OverSamu/reranker-sapbert-ncbi-disease-bce-context-title

Cambridge-SapBERT-from-PubMedBERT-fulltext trained on NCBI Disease

This is a Cross Encoder model finetuned from cambridgeltl/SapBERT-from-PubMedBERT-fulltext using the sentence-transformers library. It computes scores for pairs of texts, which can be used for text reranking and semantic search.

Model Details

Model Description

• Model Type: Cross Encoder
• Base model: cambridgeltl/SapBERT-from-PubMedBERT-fulltext
• Maximum Sequence Length: 512 tokens
• Number of Output Labels: 1 label
• Language: en

Model Sources

• Documentation: Sentence Transformers Documentation
• Documentation: Cross Encoder Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face: Cross Encoders on Hugging Face

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 CrossEncoder

# Download from the 🤗 Hub
model = CrossEncoder("cross_encoder_model_id")
# Get scores for pairs of texts
pairs = [
    ['Original mention: breast cancer.\nTitle: BRCA1 mutations in a population-based sample of young women with breast cancer.\nContext: We studied 80 women in whom breast cancer was diagnosed before the age of 35, and who were not selected on the basis of family history.', 'breast tumors'],
    ['Original mention: childhood cerebral ALD.\nTitle: Predominance of the adrenomyeloneuropathy phenotype of X-linked adrenoleukodystrophy in The Netherlands: a survey of 30 kindreds.\nContext: The phenotypic expression is highly variable, childhood cerebral ALD (CCALD) and adrenomyeloneuropathy (AMN) being the main variants.', 'x-linked adrenoleukodystrophy'],
    ['Original mention: TSD.\nTitle: The Tay-Sachs disease gene in North American Jewish populations: geographic variations and origin.\nContext: Jews with Polish and/or Russian ancestry constituted 88% of this sample and had a TSD carrier frequency of.', 'gm2 gangliosidosis, type 1'],
    ['Original mention: PWS.\nTitle: Isolation of molecular probes associated with the chromosome 15 instability in the Prader-Willi syndrome.\nContext: 2 and are shown to be deleted in DNA of one of two patients examined with the PWS.', 'syndrome, royer'],
    ['Original mention: deficiency of beta-glucocerebrosidase.\nTitle: Homozygous presence of the crossover (fusion gene) mutation identified in a type II Gaucher disease fetus: is this analogous to the Gaucher knock-out mouse model?\nGaucher disease (GD) is an inherited deficiency of beta-glucocerebrosidase (EC 3.\nContext: Homozygous presence of the crossover (fusion gene) mutation identified in a type II Gaucher disease fetus: is this analogous to the Gaucher knock-out mouse model?\nGaucher disease (GD) is an inherited deficiency of beta-glucocerebrosidase (EC 3.', 'gaucher disease, acute neuronopathic type'],
]
scores = model.predict(pairs)
print(scores.shape)
# (5,)

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    'Original mention: breast cancer.\nTitle: BRCA1 mutations in a population-based sample of young women with breast cancer.\nContext: We studied 80 women in whom breast cancer was diagnosed before the age of 35, and who were not selected on the basis of family history.',
    [
        'breast tumors',
        'x-linked adrenoleukodystrophy',
        'gm2 gangliosidosis, type 1',
        'syndrome, royer',
        'gaucher disease, acute neuronopathic type',
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Evaluation

Metrics

Cross Encoder Reranking

• Dataset: ncbi-disease-dev
• Evaluated with CrossEncoderRerankingEvaluator with these parameters:

  {
      "at_k": 10,
      "always_rerank_positives": false
  }
  

Metric	Value
map	0.9965 (+0.5546)
mrr@10	0.9981 (+0.7243)
ndcg@10	0.9979 (+0.4192)

Training Details

Training Dataset

Unnamed Dataset

• Size: 222,541 training samples
• Columns: query, answer, and label
• Approximate statistics based on the first 1000 samples:

  	query	answer	label
  type	string	string	int
  details	min: 123 characters mean: 317.4 characters max: 700 characters	min: 5 characters mean: 27.96 characters max: 168 characters	0: ~44.80% 1: ~55.20%

• Samples:

  query	answer	label
  Original mention: breast cancer. Title: BRCA1 mutations in a population-based sample of young women with breast cancer. Context: We studied 80 women in whom breast cancer was diagnosed before the age of 35, and who were not selected on the basis of family history.	breast tumors	1
  Original mention: childhood cerebral ALD. Title: Predominance of the adrenomyeloneuropathy phenotype of X-linked adrenoleukodystrophy in The Netherlands: a survey of 30 kindreds. Context: The phenotypic expression is highly variable, childhood cerebral ALD (CCALD) and adrenomyeloneuropathy (AMN) being the main variants.	x-linked adrenoleukodystrophy	1
  Original mention: TSD. Title: The Tay-Sachs disease gene in North American Jewish populations: geographic variations and origin. Context: Jews with Polish and/or Russian ancestry constituted 88% of this sample and had a TSD carrier frequency of.	gm2 gangliosidosis, type 1	1

• Loss: BinaryCrossEntropyLoss with these parameters:

  {
      "activation_fn": "torch.nn.modules.linear.Identity",
      "pos_weight": 0.7793161273002625
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 128
• per_device_eval_batch_size: 128
• learning_rate: 2e-05
• num_train_epochs: 2
• warmup_ratio: 0.05
• seed: 12
• bf16: True
• dataloader_num_workers: 4
• load_best_model_at_end: True

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 128
• per_device_eval_batch_size: 128
• 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: 2
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: None
• warmup_ratio: 0.05
• 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: 12
• data_seed: None
• jit_mode_eval: 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: 4
• 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}
• 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}
• parallelism_config: None
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch_fused
• optim_args: None
• adafactor: False
• 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
• 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
• hub_revision: None
• 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: 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
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	ncbi-disease-dev_ndcg@10
0.0006	1	0.6083	-
0.0863	150	0.567	-
0.1725	300	0.3364	-
0.2588	450	0.2209	-
0.3450	600	0.1784	-
0.4313	750	0.1435	-
0.5175	900	0.1324	-
0.6038	1050	0.1137	-
0.6901	1200	0.103	-
0.7763	1350	0.0934	-
0.8626	1500	0.0842	0.9949 (+0.4162)
0.9488	1650	0.0797	-
1.0351	1800	0.0695	-
1.1213	1950	0.0573	-
1.2076	2100	0.0613	-
1.2938	2250	0.0555	-
1.3801	2400	0.0504	-
1.4664	2550	0.0499	-
1.5526	2700	0.049	-
1.6389	2850	0.0489	-
1.7251	3000	0.0424	0.9979 (+0.4192)
1.8114	3150	0.0411	-
1.8976	3300	0.0405	-
1.9839	3450	0.0405	-
-1	-1	-	0.9979 (+0.4192)

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.12.12
• Sentence Transformers: 5.2.0
• Transformers: 4.57.5
• PyTorch: 2.9.1+cu128
• Accelerate: 1.12.0
• Datasets: 4.5.0
• 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",
}