Dathq12/bge-m3-tool-retrieval-vi

SentenceTransformer based on BAAI/bge-m3

This is a sentence-transformers model finetuned from BAAI/bge-m3. It maps inputs to a 1024-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, classification, clustering, and more.

Model Details

Model Description

• Model Type: Sentence Transformer
• Base model: BAAI/bge-m3
• Maximum Sequence Length: 192 tokens
• Output Dimensionality: 1024 dimensions
• Similarity Function: Cosine Similarity
• Supported Modality: Text

Model Sources

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

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'transformer_task': 'feature-extraction', 'modality_config': {'text': {'method': 'forward', 'method_output_name': 'last_hidden_state'}}, 'module_output_name': 'token_embeddings', 'architecture': 'XLMRobertaModel'})
  (1): Pooling({'embedding_dimension': 1024, 'pooling_mode': 'cls', 'include_prompt': True})
  (2): Normalize({'module_input_name': 'sentence_embedding', 'module_output_name': 'sentence_embedding'})
)

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("sentence_transformers_model_id")
# Run inference
queries = [
    "Hi, I'm feeling a bit bored. Can you tell me something interesting?",
]
documents = [
    'get_random_fact. Get a random fact',
    'get_random_trivia. Get a random piece of trivia',
]
query_embeddings = model.encode_query(queries)
document_embeddings = model.encode_document(documents)
print(query_embeddings.shape, document_embeddings.shape)
# [1, 1024] [2, 1024]

# Get the similarity scores for the embeddings
similarities = model.similarity(query_embeddings, document_embeddings)
print(similarities)
# tensor([[0.3718, 0.4430]])

Evaluation

Metrics

Information Retrieval

• Dataset: custom_val
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.4588
cosine_accuracy@3	0.662
cosine_accuracy@5	0.7302
cosine_accuracy@10	0.7961
cosine_precision@1	0.4588
cosine_precision@3	0.2207
cosine_precision@5	0.146
cosine_precision@10	0.0796
cosine_recall@1	0.4588
cosine_recall@3	0.662
cosine_recall@5	0.7302
cosine_recall@10	0.7961
cosine_ndcg@10	0.6287
cosine_mrr@10	0.575
cosine_map@100	0.581

Training Details

Training Dataset

Unnamed Dataset

• Size: 70,988 training samples
• Columns: anchor, positive, negative_1, negative_2, negative_3, and negative_4
• Approximate statistics based on the first 100 samples:

  	anchor	positive	negative_1	negative_2	negative_3	negative_4
  type	string	string	string	string	string	string
  modality	text	text	text	text	text	text
  details	min: 9 tokens mean: 32.96 tokens max: 89 tokens	min: 18 tokens mean: 32.62 tokens max: 76 tokens	min: 16 tokens mean: 30.79 tokens max: 58 tokens	min: 10 tokens mean: 30.11 tokens max: 61 tokens	min: 10 tokens mean: 31.28 tokens max: 64 tokens	min: 11 tokens mean: 31.76 tokens max: 92 tokens

• Samples:

  anchor	positive	negative_1	negative_2	negative_3	negative_4
  Is the email 'john.doe@example.com' part of any known data breaches?	data_breach_checker_endpoint. Checks if the given email has been compromised in any known data breaches using the Data Breach Checker API.. Tham số: email	emailvalidation. Validates if an email address exists using the RapidAPI Whois Lookup service.. Tham số: email	mailcheck. Validates the given email address against specified domain(s) using an external API.. Tham số: email, domain	emailvalidation. Validates an email address using an API service.. Tham số: email	bouncer_email_checker. Validates an email address using the Bouncer Email Checker API.. Tham số: email
  Search for movies and TV shows with the title 'The Dark Knight' without any limit on the number of results.	search_by_title. Search for Movies or TV Shows by title, with optional filters for year, media type, and limit on the number of results.. Tham số: s, l, y, m	search_movies. Search for movies based on a query. Tham số: query, limit	search. Searches for a movie or TV show based on the given title.. Tham số: title, page	search_movie. Search for a movie by title or genre. Tham số: query, max_results	search_movies. Search for movies based on title. Tham số: title
  Hi, I need to calculate the volume of a rectangular prism. The length is 5, width is 3 and height is 2.	calculate_volume. Calculate the volume of a rectangular prism. Tham số: length, width, height	calculate_cylinder_volume. Calculate the volume of a cylinder. Tham số: radius, height	calculate_cylinder_volume. Calculate the volume of a cylinder. Tham số: radius, height	calculate_area_volume. Calculate the area and volume of a three-dimensional object. Tham số: object_type, dimensions	calculate_area_volume. Calculate the area and volume of a 3D shape. Tham số: shape, dimensions

• Loss: CachedMultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim",
      "mini_batch_size": 32,
      "mini_batch_num_tokens": null,
      "gather_across_devices": false,
      "directions": [
          "query_to_doc"
      ],
      "partition_mode": "joint",
      "hardness_mode": null,
      "hardness_strength": 0.0
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 256
• learning_rate: 2e-05
• lr_scheduler_type: cosine
• warmup_steps: 0.1
• fp16: True

All Hyperparameters

Click to expand

• per_device_train_batch_size: 256
• num_train_epochs: 3
• max_steps: -1
• learning_rate: 2e-05
• lr_scheduler_type: cosine
• lr_scheduler_kwargs: None
• warmup_steps: 0.1
• optim: adamw_torch_fused
• optim_args: None
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• optim_target_modules: None
• gradient_accumulation_steps: 1
• average_tokens_across_devices: True
• max_grad_norm: 1.0
• label_smoothing_factor: 0.0
• bf16: False
• fp16: True
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• use_liger_kernel: False
• liger_kernel_config: None
• use_cache: False
• neftune_noise_alpha: None
• torch_empty_cache_steps: None
• auto_find_batch_size: False
• log_on_each_node: True
• logging_nan_inf_filter: True
• include_num_input_tokens_seen: no
• log_level: passive
• log_level_replica: warning
• disable_tqdm: False
• project: huggingface
• trackio_space_id: None
• trackio_bucket_id: None
• trackio_static_space_id: None
• per_device_eval_batch_size: 8
• prediction_loss_only: True
• eval_on_start: False
• eval_do_concat_batches: True
• eval_use_gather_object: False
• eval_accumulation_steps: None
• include_for_metrics: []
• batch_eval_metrics: False
• save_only_model: False
• save_on_each_node: False
• enable_jit_checkpoint: False
• push_to_hub: False
• hub_private_repo: None
• hub_model_id: None
• hub_strategy: every_save
• hub_always_push: False
• hub_revision: None
• load_best_model_at_end: False
• ignore_data_skip: False
• restore_callback_states_from_checkpoint: False
• full_determinism: False
• seed: 42
• data_seed: None
• use_cpu: False
• accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
• parallelism_config: None
• dataloader_drop_last: False
• dataloader_num_workers: 0
• dataloader_pin_memory: True
• dataloader_persistent_workers: False
• dataloader_prefetch_factor: None
• dataloader_multiprocessing_context: None
• dataloader_in_order: True
• remove_unused_columns: True
• label_names: None
• train_sampling_strategy: random
• length_column_name: length
• ddp_find_unused_parameters: None
• ddp_bucket_cap_mb: None
• ddp_broadcast_buffers: False
• ddp_static_graph: None
• ddp_backend: None
• ddp_timeout: 1800
• fsdp: None
• fsdp_config: None
• deepspeed: None
• debug: []
• skip_memory_metrics: True
• do_predict: False
• resume_from_checkpoint: None
• local_rank: -1
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}
• warmup_ratio: None

Training Logs

Epoch	Step	Training Loss	custom_val_cosine_ndcg@10
0.1799	50	3.7251	-
0.3597	100	2.7619	-
0.5396	150	2.1278	-
0.7194	200	1.9997	-
0.8993	250	1.8913	-
1.0791	300	1.8380	0.5842
1.2590	350	1.7996	-
1.4388	400	1.7590	-
1.6187	450	1.7098	-
1.7986	500	1.6860	-
1.9784	550	1.6494	-
2.1583	600	1.6049	0.6243
2.3381	650	1.6211	-
2.5180	700	1.6014	-
2.6978	750	1.5894	-
2.8777	800	1.6079	-
3.0	834	-	0.6287

Training Time

• Training: 6.0 hours
• Evaluation: 10.5 minutes
• Total: 6.2 hours

Framework Versions

• Python: 3.12.13
• Sentence Transformers: 6.0.0
• Transformers: 5.15.1
• PyTorch: 2.10.0+cu128
• Accelerate: 1.13.0
• Datasets: 5.0.0
• Tokenizers: 0.22.2

Additional Resources

• Training and Finetuning Embedding Models with Sentence Transformers: the end-to-end guide for training or finetuning Sentence Transformer models.
• Introduction to Matryoshka Embedding Models: variable-size embeddings that can be truncated with minimal quality loss.
• Binary and Scalar Embedding Quantization for Significantly Faster & Cheaper Retrieval: post-training compression of embedding vectors.
• Multimodal Embedding & Reranker Models with Sentence Transformers: use text, image, audio, and video models through the same API.
• Training and Finetuning Multimodal Embedding & Reranker Models with Sentence Transformers: train multimodal embedding models, with a Visual Document Retrieval walkthrough.

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",
}

CachedMultipleNegativesRankingLoss

@misc{gao2021scaling,
    title={Scaling Deep Contrastive Learning Batch Size under Memory Limited Setup},
    author={Luyu Gao and Yunyi Zhang and Jiawei Han and Jamie Callan},
    year={2021},
    eprint={2101.06983},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

MultipleNegativesRankingLoss

@misc{oord2019representationlearningcontrastivepredictive,
      title={Representation Learning with Contrastive Predictive Coding},
      author={Aaron van den Oord and Yazhe Li and Oriol Vinyals},
      year={2019},
      eprint={1807.03748},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/1807.03748},
}