SentenceTransformer based on BAAI/bge-small-en-v1.5

This is a sentence-transformers model finetuned from BAAI/bge-small-en-v1.5. It maps sentences & paragraphs to a 384-dimensional dense vector space and can be used for retrieval.

Model Details

Model Description

• Model Type: Sentence Transformer
• Base model: BAAI/bge-small-en-v1.5
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 384 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': 'BertModel'})
  (1): Pooling({'embedding_dimension': 384, 'pooling_mode': 'cls', 'include_prompt': True})
  (2): Normalize({})
)

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("RinKana/bge-small-en-v1.5-afterimage-v.0.1")
# Run inference
sentences = [
    'Represent this dialogue context for retrieving the missing turn: Speaker B is replying. Previous context: Speaker A: I am considering going for the new job that was posted yesterday . Speaker B: Are you certain that that is what you want to do ? Speaker B: Why do you think that this would be a good move ? Speaker A: I believe that this job would allow me to move up but might be a little boring for me . Following context: Speaker A: Also , the matter of pay is also a consideration . Speaker A: Yes , sometimes giving up a little to move forward is the best choice . Speaker B: I think you should definitely apply for the position .',
    'Candidate response: Yes , there are always pros and cons to making a career change .',
    'Candidate response: It may not be the best choice for me , but I am considering it .',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.7507, 0.6233],
#         [0.7507, 1.0000, 0.5267],
#         [0.6233, 0.5267, 1.0000]])

Evaluation

Metrics

Information Retrieval

• Dataset: gap-to-candidate-eval
• Evaluated with InformationRetrievalEvaluator

Training Details

Training Dataset

Unnamed Dataset

• Size: 4,744 training samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 51 tokensmean: 114.19 tokensmax: 296 tokens	min: 8 tokensmean: 19.66 tokensmax: 71 tokens	min: 9 tokensmean: 19.62 tokensmax: 59 tokens

• Samples:

  anchor	positive	negative
  Represent this dialogue context for retrieving the missing turn: Speaker B is replying. Previous context: Speaker A: Oh, yes. Could you tell me your name? Speaker B: Oh, Jerry Lynn. Speaker A: Well, what exactly is it that interests you about the job? Speaker B: I just thought that it was right up my street, you know. Speaker A: Really? Could you tell me a little about yourself? Following context: Speaker A: Do you have any special skills? Speaker B: Well, I speak two foreign languages, French and Italian. Speaker A: Well, I see.	Candidate response: I've been working abroad doing secretarial work. Previous to that I was at university. I've got a degree in English.	Candidate response: Thank you . Could you tell me the specials today ?
  Represent this dialogue context for retrieving the missing turn: Speaker B is replying. Previous context: Speaker A: We'll have to get you to a hospital for an X-ray . Speaker B: Is that really necessary ? Speaker A: It may be worse than it seems . How's it feel ? Speaker B: About the same . Speaker A: Someone will be out to take a look any minute now . Following context: Speaker A: Of course ! It's the sort of thing a hospital emergency ward is for . Will we have to wait much longer ? Speaker B: I hope not.It isn't that busy .	Candidate response: Should we be here , Alice ? I don't think it's that serious .	Candidate response: You don't have to worry about it . We have a one year warranty .
  Represent this dialogue context for retrieving the missing turn: Speaker B is replying. Previous context: Speaker A: ( Bob groans . ) What's the matter.Bob ? Speaker B: I think it's my ankle . Speaker A: What happened ? Speaker B: One of my snowshoes got caught on a rock . Speaker A: Can you stand ? Following context: Speaker A: We'll have to get you to a hospital for an X-ray . Speaker B: Is that really necessary ? Speaker A: It may be worse than it seems . How's it feel ? Speaker B: About the same . Speaker A: Someone will be out to take a look any minute now .	Candidate response: I don't think so.It ' s kind of sore .	Candidate response: You don't have to worry about it . We have a one year warranty .

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim",
      "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: 16
• per_device_eval_batch_size: 16
• num_train_epochs: 4
• warmup_steps: 0.1
• fp16: True
• load_best_model_at_end: True
• batch_sampler: no_duplicates

All Hyperparameters

• do_predict: False
• prediction_loss_only: True
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• gradient_accumulation_steps: 1
• eval_accumulation_steps: None
• torch_empty_cache_steps: None
• learning_rate: 5e-05
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 4
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: None
• warmup_ratio: None
• warmup_steps: 0.1
• log_level: passive
• log_level_replica: warning
• log_on_each_node: True
• logging_nan_inf_filter: True
• enable_jit_checkpoint: False
• save_on_each_node: False
• save_only_model: False
• restore_callback_states_from_checkpoint: False
• use_cpu: False
• seed: 42
• data_seed: None
• bf16: False
• fp16: True
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• local_rank: -1
• ddp_backend: None
• debug: []
• dataloader_drop_last: False
• dataloader_num_workers: 0
• dataloader_prefetch_factor: None
• disable_tqdm: False
• remove_unused_columns: True
• label_names: None
• load_best_model_at_end: True
• ignore_data_skip: False
• fsdp: []
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': 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
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch_fused
• optim_args: None
• group_by_length: False
• length_column_name: length
• project: huggingface
• trackio_space_id: trackio
• 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
• push_to_hub: False
• resume_from_checkpoint: None
• hub_model_id: None
• hub_strategy: every_save
• hub_private_repo: None
• hub_always_push: False
• hub_revision: None
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• include_for_metrics: []
• eval_do_concat_batches: True
• auto_find_batch_size: False
• full_determinism: False
• ddp_timeout: 1800
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• include_num_input_tokens_seen: no
• neftune_noise_alpha: None
• optim_target_modules: None
• batch_eval_metrics: False
• eval_on_start: False
• use_liger_kernel: False
• liger_kernel_config: None
• eval_use_gather_object: False
• average_tokens_across_devices: True
• use_cache: False
• prompts: None
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Training Time

• Training: 3.7 minutes

Framework Versions

• Python: 3.12.13
• Sentence Transformers: 5.4.1
• Transformers: 5.0.0
• PyTorch: 2.10.0+cu128
• Accelerate: 1.13.0
• Datasets: 5.0.0
• Tokenizers: 0.22.2

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",
}

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},
}