rgrupesh/reranker-model

CrossEncoder based on bioformers/bioformer-16L

This is a Cross Encoder model finetuned from bioformers/bioformer-16L 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: bioformers/bioformer-16L
• Maximum Sequence Length: 512 tokens
• Number of Output Labels: 1 label

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 = [
    ['cell_type: C_BEST4; tissue: descending colon; organism: Homo sapiens', 'label: smooth muscle fiber of descending colon; synonyms: non-striated muscle fiber of descending colon; definition: A smooth muscle cell that is part of the descending colon.'],
    ['cell_type: C_BEST4; tissue: descending colon; organism: Homo sapiens', 'label: smooth muscle cell of colon; synonyms: non-striated muscle fiber of colon; definition: A smooth muscle cell that is part of the colon.'],
    ['cell_type: C_BEST4; tissue: descending colon; organism: Homo sapiens', 'label: BEST4+ colonocyte; definition: An enterocyte of the human colon expressing bestrophin-4 (BEST4) calcium-activated ion channels.  These cells have a distinct transcriptomic profile compared to other enterocytes and are scattered thr'],
    ['cell_type: C_BEST4; tissue: descending colon; organism: Homo sapiens', 'label: smooth muscle fiber of ascending colon; synonyms: non-striated muscle fiber of ascending colon; definition: A smooth muscle cell that is part of the ascending colon.'],
    ['cell_type: C_BEST4; tissue: descending colon; organism: Homo sapiens', 'label: type D cell of colon; synonyms: colon D-cell | colonic delta cell | delta cell of colon; definition: A D cell located in the colon.'],
]
scores = model.predict(pairs)
print(scores.shape)
# (5,)

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    'cell_type: C_BEST4; tissue: descending colon; organism: Homo sapiens',
    [
        'label: smooth muscle fiber of descending colon; synonyms: non-striated muscle fiber of descending colon; definition: A smooth muscle cell that is part of the descending colon.',
        'label: smooth muscle cell of colon; synonyms: non-striated muscle fiber of colon; definition: A smooth muscle cell that is part of the colon.',
        'label: BEST4+ colonocyte; definition: An enterocyte of the human colon expressing bestrophin-4 (BEST4) calcium-activated ion channels.  These cells have a distinct transcriptomic profile compared to other enterocytes and are scattered thr',
        'label: smooth muscle fiber of ascending colon; synonyms: non-striated muscle fiber of ascending colon; definition: A smooth muscle cell that is part of the ascending colon.',
        'label: type D cell of colon; synonyms: colon D-cell | colonic delta cell | delta cell of colon; definition: A D cell located in the colon.',
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Evaluation

Metrics

Cross Encoder Binary Classification

• Dataset: dev
• Evaluated with CEBinaryClassificationEvaluator

Metric	Value
accuracy	0.975
accuracy_threshold	0.9519
f1	0.8237
f1_threshold	0.682
precision	0.7958
recall	0.8536
average_precision	0.8867

Training Details

Training Dataset

Unnamed Dataset

• Size: 2,401,485 training samples
• Columns: sentence1, sentence2, and label
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2	label
  type	string	string	float
  details	min: 53 characters mean: 63.88 characters max: 77 characters	min: 24 characters mean: 222.07 characters max: 681 characters	min: 0.0 mean: 0.07 max: 1.0

• Samples:

  sentence1	sentence2	label
  cell_type: cDC1; tissue: tonsil; organism: Homo sapiens	label: tonsil germinal center B cell; definition: Any germinal center B cell that is part of a tonsil.	0.0
  cell_type: cDC1; tissue: tonsil; organism: Homo sapiens	label: germinal center T cell; synonyms: T follicular helper cell of germinal center of lymph node	T follicular helper cell of lymph node germinal center
  cell_type: cDC1; tissue: tonsil; organism: Homo sapiens	label: tonsil squamous cell; synonyms: tonsil squamous epithelial cell	tonsil squamous epithelial cells

• Loss: BinaryCrossEntropyLoss with these parameters:

  {
      "activation_fn": "torch.nn.modules.linear.Identity",
      "pos_weight": 5.0
  }
  

Evaluation Dataset

Unnamed Dataset

• Size: 132,647 evaluation samples
• Columns: sentence1, sentence2, and label
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2	label
  type	string	string	float
  details	min: 51 characters mean: 71.85 characters max: 100 characters	min: 23 characters mean: 221.83 characters max: 681 characters	min: 0.0 mean: 0.07 max: 1.0

• Samples:

  sentence1	sentence2	label
  cell_type: C_BEST4; tissue: descending colon; organism: Homo sapiens	label: smooth muscle fiber of descending colon; synonyms: non-striated muscle fiber of descending colon; definition: A smooth muscle cell that is part of the descending colon.	0.0
  cell_type: C_BEST4; tissue: descending colon; organism: Homo sapiens	label: smooth muscle cell of colon; synonyms: non-striated muscle fiber of colon; definition: A smooth muscle cell that is part of the colon.	0.0
  cell_type: C_BEST4; tissue: descending colon; organism: Homo sapiens	label: BEST4+ colonocyte; definition: An enterocyte of the human colon expressing bestrophin-4 (BEST4) calcium-activated ion channels. These cells have a distinct transcriptomic profile compared to other enterocytes and are scattered thr	0.0

• Loss: BinaryCrossEntropyLoss with these parameters:

  {
      "activation_fn": "torch.nn.modules.linear.Identity",
      "pos_weight": 5.0
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 32
• learning_rate: 2e-05
• weight_decay: 0.01
• warmup_ratio: 0.1
• fp16: True
• 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: 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: 0.01
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 3
• 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}
• 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
• 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: False
• 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: False
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss	dev_average_precision
2.0947	78600	0.1805	-	-
2.0973	78700	0.1498	-	-
2.1000	78800	0.1899	-	-
2.1027	78900	0.1924	-	-
2.1053	79000	0.1521	0.2484	0.8835
2.1080	79100	0.1988	-	-
2.1106	79200	0.1702	-	-
2.1133	79300	0.1824	-	-
2.1160	79400	0.1688	-	-
2.1186	79500	0.1809	0.2202	0.8798
2.1213	79600	0.1858	-	-
2.1240	79700	0.1642	-	-
2.1266	79800	0.191	-	-
2.1293	79900	0.1655	-	-
2.1320	80000	0.1789	0.1960	0.8772
2.1346	80100	0.2002	-	-
2.1373	80200	0.2139	-	-
2.1400	80300	0.199	-	-
2.1426	80400	0.165	-	-
2.1453	80500	0.1848	0.2662	0.8782
2.1480	80600	0.2118	-	-
2.1506	80700	0.177	-	-
2.1533	80800	0.169	-	-
2.1560	80900	0.1748	-	-
2.1586	81000	0.1532	0.2608	0.8839
2.1613	81100	0.1618	-	-
2.1639	81200	0.1679	-	-
2.1666	81300	0.1808	-	-
2.1693	81400	0.1685	-	-
2.1719	81500	0.1901	0.2348	0.8830
2.1746	81600	0.184	-	-
2.1773	81700	0.1667	-	-
2.1799	81800	0.1816	-	-
2.1826	81900	0.159	-	-
2.1853	82000	0.1553	0.2375	0.8870
2.1879	82100	0.1572	-	-
2.1906	82200	0.1629	-	-
2.1933	82300	0.194	-	-
2.1959	82400	0.1685	-	-
2.1986	82500	0.1859	0.2293	0.8798
2.2013	82600	0.1483	-	-
2.2039	82700	0.2056	-	-
2.2066	82800	0.1799	-	-
2.2093	82900	0.1645	-	-
2.2119	83000	0.1633	0.2365	0.8839
2.2146	83100	0.1714	-	-
2.2172	83200	0.1793	-	-
2.2199	83300	0.154	-	-
2.2226	83400	0.151	-	-
2.2252	83500	0.1841	0.2031	0.8854
2.2279	83600	0.1709	-	-
2.2306	83700	0.1875	-	-
2.2332	83800	0.2041	-	-
2.2359	83900	0.1812	-	-
2.2386	84000	0.1779	0.2351	0.8889
2.2412	84100	0.224	-	-
2.2439	84200	0.1624	-	-
2.2466	84300	0.1361	-	-
2.2492	84400	0.1371	-	-
2.2519	84500	0.2314	0.2933	0.8862
2.2546	84600	0.1889	-	-
2.2572	84700	0.2079	-	-
2.2599	84800	0.1783	-	-
2.2626	84900	0.1551	-	-
2.2652	85000	0.1606	0.2096	0.8915
2.2679	85100	0.1887	-	-
2.2705	85200	0.1829	-	-
2.2732	85300	0.18	-	-
2.2759	85400	0.1935	-	-
2.2785	85500	0.1934	0.2301	0.8867

Framework Versions

• Python: 3.11.13
• Sentence Transformers: 4.1.0
• Transformers: 4.53.3
• PyTorch: 2.6.0+cu124
• Accelerate: 1.9.0
• Datasets: 4.4.1
• Tokenizers: 0.21.2

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",
}