wwydmanski/modernbert-pubmed-v0.1

SentenceTransformer based on answerdotai/ModernBERT-base

This is a sentence-transformers model finetuned from answerdotai/ModernBERT-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: answerdotai/ModernBERT-base
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 dimensions
• 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: ModernBertModel 
  (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 = [
    'EHL tendon reconstruction',
    'A Combined Surgical Approach for Extensor Hallucis Longus Reconstruction: Two Case Reports. ',
    'Flexor tendon reconstruction. ',
]
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]

Evaluation

Metrics

Triplet

• Dataset: triplet-dev
• Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	0.887

Training Details

Training Dataset

json

• Dataset: json
• Size: 10,053 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: 8.86 tokens max: 34 tokens	min: 4 tokens mean: 21.84 tokens max: 62 tokens	min: 3 tokens mean: 13.65 tokens max: 50 tokens

• Samples:

  anchor	positive	negative
  COM-induced secretome changes in U937 monocytes	Characterization of calcium oxalate crystal-induced changes in the secretome of U937 human monocytes.	Monocytes.
  Metamaterials	Sound attenuation optimization using metaporous materials tuned on exceptional points.	Metamaterials: A cat's eye for all directions.
  Pediatric Parasitology	Parasitic infections among school age children 6 to 11-years-of-age in the Eastern province.	[DIALOGUE ON PEDIATRIC PARASITOLOGY].

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 64
• per_device_eval_batch_size: 64
• learning_rate: 0.0002
• num_train_epochs: 2
• 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: steps
• prediction_loss_only: True
• per_device_train_batch_size: 64
• per_device_eval_batch_size: 64
• 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: 0.0002
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 2
• 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: 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
• 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
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Click to expand

Epoch	Step	Training Loss	triplet-dev_cosine_accuracy
0	0	-	0.457
0.0189	1	5.2934	-
0.0377	2	5.2413	-
0.0566	3	4.9969	-
0.0755	4	4.5579	-
0.0943	5	3.9145	-
0.1132	6	3.3775	-
0.1321	7	2.8787	-
0.1509	8	3.0147	-
0.1698	9	2.7166	-
0.1887	10	2.7875	-
0.2075	11	2.3848	-
0.2264	12	2.1921	-
0.2453	13	1.7009	-
0.2642	14	1.7649	-
0.2830	15	1.7948	-
0.3019	16	1.5384	-
0.3208	17	1.6039	-
0.3396	18	1.3364	-
0.3585	19	1.3852	-
0.3774	20	1.2427	-
0.3962	21	1.3216	-
0.4151	22	1.4202	-
0.4340	23	1.2754	-
0.4528	24	1.281	-
0.4717	25	1.1709	0.815
0.4906	26	1.2363	-
0.5094	27	1.2169	-
0.5283	28	1.1495	-
0.5472	29	1.0066	-
0.5660	30	1.0478	-
0.5849	31	1.1511	-
0.6038	32	0.9992	-
0.6226	33	1.095	-
0.6415	34	1.1699	-
0.6604	35	0.9866	-
0.6792	36	1.1303	-
0.6981	37	1.1126	-
0.7170	38	0.889	-
0.7358	39	1.0355	-
0.7547	40	1.0129	-
0.7736	41	1.118	-
0.7925	42	0.8494	-
0.8113	43	1.0829	-
0.8302	44	0.8751	-
0.8491	45	0.8115	-
0.8679	46	0.8579	-
0.8868	47	1.1111	-
0.9057	48	0.9032	-
0.9245	49	1.0394	-
0.9434	50	0.9691	0.862
0.9623	51	1.023	-
0.9811	52	0.9465	-
1.0	53	0.6713	-
1.0189	54	0.9773	-
1.0377	55	0.8693	-
1.0566	56	0.7187	-
1.0755	57	0.805	-
1.0943	58	0.728	-
1.1132	59	1.0967	-
1.1321	60	0.7036	-
1.1509	61	0.8213	-
1.1698	62	0.57	-
1.1887	63	0.7006	-
1.2075	64	0.5091	-
1.2264	65	0.5758	-
1.2453	66	0.4484	-
1.2642	67	0.397	-
1.2830	68	0.6172	-
1.3019	69	0.513	-
1.3208	70	0.4447	-
1.3396	71	0.3205	-
1.3585	72	0.5881	-
1.3774	73	0.2543	-
1.3962	74	0.3648	-
1.4151	75	0.4849	0.876
1.4340	76	0.3455	-
1.4528	77	0.3424	-
1.4717	78	0.224	-
1.4906	79	0.18	-
1.5094	80	0.2255	-
1.5283	81	0.3024	-
1.5472	82	0.1835	-
1.5660	83	0.1946	-
1.5849	84	0.1958	-
1.6038	85	0.1568	-
1.6226	86	0.1626	-
1.6415	87	0.1774	-
1.6604	88	0.1934	-
1.6792	89	0.2426	-
1.6981	90	0.2958	-
1.7170	91	0.1606	-
1.7358	92	0.2281	-
1.7547	93	0.1786	-
1.7736	94	0.2241	-
1.7925	95	0.1909	-
1.8113	96	0.236	-
1.8302	97	0.1332	-
1.8491	98	0.1247	-
1.8679	99	0.156	-
1.8868	100	0.2152	0.889
1.9057	101	0.1549	-
1.9245	102	0.2226	-
1.9434	103	0.21	-
1.9623	104	0.2139	-
1.9811	105	0.1864	-
2.0	106	0.0719	0.887

Framework Versions

• Python: 3.12.3
• Sentence Transformers: 3.3.1
• Transformers: 4.48.0.dev0
• PyTorch: 2.5.1
• Accelerate: 1.2.1
• Datasets: 3.2.0
• Tokenizers: 0.21.0

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}
}