Sailesh9999/bge-base-financial-matryoshka

BGE base Financial Matryoshka

This is a sentence-transformers model finetuned from BAAI/bge-base-en-v1.5. It maps sentences & paragraphs to a 768-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-base-en-v1.5
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 tokens
• Similarity Function: Cosine Similarity
• Language: en
• License: apache-2.0

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}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 768, '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("Sailesh9999/bge-base-financial-matryoshka")
# Run inference
sentences = [
    'Chipotle retains an independent third-party compensation consultant each year to conduct a pay equity analysis of its U.S. and Canadian workforce, including factors of pay such as grade level, tenure in role, and external market conditions like geographic location, to ensure consistency and equitable treatment among employees.',
    'How does Chipotle ensure pay equity among its employees?',
    'How can one locate information on legal proceedings within the Consolidated Financial Statements?',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Information Retrieval

• Dataset: dim_768
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6986
cosine_accuracy@3	0.8343
cosine_accuracy@5	0.8629
cosine_accuracy@10	0.9
cosine_precision@1	0.6986
cosine_precision@3	0.2781
cosine_precision@5	0.1726
cosine_precision@10	0.09
cosine_recall@1	0.6986
cosine_recall@3	0.8343
cosine_recall@5	0.8629
cosine_recall@10	0.9
cosine_ndcg@10	0.8029
cosine_mrr@10	0.7715
cosine_map@10	0.7715

Information Retrieval

• Dataset: dim_512
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6843
cosine_accuracy@3	0.8271
cosine_accuracy@5	0.8629
cosine_accuracy@10	0.8929
cosine_precision@1	0.6843
cosine_precision@3	0.2757
cosine_precision@5	0.1726
cosine_precision@10	0.0893
cosine_recall@1	0.6843
cosine_recall@3	0.8271
cosine_recall@5	0.8629
cosine_recall@10	0.8929
cosine_ndcg@10	0.7943
cosine_mrr@10	0.7621
cosine_map@10	0.7621

Information Retrieval

• Dataset: dim_256
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6871
cosine_accuracy@3	0.8157
cosine_accuracy@5	0.8614
cosine_accuracy@10	0.8929
cosine_precision@1	0.6871
cosine_precision@3	0.2719
cosine_precision@5	0.1723
cosine_precision@10	0.0893
cosine_recall@1	0.6871
cosine_recall@3	0.8157
cosine_recall@5	0.8614
cosine_recall@10	0.8929
cosine_ndcg@10	0.7936
cosine_mrr@10	0.7614
cosine_map@10	0.7614

Information Retrieval

• Dataset: dim_128
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6757
cosine_accuracy@3	0.8171
cosine_accuracy@5	0.8514
cosine_accuracy@10	0.8814
cosine_precision@1	0.6757
cosine_precision@3	0.2724
cosine_precision@5	0.1703
cosine_precision@10	0.0881
cosine_recall@1	0.6757
cosine_recall@3	0.8171
cosine_recall@5	0.8514
cosine_recall@10	0.8814
cosine_ndcg@10	0.7843
cosine_mrr@10	0.7526
cosine_map@10	0.7526

Information Retrieval

• Dataset: dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.64
cosine_accuracy@3	0.79
cosine_accuracy@5	0.8271
cosine_accuracy@10	0.87
cosine_precision@1	0.64
cosine_precision@3	0.2633
cosine_precision@5	0.1654
cosine_precision@10	0.087
cosine_recall@1	0.64
cosine_recall@3	0.79
cosine_recall@5	0.8271
cosine_recall@10	0.87
cosine_ndcg@10	0.7595
cosine_mrr@10	0.7237
cosine_map@10	0.7237

Training Details

Training Dataset

Unnamed Dataset

• Size: 6,300 training samples
• Columns: positive and anchor
• Approximate statistics based on the first 1000 samples:

  	positive	anchor
  type	string	string
  details	min: 7 tokens mean: 46.55 tokens max: 439 tokens	min: 9 tokens mean: 20.43 tokens max: 46 tokens

• Samples:

  positive	anchor
  Americas	$
  Item 1 Business typically includes detailed information about the organization's operations, the nature of the business, and its strategic direction.	What is the title of the section that potentially discusses the operations or nature of a business in a document?
  Operating expenses as a percentage of total revenues decreased to 15.3% in 2023 compared to 15.9% in 2022.	What was the operating expenses as a percentage of total revenues in 2023?

• Loss: MatryoshkaLoss with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          768,
          512,
          256,
          128,
          64
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: epoch
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 16
• gradient_accumulation_steps: 16
• learning_rate: 2e-05
• num_train_epochs: 4
• lr_scheduler_type: cosine
• warmup_ratio: 0.1
• bf16: True
• tf32: True
• load_best_model_at_end: True
• optim: adamw_torch_fused
• batch_sampler: no_duplicates

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: 16
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 16
• eval_accumulation_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: 4
• 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: None
• jit_mode_eval: False
• use_ipex: False
• bf16: True
• fp16: False
• fp16_opt_level: O1
• half_precision_backend: auto
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: True
• 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_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: False
• hub_always_push: False
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• include_inputs_for_metrics: False
• 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
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	dim_128_cosine_map@10	dim_256_cosine_map@10	dim_512_cosine_map@10	dim_64_cosine_map@10	dim_768_cosine_map@10
0.8122	10	1.5638	-	-	-	-	-
0.9746	12	-	0.7308	0.7547	0.7547	0.7004	0.7624
1.6244	20	0.6662	-	-	-	-	-
1.9492	24	-	0.7468	0.7586	0.7624	0.7195	0.7655
2.4365	30	0.4634	-	-	-	-	-
2.9239	36	-	0.7525	0.7620	0.7614	0.7237	0.7717
3.2487	40	0.387	-	-	-	-	-
3.8985	48	-	0.7526	0.7614	0.7621	0.7237	0.7715

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.9.18
• Sentence Transformers: 3.0.1
• Transformers: 4.41.2
• PyTorch: 2.1.2+cu121
• Accelerate: 0.29.3
• Datasets: 2.19.1
• Tokenizers: 0.19.1

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",
}

MatryoshkaLoss

@misc{kusupati2024matryoshka,
    title={Matryoshka Representation Learning}, 
    author={Aditya Kusupati and Gantavya Bhatt and Aniket Rege and Matthew Wallingford and Aditya Sinha and Vivek Ramanujan and William Howard-Snyder and Kaifeng Chen and Sham Kakade and Prateek Jain and Ali Farhadi},
    year={2024},
    eprint={2205.13147},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

MultipleNegativesRankingLoss

@misc{henderson2017efficient,
    title={Efficient Natural Language Response Suggestion for Smart Reply}, 
    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
    year={2017},
    eprint={1705.00652},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}