rshn-krn/bge-large-legal-billing

SentenceTransformer based on BAAI/bge-large-en-v1.5

This is a sentence-transformers model finetuned from BAAI/bge-large-en-v1.5. It maps sentences & paragraphs to a 1024-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: BAAI/bge-large-en-v1.5
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 1024 dimensions
• Similarity Function: Cosine Similarity

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': True, 'architecture': 'BertModel'})
  (1): Pooling({'word_embedding_dimension': 1024, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  (2): Normalize()
)

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("rshn-krn/bge-large-legal-billing")
# Run inference
sentences = [
    'Carrier: Coaction Global, Inc.\nAttorney: Senior Counsel | Rate: $245/hr | Units: 0.1\nTask: L120 - Analysis/Strategy | Activity: A103 - Draft/revise\nNarrative: Draft Litigation Status Report regarding the deposition testimony of California Highway Patrol Officer A.S. Johnson with attention to Section I. Testimony, subsection G. Witness Statements, subsection 3. Tuna Taleni ',
    'Carrier: Coaction Global, Inc.\nAttorney: Senior Counsel | Rate: $245/hr | Units: 0.3\nTask: L120 - Analysis/Strategy | Activity: A103 - Draft/revise\nNarrative: Draft Litigation Status Report regarding the deposition testimony of California Highway Patrol Officer A.S. Johnson with attention to Section I. Testimony, subsection G. Witness Statements, subsection 4. Lumafale Oti ',
    'Carrier: Mitsui Sumitomo Marine Management, Inc.\nAttorney: Senior Counsel | Rate: $325/hr | Units: 0.1\nTask: L110 - Fact Investigation/Development | Activity: A107 - Communicat/OUT\nNarrative: Email to current counsel for Google defendants regarding status of dismissal of non-related entities in our to evaluate the case for insured.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.9998, 0.4579],
#         [0.9998, 1.0000, 0.4567],
#         [0.4579, 0.4567, 1.0000]])

Training Details

Training Dataset

Unnamed Dataset

• Size: 28,621 training samples
• Columns: anchor and positive
• Approximate statistics based on the first 1000 samples:

  	anchor	positive
  type	string	string
  details	min: 50 tokens mean: 81.19 tokens max: 210 tokens	min: 44 tokens mean: 81.51 tokens max: 211 tokens

• Samples:

  anchor	positive
  Carrier: Golden Bear Insurance Company Attorney: Partner | Rate: $210/hr | Units: 0.3 Task: L120 - Analysis/Strategy | Activity: A104 - Review/analyze Narrative: Review and analysis of asserted claims with particular attention to elements necessary to establish and likely defenses available to rebut claims and related initial pleadings, investigation, and discovery to leverage same so as to strategically implement ideal defense strategy	Carrier: Golden Bear Insurance Company Attorney: Partner | Rate: $250/hr | Units: 0.1 Task: L440 - Other Trial Preparation and Support | Activity: A104 - Review/analyze Narrative: Analyze the notice from the civil division manager for Monmouth County ordering a trial in this matter in April of 2025
  Carrier: Aspen Specialty Insurance Company Attorney: Associate | Rate: $240/hr | Units: 0.1 Task: L110 - Fact Investigation/Development | Activity: A103 - Draft/revise Narrative: Draft and evaluate electronic correspondence to Ross McKissick, Plaintiff's counsel, discussing the nature and scope of Plaintiff's meet and confer issues for purposes of defense of same.	Carrier: Aspen Specialty Insurance Company Attorney: Associate | Rate: $240/hr | Units: 1.4 Task: L310 - Written Discovery | Activity: A103 - Draft/revise Narrative: Prepare Defendant's First Request for Production to Plaintiff.
  Carrier: Argo Group U.S. Attorney: Paralegal | Rate: $135/hr | Units: 0.2 Task: L110 - Fact Investigation/Development | Activity: A103 - Draft/revise Narrative: Draft r letter to American Pacific Concrete LLC to be in compliance with order issued by Arbitrator and to advise possible default can be entered if failure to respond.	Carrier: Argo Group U.S. Attorney: Paralegal | Rate: $135/hr | Units: 0.2 Task: L130 - Experts/Consultants | Activity: A108 - Communicat/MISC Narrative: Emails (4) to and from retained expert Jeff Ambrosia regarding meeting of experts of all parties to discuss landscaping issues in compliance with Arbitrator's order.

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim",
      "gather_across_devices": false,
      "directions": [
          "query_to_doc"
      ],
      "partition_mode": "joint",
      "hardness_mode": null,
      "hardness_strength": 0.0
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 32
• warmup_steps: 0.1
• gradient_accumulation_steps: 4
• fp16: True
• gradient_checkpointing: True

All Hyperparameters

Click to expand

• per_device_train_batch_size: 32
• num_train_epochs: 3
• max_steps: -1
• learning_rate: 5e-05
• lr_scheduler_type: linear
• lr_scheduler_kwargs: None
• warmup_steps: 0.1
• optim: adamw_torch_fused
• optim_args: None
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• optim_target_modules: None
• gradient_accumulation_steps: 4
• average_tokens_across_devices: True
• max_grad_norm: 1.0
• label_smoothing_factor: 0.0
• bf16: False
• fp16: True
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• gradient_checkpointing: True
• gradient_checkpointing_kwargs: None
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• use_liger_kernel: False
• liger_kernel_config: None
• use_cache: False
• neftune_noise_alpha: None
• torch_empty_cache_steps: None
• auto_find_batch_size: False
• log_on_each_node: True
• logging_nan_inf_filter: True
• include_num_input_tokens_seen: no
• log_level: passive
• log_level_replica: warning
• disable_tqdm: False
• project: huggingface
• trackio_space_id: trackio
• eval_strategy: no
• per_device_eval_batch_size: 8
• prediction_loss_only: True
• eval_on_start: False
• eval_do_concat_batches: True
• eval_use_gather_object: False
• eval_accumulation_steps: None
• include_for_metrics: []
• batch_eval_metrics: False
• save_only_model: False
• save_on_each_node: False
• enable_jit_checkpoint: False
• push_to_hub: False
• hub_private_repo: None
• hub_model_id: None
• hub_strategy: every_save
• hub_always_push: False
• hub_revision: None
• load_best_model_at_end: False
• ignore_data_skip: False
• restore_callback_states_from_checkpoint: False
• full_determinism: False
• seed: 42
• data_seed: None
• use_cpu: False
• 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
• dataloader_drop_last: False
• dataloader_num_workers: 0
• dataloader_pin_memory: True
• dataloader_persistent_workers: False
• dataloader_prefetch_factor: None
• remove_unused_columns: True
• label_names: None
• train_sampling_strategy: random
• length_column_name: length
• ddp_find_unused_parameters: None
• ddp_bucket_cap_mb: None
• ddp_broadcast_buffers: False
• ddp_backend: None
• ddp_timeout: 1800
• fsdp: []
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
• deepspeed: None
• debug: []
• skip_memory_metrics: True
• do_predict: False
• resume_from_checkpoint: None
• warmup_ratio: None
• local_rank: -1
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss
0.0447	10	1.5338
0.0894	20	0.4740
0.1341	30	0.4362
0.1788	40	0.3799
0.2235	50	0.4078
0.2682	60	0.3891
0.3128	70	0.4182
0.3575	80	0.3938
0.4022	90	0.4459
0.4469	100	0.4123
0.4916	110	0.3640
0.5363	120	0.4194
0.5810	130	0.3928
0.6257	140	0.4267
0.6704	150	0.4228
0.7151	160	0.4358
0.7598	170	0.4309
0.8045	180	0.4161
0.8492	190	0.4289
0.8939	200	0.4091
0.9385	210	0.3994
0.9832	220	0.4184
1.0268	230	0.4119
1.0715	240	0.4279
1.1162	250	0.3907
1.1609	260	0.4242
1.2056	270	0.4049
1.2503	280	0.3787
1.2950	290	0.4150
1.3397	300	0.4472
1.3844	310	0.3944
1.4291	320	0.4288
1.4737	330	0.3718
1.5184	340	0.4148
1.5631	350	0.4160
1.6078	360	0.3907
1.6525	370	0.3918
1.6972	380	0.3777
1.7419	390	0.4300
1.7866	400	0.3913
1.8313	410	0.4205
1.8760	420	0.3863
1.9207	430	0.4370
1.9654	440	0.4225
2.0089	450	0.4057
2.0536	460	0.3843
2.0983	470	0.4034
2.1430	480	0.4115
2.1877	490	0.4128
2.2324	500	0.4028
2.2771	510	0.4198
2.3218	520	0.3613
2.3665	530	0.4017
2.4112	540	0.3639
2.4559	550	0.3978
2.5006	560	0.3982
2.5453	570	0.4059
2.5899	580	0.4175
2.6346	590	0.4510
2.6793	600	0.4210
2.7240	610	0.4098
2.7687	620	0.4082
2.8134	630	0.3970
2.8581	640	0.3846
2.9028	650	0.4155
2.9475	660	0.4071
2.9922	670	0.4293

Framework Versions

• Python: 3.11.10
• Sentence Transformers: 5.3.0
• Transformers: 5.4.0
• PyTorch: 2.11.0+cu130
• Accelerate: 1.13.0
• Datasets: 4.8.4
• 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",
}

MultipleNegativesRankingLoss

@misc{oord2019representationlearningcontrastivepredictive,
      title={Representation Learning with Contrastive Predictive Coding},
      author={Aaron van den Oord and Yazhe Li and Oriol Vinyals},
      year={2019},
      eprint={1807.03748},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/1807.03748},
}