mghuibregtse/biolinkbert-large-simcse-rat

SentenceTransformer based on michiyasunaga/BioLinkBERT-large

This is a sentence-transformers model finetuned from michiyasunaga/BioLinkBERT-large. It maps sentences & paragraphs to a 1024-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: michiyasunaga/BioLinkBERT-large
• Maximum Sequence Length: 128 tokens
• Output Dimensionality: 1024 dimensions
• Similarity Function: Cosine Similarity

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: BertModel 
  (1): Pooling({'word_embedding_dimension': 1024, '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 = [
    'coactivator-associated arginine methyltransferase 1',
    'tRNA methyltransferase 13 homolog',
    'small nucleolar RNA SNORA17',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

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

Evaluation

Metrics

Semantic Similarity

• Dataset: val-eval
• Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.7654
spearman_cosine	0.7408

Training Details

Training Dataset

Unnamed Dataset

• Size: 180,794 training samples
• Columns: text1, text2, and label
• Approximate statistics based on the first 1000 samples:

  	text1	text2	label
  type	string	string	int
  details	min: 3 tokens mean: 8.47 tokens max: 16 tokens	min: 3 tokens mean: 7.63 tokens max: 21 tokens	0: ~52.50% 1: ~47.50%

• Samples:

  text1	text2	label
  ENSRNOG00000007053	mediator complex subunit 7	1
  ENSRNOG00000060932	small nucleolar RNA SNORA55	1
  ENSRNOG00000015213	ENSRNOG00000024039	0

• Loss: ContrastiveLoss with these parameters:

  {
      "distance_metric": "SiameseDistanceMetric.COSINE_DISTANCE",
      "margin": 1.0,
      "size_average": true
  }
  

Evaluation Dataset

Unnamed Dataset

• Size: 22,599 evaluation samples
• Columns: text1, text2, and label
• Approximate statistics based on the first 1000 samples:

  	text1	text2	label
  type	string	string	int
  details	min: 3 tokens mean: 8.47 tokens max: 27 tokens	min: 3 tokens mean: 7.73 tokens max: 22 tokens	0: ~54.50% 1: ~45.50%

• Samples:

  text1	text2	label
  ENSRNOG00000001350	Naa25	1
  ENSRNOG00000019570	Gng3	1
  AABR07040892.1	ENSRNOG00000039203	1

• Loss: ContrastiveLoss with these parameters:

  {
      "distance_metric": "SiameseDistanceMetric.COSINE_DISTANCE",
      "margin": 1.0,
      "size_average": true
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 256
• per_device_eval_batch_size: 256
• learning_rate: 3e-05
• num_train_epochs: 100
• warmup_ratio: 0.1
• fp16: True
• load_best_model_at_end: True
• ddp_find_unused_parameters: False

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 256
• per_device_eval_batch_size: 256
• 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: 3e-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: 100
• 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: True
• 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: True
• 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}
• tp_size: 0
• 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: False
• 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
• 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: batch_sampler
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss	val-eval_spearman_cosine
0.7072	500	0.1291	-	-
0.9986	706	-	0.1366	-0.0396
1.4144	1000	0.1134	-	-
1.9972	1412	-	0.1040	0.3382
2.1216	1500	0.1066	-	-
2.8289	2000	0.0943	-	-
2.9958	2118	-	0.0867	0.5349
3.5361	2500	0.0863	-	-
3.9943	2824	-	0.0825	0.5669
4.2433	3000	0.0827	-	-
4.9505	3500	0.0806	-	-
4.9929	3530	-	0.0810	0.5764
5.6577	4000	0.0782	-	-
5.9915	4236	-	0.0785	0.5923
6.3649	4500	0.0774	-	-
6.9901	4942	-	0.0774	0.6017
7.0721	5000	0.0758	-	-
7.7793	5500	0.0735	-	-
7.9887	5648	-	0.0773	0.6034
8.4866	6000	0.0719	-	-
8.9873	6354	-	0.0765	0.6052
9.1938	6500	0.0701	-	-
9.9010	7000	0.0685	-	-
9.9859	7060	-	0.0753	0.6165
10.6082	7500	0.0651	-	-
10.9844	7766	-	0.0742	0.6215
11.3154	8000	0.0634	-	-
11.9830	8472	-	0.0730	0.6345
12.0226	8500	0.0612	-	-
12.7298	9000	0.0567	-	-
12.9816	9178	-	0.0720	0.6401
13.4371	9500	0.0538	-	-
13.9802	9884	-	0.0708	0.6514
14.1443	10000	0.0517	-	-
14.8515	10500	0.048	-	-
14.9788	10590	-	0.0691	0.6616
15.5587	11000	0.0436	-	-
15.9774	11296	-	0.0681	0.6692
16.2659	11500	0.0417	-	-
16.9731	12000	0.0394	-	-
16.9760	12002	-	0.0659	0.6819
17.6803	12500	0.0345	-	-
17.9745	12708	-	0.0636	0.6954
18.3876	13000	0.033	-	-
18.9731	13414	-	0.0621	0.7027
19.0948	13500	0.0313	-	-
19.8020	14000	0.028	-	-
19.9717	14120	-	0.0615	0.7066
20.5092	14500	0.0258	-	-
20.9703	14826	-	0.0598	0.7144
21.2164	15000	0.0249	-	-
21.9236	15500	0.0231	-	-
21.9689	15532	-	0.0587	0.7191
22.6308	16000	0.0207	-	-
22.9675	16238	-	0.0582	0.7215
23.3380	16500	0.0199	-	-
23.9661	16944	-	0.0575	0.7245
24.0453	17000	0.0194	-	-
24.7525	17500	0.0169	-	-
24.9646	17650	-	0.0562	0.7293
25.4597	18000	0.0161	-	-
25.9632	18356	-	0.0557	0.7327
26.1669	18500	0.0159	-	-
26.8741	19000	0.0146	-	-
26.9618	19062	-	0.0550	0.7342
27.5813	19500	0.0134	-	-
27.9604	19768	-	0.0551	0.7340
28.2885	20000	0.0132	-	-
28.9590	20474	-	0.0544	0.7373
28.9958	20500	0.0127	-	-
29.7030	21000	0.0112	-	-
29.9576	21180	-	0.0538	0.7387
30.4102	21500	0.011	-	-
30.9562	21886	-	0.0534	0.7403
31.1174	22000	0.0109	-	-
31.8246	22500	0.0099	-	-
31.9547	22592	-	0.0536	0.7402
32.5318	23000	0.0094	-	-
32.9533	23298	-	0.0530	0.7421
33.2390	23500	0.0093	-	-
33.9463	24000	0.0091	-	-
33.9519	24004	-	0.0528	0.7425
34.6535	24500	0.0081	-	-
34.9505	24710	-	0.0524	0.7435
35.3607	25000	0.0081	-	-
35.9491	25416	-	0.0529	0.7421
36.0679	25500	0.008	-	-
36.7751	26000	0.0072	-	-
36.9477	26122	-	0.0526	0.7426
37.4823	26500	0.007	-	-
37.9463	26828	-	0.0522	0.7439
38.1895	27000	0.007	-	-
38.8967	27500	0.0067	-	-
38.9448	27534	-	0.0529	0.7416
39.6040	28000	0.0062	-	-
39.9434	28240	-	0.0523	0.7425
40.3112	28500	0.0062	-	-
40.9420	28946	-	0.0529	0.7408

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 4.1.0
• Transformers: 4.51.3
• PyTorch: 2.6.0+cu124
• Accelerate: 1.6.0
• Datasets: 3.5.1
• Tokenizers: 0.21.1

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",
}

ContrastiveLoss

@inproceedings{hadsell2006dimensionality,
    author={Hadsell, R. and Chopra, S. and LeCun, Y.},
    booktitle={2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)},
    title={Dimensionality Reduction by Learning an Invariant Mapping},
    year={2006},
    volume={2},
    number={},
    pages={1735-1742},
    doi={10.1109/CVPR.2006.100}
}