Stevenf232/SapBERT_MultipleNegativesRankingLoss_BC5CDR_Context

SentenceTransformer based on cambridgeltl/SapBERT-from-PubMedBERT-fulltext

This is a sentence-transformers model finetuned from cambridgeltl/SapBERT-from-PubMedBERT-fulltext on the bc5_cdr_me_sh2015_complete 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: cambridgeltl/SapBERT-from-PubMedBERT-fulltext
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • bc5_cdr_me_sh2015_complete

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': 512, 'do_lower_case': False, 'architecture': 'BertModel'})
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, '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("Stevenf232/SapBERT_MultipleNegativesRankingLoss_BC5CDR_Context")
# Run inference
sentences = [
    'insomnia [SEP] pressive symptoms was admitted to a psychiatric hospital due to insomnia, loss of appetite, exhaustion, and agitation. Medical treatment',
    'Sleep Initiation and Maintenance Disorders [SEP] Disorders characterized by impairment of the ability to initiate or maintain sleep. This may occur as a primary disorder or in a',
    'Atrioventricular Block [SEP] Impaired impulse conduction from HEART ATRIA to HEART VENTRICLES. AV block can mean delayed or completely blocked impulse conduc',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.8093, 0.1453],
#         [0.8093, 1.0000, 0.1948],
#         [0.1453, 0.1948, 1.0000]])

Training Details

Training Dataset

bc5_cdr_me_sh2015_complete

• Dataset: bc5_cdr_me_sh2015_complete at f40f655
• Size: 5,424 training samples
• Columns: sentence1, sentence2, and label
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2	label
  type	string	string	int
  details	min: 9 tokens mean: 29.07 tokens max: 79 tokens	min: 4 tokens mean: 25.04 tokens max: 43 tokens	1: 100.00%

• Samples:

  sentence1	sentence2	label
  Naloxone [SEP] Naloxone reverses the antihypertensive effect of clonidine.	Naloxone [SEP] A specific opiate antagonist that has no agonist activity. It is a competitive antagonist at mu, delta, and kappa opioid recepto	1
  clonidine [SEP] Naloxone reverses the antihypertensive effect of clonidine.	Clonidine [SEP] An imidazoline sympatholytic agent that stimulates ALPHA-2 ADRENERGIC RECEPTORS and central IMIDAZOLINE RECEPTORS. It is commonl	1
  hypertensive [SEP] In unanesthetized, spontaneously hypertensive rats the decrease in blood pressure and heart rate produced by	Hypertension [SEP] Persistently high systemic arterial BLOOD PRESSURE. Based on multiple readings (BLOOD PRESSURE DETERMINATION), hypertension is c	1

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim",
      "gather_across_devices": false
  }
  

Evaluation Dataset

bc5_cdr_me_sh2015_complete

• Dataset: bc5_cdr_me_sh2015_complete at f40f655
• Size: 5,445 evaluation samples
• Columns: sentence1, sentence2, and label
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2	label
  type	string	string	int
  details	min: 11 tokens mean: 30.69 tokens max: 166 tokens	min: 4 tokens mean: 24.66 tokens max: 62 tokens	1: 100.00%

• Samples:

  sentence1	sentence2	label
  Tricuspid valve regurgitation [SEP] Tricuspid valve regurgitation and lithium carbonate toxicity in a newborn infant.	Tricuspid Valve Insufficiency [SEP] Backflow of blood from the RIGHT VENTRICLE into the RIGHT ATRIUM due to imperfect closure of the TRICUSPID VALVE.	1
  lithium carbonate [SEP] Tricuspid valve regurgitation and lithium carbonate toxicity in a newborn infant.	Lithium Carbonate [SEP] A lithium salt, classified as a mood-stabilizing agent. Lithium ion alters the metabolism of BIOGENIC MONOAMINES in the CENTRAL	1
  toxicity [SEP] Tricuspid valve regurgitation and lithium carbonate toxicity in a newborn infant.	Drug-Related Side Effects and Adverse Reactions [SEP] Disorders that result from the intended use of PHARMACEUTICAL PREPARATIONS. Included in this heading are a broad variety of chem	1

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim",
      "gather_across_devices": false
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 64
• per_device_eval_batch_size: 64
• learning_rate: 2e-05
• max_steps: 200
• warmup_ratio: 0.1
• warmup_steps: 0.1
• fp16: True

All Hyperparameters

Click to expand

• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 64
• per_device_eval_batch_size: 64
• gradient_accumulation_steps: 1
• eval_accumulation_steps: None
• torch_empty_cache_steps: None
• learning_rate: 2e-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: 3
• max_steps: 200
• lr_scheduler_type: linear
• lr_scheduler_kwargs: None
• warmup_ratio: 0.1
• 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: False
• 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: batch_sampler
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	Validation Loss
0.1176	10	2.6695	2.4324
0.2353	20	2.2030	1.8628
0.3529	30	1.6394	1.5455
0.4706	40	1.5937	1.3570
0.5882	50	1.3294	1.2489
0.7059	60	1.2576	1.1594
0.8235	70	1.0213	1.1042
0.9412	80	1.0295	1.0672
1.0588	90	0.8890	1.0293
1.1765	100	0.9259	1.0030
1.2941	110	0.8096	0.9743
1.4118	120	0.7438	0.9587
1.5294	130	0.7797	0.9442
1.6471	140	0.7999	0.9265
1.7647	150	0.7323	0.9142
1.8824	160	0.7510	0.9070
2.0	170	0.7297	0.9032
2.1176	180	0.6434	0.8985
2.2353	190	0.5984	0.8967
2.3529	200	0.6603	0.8959

Framework Versions

• Python: 3.12.12
• Sentence Transformers: 5.2.3
• Transformers: 5.0.0
• PyTorch: 2.10.0+cu128
• Accelerate: 1.12.0
• Datasets: 4.0.0
• 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{henderson2017efficient,
    title={Efficient Natural Language Response Suggestion for Smart Reply},
    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
    year={2017},
    eprint={1705.00652},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}