_mps_device: False
seed: 42
data_seed: None
jit_mode_eval: False
use_ipex: False
bf16: False
fp16: False
fp16_opt_level: O1
half_precision_backend: auto
bf16_full_eval: False
fp16_full_eval: False
tf32: None
local_rank: 0
ddp_backend: None
tpu_num_cores: None
tpu_metrics_debug: False
debug: []
dataloader_drop_last: False
dataloader_num_workers: 0
dataloader_prefetch_factor: None
past_index: -1
disable_tqdm: False
remove_unused_columns: True
label_names: None
load_best_model_at_end: False
ignore_data_skip: False
fsdp: []
fsdp_min_num_params: 0
fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
fsdp_transformer_layer_cls_to_wrap: None
accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
deepspeed: None
label_smoothing_factor: 0.0
optim: adamw_torch
optim_args: None
adafactor: False
group_by_length: False
length_column_name: length
ddp_find_unused_parameters: None
ddp_bucket_cap_mb: None
ddp_broadcast_buffers: False
dataloader_pin_memory: True
dataloader_persistent_workers: False
skip_memory_metrics: True
use_legacy_prediction_loop: False
push_to_hub: False
resume_from_checkpoint: None
hub_model_id: None
hub_strategy: every_save
hub_private_repo: None
hub_always_push: False
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
include_inputs_for_metrics: False
include_for_metrics: []
eval_do_concat_batches: True
fp16_backend: auto
push_to_hub_model_id: None
push_to_hub_organization: None
mp_parameters:
auto_find_batch_size: False
full_determinism: False
torchdynamo: None
ray_scope: last
ddp_timeout: 1800
torch_compile: False
torch_compile_backend: None
torch_compile_mode: None
dispatch_batches: None
split_batches: None
include_tokens_per_second: False
include_num_input_tokens_seen: False
neftune_noise_alpha: None
optim_target_modules: None
batch_eval_metrics: False
eval_on_start: False
use_liger_kernel: False
eval_use_gather_object: False
average_tokens_across_devices: False
prompts: None
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss	amharic-xlmr-nli-dev_cosine_accuracy	amharic-xlmr-finetuned-dev_cosine_accuracy
0	0	-	-	0.6432	-
0.05	100	4.6673	4.3076	0.8492	-
0.1	200	4.1006	3.6344	0.821	-
0.15	300	3.843	4.1666	0.7652	-
0.2	400	4.0508	3.8094	0.815	-
0.25	500	3.9858	-	-	0.8237
0.2	100	4.15	-	-	-
0.4	200	4.1811	-	-	-
0.6	300	4.3359	-	-	-
0.8	400	4.382	-	-	-
1.0	500	3.6309	3.5175	-	0.858
1.198	600	4.1283	-	-	-
1.3980	700	4.0372	-	-	-
1.5980	800	4.2113	-	-	-
1.798	900	4.059	-	-	-
1.998	1000	3.4594	3.5366	-	0.8565
2.196	1100	4.0407	-	-	-
2.396	1200	3.9531	-	-	-
2.596	1300	4.1321	-	-	-
2.7960	1400	3.9537	-	-	-
2.996	1500	3.4291	3.5476	-	0.8575

Framework Versions

Python: 3.11.11
Sentence Transformers: 3.3.1
Transformers: 4.47.1
PyTorch: 2.5.1+cu121
Accelerate: 1.2.1
Datasets: 3.2.0
Tokenizers: 0.21.0

Citation

BibTeX

Sentence Transformers

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084",
}

TripletLoss

@misc{hermans2017defense,
    title={In Defense of the Triplet Loss for Person Re-Identification},
    author={Alexander Hermans and Lucas Beyer and Bastian Leibe},
    year={2017},
    eprint={1703.07737},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}