jayshah5696/er2-ce-bge-reranker-m3-ft

CrossEncoder based on BAAI/bge-reranker-v2-m3

This is a Cross Encoder model finetuned from BAAI/bge-reranker-v2-m3 on the entity-resolution-ce-pairs-v2 dataset 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: BAAI/bge-reranker-v2-m3
• Maximum Sequence Length: 8192 tokens
• Number of Output Labels: 1 label
• Training Dataset:
  • entity-resolution-ce-pairs-v2

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("jayshah5696/er2-ce-bge-reranker-m3-ft")
# Get scores for pairs of texts
pairs = [
    ["{'entity_id': 'E2-013271', 'first_name': 'Ramswaroop', 'last_name': 'Kakoti', 'company': 'KDS APPLIANCES RBS', 'title': 'Highway Painter Helper', 'email': 'ramswaroop_kakoti@kdsappliancesrbs.com', 'country': 'GB', 'ethnicity_group': 'indian', 'name_script': 'latin'}", "{'entity_id': 'E2-013271', 'first_name': 'Ramsaroop', 'last_name': 'Kakoti', 'company': 'KDS APPLIANCES RBS', 'title': 'Highway Painter Helper', 'email': 'ramswaroop_kakoti@kdsappliancesrbs.com', 'country': 'GB', 'ethnicity_group': 'indian', 'name_script': 'latin'}"],
    ["{'entity_id': 'E2-036160', 'first_name': 'Lole', 'last_name': 'Camaney', 'company': 'Monroe Capital CFO I Ltd.', 'title': 'Computer Technology Trainer', 'email': 'lole@monroecapitalcfoi.com', 'country': 'KY', 'ethnicity_group': 'hispanic', 'name_script': 'latin'}", "{'entity_id': 'E2-045966', 'first_name': 'Lali', 'last_name': 'Roig', 'company': 'AQUA AEREM (DBH) PTY LTD', 'title': 'Clamp Operator', 'email': 'laliroig@aquaaeremdbh.com', 'country': 'AU', 'ethnicity_group': 'other', 'name_script': 'latin'}"],
    ["{'entity_id': 'E2-045272', 'first_name': 'Rim', 'last_name': 'Hue', 'company': 'Edizioni ZYX Music SRL', 'title': 'Radiologic Technology Teacher', 'email': 'rim_hue@edizionizyxmusic.com', 'country': 'IT', 'ethnicity_group': 'other', 'name_script': 'latin'}", "{'entity_id': 'E2-043277', 'first_name': 'Less', 'last_name': 'Hue', 'company': 'EXERGY S.P.A. - IN LIQUIDAZIONE', 'title': 'Personnel Quality Assurance Auditor', 'email': 'less_hue@exergyspainliquidazione.com', 'country': 'IT', 'ethnicity_group': 'other', 'name_script': 'latin'}"],
    ["{'entity_id': 'E2-010159', 'first_name': 'Brianna', 'last_name': 'Cruzmartinez', 'company': 'S PARIKH AND CO', 'title': 'General Manager (GM)', 'email': 'bcruzmartinez@sparikhand.com', 'country': 'IN', 'ethnicity_group': 'us_uk_english', 'name_script': 'latin'}", "{'entity_id': 'E2-010442', 'first_name': 'Brian', 'last_name': 'Alves', 'company': '中州國際證券有限公司', 'title': 'Soaking Pits Supervisor', 'email': 'brian.alves@example.com', 'country': 'HK', 'ethnicity_group': 'us_uk_english', 'name_script': 'latin'}"],
    ["{'entity_id': 'E2-015732', 'first_name': 'Ikka', 'last_name': 'Dhurve', 'company': 'Tiburon Unternehmensaufbau GmbH', 'title': 'Freight Elevator Operator', 'email': 'idhurve@tiburonunternehmensaufbau.com', 'country': 'DE', 'ethnicity_group': 'indian', 'name_script': 'latin'}", "{'entity_id': 'E2-015732', 'first_name': 'Ikka', 'last_name': 'Dhurve', 'company': 'Tiburon Unternehmensaufbau', 'title': 'Freight Elevator Operator', 'email': 'idhurve@tiburonunternehmensaufbau.com', 'country': 'DE', 'ethnicity_group': 'indian', 'name_script': 'latin'}"],
]
scores = model.predict(pairs)
print(scores.shape)
# (5,)

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    "{'entity_id': 'E2-013271', 'first_name': 'Ramswaroop', 'last_name': 'Kakoti', 'company': 'KDS APPLIANCES RBS', 'title': 'Highway Painter Helper', 'email': 'ramswaroop_kakoti@kdsappliancesrbs.com', 'country': 'GB', 'ethnicity_group': 'indian', 'name_script': 'latin'}",
    [
        "{'entity_id': 'E2-013271', 'first_name': 'Ramsaroop', 'last_name': 'Kakoti', 'company': 'KDS APPLIANCES RBS', 'title': 'Highway Painter Helper', 'email': 'ramswaroop_kakoti@kdsappliancesrbs.com', 'country': 'GB', 'ethnicity_group': 'indian', 'name_script': 'latin'}",
        "{'entity_id': 'E2-045966', 'first_name': 'Lali', 'last_name': 'Roig', 'company': 'AQUA AEREM (DBH) PTY LTD', 'title': 'Clamp Operator', 'email': 'laliroig@aquaaeremdbh.com', 'country': 'AU', 'ethnicity_group': 'other', 'name_script': 'latin'}",
        "{'entity_id': 'E2-043277', 'first_name': 'Less', 'last_name': 'Hue', 'company': 'EXERGY S.P.A. - IN LIQUIDAZIONE', 'title': 'Personnel Quality Assurance Auditor', 'email': 'less_hue@exergyspainliquidazione.com', 'country': 'IT', 'ethnicity_group': 'other', 'name_script': 'latin'}",
        "{'entity_id': 'E2-010442', 'first_name': 'Brian', 'last_name': 'Alves', 'company': '中州國際證券有限公司', 'title': 'Soaking Pits Supervisor', 'email': 'brian.alves@example.com', 'country': 'HK', 'ethnicity_group': 'us_uk_english', 'name_script': 'latin'}",
        "{'entity_id': 'E2-015732', 'first_name': 'Ikka', 'last_name': 'Dhurve', 'company': 'Tiburon Unternehmensaufbau', 'title': 'Freight Elevator Operator', 'email': 'idhurve@tiburonunternehmensaufbau.com', 'country': 'DE', 'ethnicity_group': 'indian', 'name_script': 'latin'}",
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Evaluation

Metrics

Cross Encoder Binary Classification

• Dataset: bge_reranker_m3
• Evaluated with CEBinaryClassificationEvaluator

Metric	Value
accuracy	0.9992
accuracy_threshold	0.9999
f1	0.9992
f1_threshold	0.9999
precision	1.0
recall	0.9984
average_precision	0.9994

Training Details

Training Dataset

entity-resolution-ce-pairs-v2

• Dataset: entity-resolution-ce-pairs-v2 at 8b9c268
• Size: 33,290 training samples
• Columns: label, text_a, and text_b
• Approximate statistics based on the first 1000 samples:

  	label	text_a	text_b
  type	int	string	string
  details	0: ~52.80% 1: ~47.20%	min: 210 characters mean: 269.24 characters max: 481 characters	min: 200 characters mean: 268.38 characters max: 731 characters

• Samples:

  label	text_a	text_b
  0	{'entity_id': 'E2-011894', 'first_name': 'Noel', 'last_name': 'Muldowney', 'company': 'HIPP UK LIMITED', 'title': 'Flight Teacher', 'email': 'nmuldowney@hipp.com', 'country': 'GB', 'ethnicity_group': 'us_uk_english', 'name_script': 'latin'}	{'entity_id': 'E2-003538', 'first_name': 'Neal', 'last_name': 'Ellenberger', 'company': 'Hawthorns Park Ltd', 'title': 'Rip Sawyer', 'email': 'nealellenberger@hawthornspark.com', 'country': 'GB', 'ethnicity_group': 'us_uk_english', 'name_script': 'latin'}
  1	{'entity_id': 'E2-024730', 'first_name': '正平', 'last_name': 'Yan', 'company': 'TRUSTEE OF M I L WADLEY DECEASED WILL TRUST', 'title': 'Radial Drill Press Operator', 'email': '正平yan@trusteeofmilwadleydeceasedwilltrust.com', 'country': 'GB', 'ethnicity_group': 'chinese', 'name_script': 'cjk'}	{'entity_id': 'E2-024730', 'first_name': '正平', 'last_name': 'Yan', 'company': 'TRUSTEE OF M I L WADLEY DECEASED WILL TRUST Inc', 'title': 'Radial Drill Press Operator', 'email': '正平yan@trusteeofmilwadleydeceasedwilltrust.com', 'country': 'GB', 'ethnicity_group': 'chinese', 'name_script': 'cjk'}
  0	{'entity_id': 'E2-044064', 'first_name': 'Gîte', 'last_name': 'Gi', 'company': 'VECMEDICAL SPAIN SL', 'title': 'Shellfish Bed Worker', 'email': 'gîte.gi@vecmedicalspain.com', 'country': 'ES', 'ethnicity_group': 'other', 'name_script': 'latin'}	{'entity_id': 'E2-030776', 'first_name': 'Anyelo', 'last_name': 'Dela Cruz', 'company': 'AVURNAVE S.L', 'title': 'Nail Assembly Machine Operator', 'email': 'adela cruz@avurnave.com', 'country': 'ES', 'ethnicity_group': 'hispanic', 'name_script': 'latin'}

• Loss: BinaryCrossEntropyLoss with these parameters:

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

Evaluation Dataset

entity-resolution-ce-pairs-v2

• Dataset: entity-resolution-ce-pairs-v2 at 8b9c268
• Size: 3,802 evaluation samples
• Columns: label, text_a, and text_b
• Approximate statistics based on the first 1000 samples:

  	label	text_a	text_b
  type	int	string	string
  details	0: ~49.10% 1: ~50.90%	min: 209 characters mean: 268.91 characters max: 596 characters	min: 197 characters mean: 267.52 characters max: 596 characters

• Samples:

  label	text_a	text_b
  1	{'entity_id': 'E2-013271', 'first_name': 'Ramswaroop', 'last_name': 'Kakoti', 'company': 'KDS APPLIANCES RBS', 'title': 'Highway Painter Helper', 'email': 'ramswaroop_kakoti@kdsappliancesrbs.com', 'country': 'GB', 'ethnicity_group': 'indian', 'name_script': 'latin'}	{'entity_id': 'E2-013271', 'first_name': 'Ramsaroop', 'last_name': 'Kakoti', 'company': 'KDS APPLIANCES RBS', 'title': 'Highway Painter Helper', 'email': 'ramswaroop_kakoti@kdsappliancesrbs.com', 'country': 'GB', 'ethnicity_group': 'indian', 'name_script': 'latin'}
  0	{'entity_id': 'E2-036160', 'first_name': 'Lole', 'last_name': 'Camaney', 'company': 'Monroe Capital CFO I Ltd.', 'title': 'Computer Technology Trainer', 'email': 'lole@monroecapitalcfoi.com', 'country': 'KY', 'ethnicity_group': 'hispanic', 'name_script': 'latin'}	{'entity_id': 'E2-045966', 'first_name': 'Lali', 'last_name': 'Roig', 'company': 'AQUA AEREM (DBH) PTY LTD', 'title': 'Clamp Operator', 'email': 'laliroig@aquaaeremdbh.com', 'country': 'AU', 'ethnicity_group': 'other', 'name_script': 'latin'}
  0	{'entity_id': 'E2-045272', 'first_name': 'Rim', 'last_name': 'Hue', 'company': 'Edizioni ZYX Music SRL', 'title': 'Radiologic Technology Teacher', 'email': 'rim_hue@edizionizyxmusic.com', 'country': 'IT', 'ethnicity_group': 'other', 'name_script': 'latin'}	{'entity_id': 'E2-043277', 'first_name': 'Less', 'last_name': 'Hue', 'company': 'EXERGY S.P.A. - IN LIQUIDAZIONE', 'title': 'Personnel Quality Assurance Auditor', 'email': 'less_hue@exergyspainliquidazione.com', 'country': 'IT', 'ethnicity_group': 'other', 'name_script': 'latin'}

• Loss: BinaryCrossEntropyLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: epoch
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 64
• learning_rate: 2e-05
• num_train_epochs: 5
• 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: epoch
• prediction_loss_only: True
• per_device_train_batch_size: 32
• 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: 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: 5
• 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
• 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: batch_sampler
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	Validation Loss	bge_reranker_m3_average_precision
0.0961	100	0.0224	-	-
0.1921	200	0.0057	-	-
0.2882	300	0.0182	-	-
0.3842	400	0.0167	-	-
0.4803	500	0.0097	-	-
0.5764	600	0.0095	-	-
0.6724	700	0.0093	-	-
0.7685	800	0.0049	-	-
0.8646	900	0.0091	-	-
0.9606	1000	0.0144	-	-
1.0	1041	-	0.0059	0.9999
1.0567	1100	0.007	-	-
1.1527	1200	0.0188	-	-
1.2488	1300	0.0071	-	-
1.3449	1400	0.0144	-	-
1.4409	1500	0.01	-	-
1.5370	1600	0.0064	-	-
1.6330	1700	0.0094	-	-
1.7291	1800	0.0183	-	-
1.8252	1900	0.0069	-	-
1.9212	2000	0.0136	-	-
2.0	2082	-	0.0054	0.9999
2.0173	2100	0.0087	-	-
2.1134	2200	0.0074	-	-
2.2094	2300	0.0094	-	-
2.3055	2400	0.0028	-	-
2.4015	2500	0.0142	-	-
2.4976	2600	0.0178	-	-
2.5937	2700	0.0129	-	-
2.6897	2800	0.0135	-	-
2.7858	2900	0.007	-	-
2.8818	3000	0.0048	-	-
2.9779	3100	0.0115	-	-
3.0	3123	-	0.0060	0.9998
3.0740	3200	0.0122	-	-
3.1700	3300	0.0063	-	-
3.2661	3400	0.0105	-	-
3.3622	3500	0.0088	-	-
3.4582	3600	0.0165	-	-
3.5543	3700	0.0028	-	-
3.6503	3800	0.0129	-	-
3.7464	3900	0.0017	-	-
3.8425	4000	0.0047	-	-
3.9385	4100	0.0053	-	-
4.0	4164	-	0.0093	0.9995
4.0346	4200	0.0033	-	-
4.1306	4300	0.0034	-	-
4.2267	4400	0.0002	-	-
4.3228	4500	0.0025	-	-
4.4188	4600	0.0045	-	-
4.5149	4700	0.0132	-	-
4.6110	4800	0.0004	-	-
4.7070	4900	0.0043	-	-
4.8031	5000	0.0031	-	-
4.8991	5100	0.0001	-	-
4.9952	5200	0.001	-	-
5.0	5205	-	0.0108	0.9994

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.11.12
• Sentence Transformers: 5.0.0
• Transformers: 4.49.0
• PyTorch: 2.6.0+cu124
• Accelerate: 1.13.0
• Datasets: 4.8.3
• Tokenizers: 0.21.4

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",
}