varadsrivastava/findocranker-mpnet-base-v2

CrossEncoder based on sentence-transformers/all-mpnet-base-v2

This is a Cross Encoder model finetuned from sentence-transformers/all-mpnet-base-v2 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: sentence-transformers/all-mpnet-base-v2
• 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("varadsrivastava/findocranker-mpnet-base-v2")
# Get scores for pairs of texts
pairs = [
    ['What did Fifth Third Bancorp’s leadership say about Fifth Third Bancorp’s dividend policy?', '[DOC=10-K | annual report | comprehensive business overview, risks, financials | 100-300 pages]'],
    ['What did Fifth Third Bancorp’s leadership say about Fifth Third Bancorp’s dividend policy?', '[DOC=10-Q | quarterly report | interim financials, MD&A updates | 30-60 pages]'],
    ['What did Fifth Third Bancorp’s leadership say about Fifth Third Bancorp’s dividend policy?', '[DOC=DEF-14A | proxy statement | governance, compensation, shareholder voting matters | annual filing]'],
    ['What did Fifth Third Bancorp’s leadership say about Fifth Third Bancorp’s dividend policy?', '[DOC=8-K | current report | material events, timely disclosures | ad-hoc filing]'],
    ['What did Fifth Third Bancorp’s leadership say about Fifth Third Bancorp’s dividend policy?', '[DOC=Earnings | earnings call transcript | forward guidance, Q&A, management commentary | quarterly]'],
]
scores = model.predict(pairs)
print(scores.shape)
# (5,)

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    'What did Fifth Third Bancorp’s leadership say about Fifth Third Bancorp’s dividend policy?',
    [
        '[DOC=10-K | annual report | comprehensive business overview, risks, financials | 100-300 pages]',
        '[DOC=10-Q | quarterly report | interim financials, MD&A updates | 30-60 pages]',
        '[DOC=DEF-14A | proxy statement | governance, compensation, shareholder voting matters | annual filing]',
        '[DOC=8-K | current report | material events, timely disclosures | ad-hoc filing]',
        '[DOC=Earnings | earnings call transcript | forward guidance, Q&A, management commentary | quarterly]',
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Training Details

Training Dataset

Unnamed Dataset

• Size: 3,943 training samples
• Columns: query, docs, and labels
• Approximate statistics based on the first 1000 samples:

  	query	docs	labels
  type	string	list	list
  details	min: 59 characters mean: 104.63 characters max: 181 characters	size: 5 elements	size: 5 elements

• Samples:

  query	docs	labels
  What did Fifth Third Bancorp’s leadership say about Fifth Third Bancorp’s dividend policy?	['[DOC=10-K | annual report | comprehensive business overview, risks, financials | 100-300 pages]', '[DOC=10-Q | quarterly report | interim financials, MD&A updates | 30-60 pages]', '[DOC=DEF-14A | proxy statement | governance, compensation, shareholder voting matters | annual filing]', '[DOC=8-K | current report | material events, timely disclosures | ad-hoc filing]', '[DOC=Earnings | earnings call transcript | forward guidance, Q&A, management commentary | quarterly]']	[4, 3, 2, 1, 0]
  How did Qualcomm’s management describe forecasted capital allocation between developing new semiconductor technologies and potential acquisitions?	['[DOC=10-K | annual report | comprehensive business overview, risks, financials | 100-300 pages]', '[DOC=10-Q | quarterly report | interim financials, MD&A updates | 30-60 pages]', '[DOC=8-K | current report | material events, timely disclosures | ad-hoc filing]', '[DOC=DEF-14A | proxy statement | governance, compensation, shareholder voting matters | annual filing]', '[DOC=Earnings | earnings call transcript | forward guidance, Q&A, management commentary | quarterly]']	[4, 3, 2, 1, 0]
  What did GE HealthCare Technologies Inc.’s leadership say about GE HealthCare Technologies Inc.’s dividend policy?	['[DOC=10-K | annual report | comprehensive business overview, risks, financials | 100-300 pages]', '[DOC=8-K | current report | material events, timely disclosures | ad-hoc filing]', '[DOC=Earnings | earnings call transcript | forward guidance, Q&A, management commentary | quarterly]', '[DOC=10-Q | quarterly report | interim financials, MD&A updates | 30-60 pages]', '[DOC=DEF-14A | proxy statement | governance, compensation, shareholder voting matters | annual filing]']	[4, 3, 2, 1, 0]

• Loss: ListNetLoss with these parameters:

  {
      "activation_fn": "torch.nn.modules.linear.Identity",
      "mini_batch_size": null
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 4
• gradient_accumulation_steps: 4
• learning_rate: 2e-05
• weight_decay: 0.01
• num_train_epochs: 5
• lr_scheduler_type: cosine
• warmup_ratio: 0.1
• data_seed: 42
• fp16: True
• dataloader_num_workers: 2

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: no
• prediction_loss_only: True
• per_device_train_batch_size: 4
• per_device_eval_batch_size: 8
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 4
• 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: 5
• max_steps: -1
• lr_scheduler_type: cosine
• 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: 42
• 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: 2
• 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}
• 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
• 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
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss
0.1014	25	1.6085
0.2028	50	1.5942
0.3043	75	1.4848
0.4057	100	1.405
0.5071	125	1.4059
0.6085	150	1.3635
0.7099	175	1.3535
0.8114	200	1.3472
0.9128	225	1.3368
1.0122	250	1.3291
1.1136	275	1.2947
1.2150	300	1.3202
1.3164	325	1.3245
1.4178	350	1.321
1.5193	375	1.298
1.6207	400	1.307
1.7221	425	1.325
1.8235	450	1.3332
1.9249	475	1.301
2.0243	500	1.3106
2.1258	525	1.2973
2.2272	550	1.2995
2.3286	575	1.2978
2.4300	600	1.3109
2.5314	625	1.298
2.6329	650	1.307
2.7343	675	1.2969
2.8357	700	1.2762
2.9371	725	1.2917
3.0365	750	1.2545
3.1379	775	1.271
3.2394	800	1.2609
3.3408	825	1.2694
3.4422	850	1.2906
3.5436	875	1.2951
3.6450	900	1.2852
3.7465	925	1.2788
3.8479	950	1.283
3.9493	975	1.2727
4.0487	1000	1.263
4.1501	1025	1.2662
4.2515	1050	1.2628
4.3529	1075	1.2511
4.4544	1100	1.2788
4.5558	1125	1.2671
4.6572	1150	1.2648
4.7586	1175	1.2694
4.8600	1200	1.2648
4.9615	1225	1.2678

Framework Versions

• Python: 3.12.11
• Sentence Transformers: 5.1.0
• Transformers: 4.56.1
• PyTorch: 2.8.0+cu126
• Accelerate: 1.10.1
• Datasets: 4.0.0
• Tokenizers: 0.22.0

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",
}

ListNetLoss

@inproceedings{cao2007learning,
    title={Learning to Rank: From Pairwise Approach to Listwise Approach},
    author={Cao, Zhe and Qin, Tao and Liu, Tie-Yan and Tsai, Ming-Feng and Li, Hang},
    booktitle={Proceedings of the 24th international conference on Machine learning},
    pages={129--136},
    year={2007}
}