wwydmanski/specter2_pubmed-v0.4

SentenceTransformer based on allenai/specter2_base

This is a sentence-transformers model finetuned from allenai/specter2_base on the json 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: allenai/specter2_base
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 tokens
• Similarity Function: Cosine Similarity
• Training Dataset:
  • json

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}) with Transformer model: BertModel 
  (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("sentence_transformers_model_id")
# Run inference
sentences = [
    'Kawasaki disease immunoprophylaxis',
    '[Effect of immunoglobulin in the prevention of coronary artery aneurysms in Kawasaki disease]. ',
    'Management of Kawasaki disease. ',
]
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]

Training Details

Training Dataset

json

• Dataset: json
• Size: 8,705 training samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 4 tokens mean: 7.6 tokens max: 18 tokens	min: 6 tokens mean: 19.26 tokens max: 42 tokens	min: 4 tokens mean: 11.72 tokens max: 46 tokens

• Samples:

  anchor	positive	negative
  Telehealth challenges	[Technological transformations and evolution of the medical practice: current status, issues and perspectives for the development of telemedicine].	The untapped potential of Telehealth.
  Racial disparities in mental health treatment	Relationships between stigma, depression, and treatment in white and African American primary care patients.	Mental Health Care Disparities Now and in the Future.
  Iatrogenic hyperkalemia in elderly patients with cardiovascular disease	Iatrogenic hyperkalemia as a serious problem in therapy of cardiovascular diseases in elderly patients.	The cardiovascular implications of hypokalemia.

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 32
• per_device_eval_batch_size: 32
• learning_rate: 2e-05
• num_train_epochs: 1
• lr_scheduler_type: cosine_with_restarts
• warmup_ratio: 0.1
• bf16: True
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: no
• prediction_loss_only: True
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 32
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• 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: 1
• max_steps: -1
• lr_scheduler_type: cosine_with_restarts
• 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: True
• 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: False
• hub_always_push: False
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• include_inputs_for_metrics: False
• 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
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss
0.0110	1	2.9861
0.0220	2	2.9379
0.0330	3	3.0613
0.0440	4	2.8081
0.0549	5	2.6516
0.0659	6	2.3688
0.0769	7	2.0502
0.0879	8	1.7557
0.0989	9	1.5316
0.1099	10	1.2476
0.1209	11	1.1529
0.1319	12	0.9483
0.1429	13	0.7187
0.1538	14	0.6824
0.1648	15	0.593
0.1758	16	0.4593
0.1868	17	0.3737
0.1978	18	0.5082
0.2088	19	0.4232
0.2198	20	0.3089
0.2308	21	0.2057
0.2418	22	0.2358
0.2527	23	0.2291
0.2637	24	0.2707
0.2747	25	0.1359
0.2857	26	0.2294
0.2967	27	0.157
0.3077	28	0.0678
0.3187	29	0.1022
0.3297	30	0.0713
0.3407	31	0.0899
0.3516	32	0.1385
0.3626	33	0.0809
0.3736	34	0.1053
0.3846	35	0.0925
0.3956	36	0.0675
0.4066	37	0.0841
0.4176	38	0.0366
0.4286	39	0.0768
0.4396	40	0.0529
0.4505	41	0.0516
0.4615	42	0.0342
0.4725	43	0.0456
0.4835	44	0.0344
0.4945	45	0.1337
0.5055	46	0.0883
0.5165	47	0.0691
0.5275	48	0.0322
0.5385	49	0.0731
0.5495	50	0.0376
0.5604	51	0.0464
0.5714	52	0.0173
0.5824	53	0.0516
0.5934	54	0.0703
0.6044	55	0.0273
0.6154	56	0.0374
0.6264	57	0.0292
0.6374	58	0.1195
0.6484	59	0.0852
0.6593	60	0.0697
0.6703	61	0.0653
0.6813	62	0.0426
0.6923	63	0.0288
0.7033	64	0.0344
0.7143	65	0.104
0.7253	66	0.0251
0.7363	67	0.0095
0.7473	68	0.0208
0.7582	69	0.0814
0.7692	70	0.0813
0.7802	71	0.0508
0.7912	72	0.032
0.8022	73	0.0879
0.8132	74	0.095
0.8242	75	0.0932
0.8352	76	0.0868
0.8462	77	0.0231
0.8571	78	0.0144
0.8681	79	0.0179
0.8791	80	0.0457
0.8901	81	0.0935
0.9011	82	0.0658
0.9121	83	0.0553
0.9231	84	0.003
0.9341	85	0.0036
0.9451	86	0.0034
0.9560	87	0.0032
0.9670	88	0.0026
0.9780	89	0.0042
0.9890	90	0.0024
1.0	91	0.0022

Framework Versions

• Python: 3.9.19
• Sentence Transformers: 3.1.1
• Transformers: 4.45.2
• PyTorch: 2.5.0
• Accelerate: 1.0.1
• Datasets: 2.19.0
• Tokenizers: 0.20.3

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}
}