Hritikmore/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("Hritikmore/bge-base-financial-matryoshka")
# Run inference
sentences = [
    'The effective duration of our total AFS and HTM investments securities as of December 31, 2023 is approximately 3.9 years.',
    'What are the effective durations of the total Available-for-Sale (AFS) and Held-to-Maturity (HTM) investment securities as of December 31, 2023?',
    'What was the net unit growth percentage for Hilton in the year ended December 31, 2023?',
]
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	0.7286
cosine_accuracy@3	0.8486
cosine_accuracy@5	0.8886
cosine_accuracy@10	0.9214
cosine_precision@1	0.7286
cosine_precision@3	0.2829
cosine_precision@5	0.1777
cosine_precision@10	0.0921
cosine_recall@1	0.7286
cosine_recall@3	0.8486
cosine_recall@5	0.8886
cosine_recall@10	0.9214
cosine_ndcg@10	0.8274
cosine_mrr@10	0.797
cosine_map@100	0.7999

Information Retrieval

• Dataset: dim_512
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.72
cosine_accuracy@3	0.8443
cosine_accuracy@5	0.8786
cosine_accuracy@10	0.92
cosine_precision@1	0.72
cosine_precision@3	0.2814
cosine_precision@5	0.1757
cosine_precision@10	0.092
cosine_recall@1	0.72
cosine_recall@3	0.8443
cosine_recall@5	0.8786
cosine_recall@10	0.92
cosine_ndcg@10	0.8214
cosine_mrr@10	0.7897
cosine_map@100	0.7927

Information Retrieval

• Dataset: dim_256
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.7214
cosine_accuracy@3	0.8386
cosine_accuracy@5	0.8743
cosine_accuracy@10	0.9129
cosine_precision@1	0.7214
cosine_precision@3	0.2795
cosine_precision@5	0.1749
cosine_precision@10	0.0913
cosine_recall@1	0.7214
cosine_recall@3	0.8386
cosine_recall@5	0.8743
cosine_recall@10	0.9129
cosine_ndcg@10	0.8191
cosine_mrr@10	0.7889
cosine_map@100	0.7921

Information Retrieval

• Dataset: dim_128
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6971
cosine_accuracy@3	0.8329
cosine_accuracy@5	0.8671
cosine_accuracy@10	0.9057
cosine_precision@1	0.6971
cosine_precision@3	0.2776
cosine_precision@5	0.1734
cosine_precision@10	0.0906
cosine_recall@1	0.6971
cosine_recall@3	0.8329
cosine_recall@5	0.8671
cosine_recall@10	0.9057
cosine_ndcg@10	0.8054
cosine_mrr@10	0.7729
cosine_map@100	0.7762

Information Retrieval

• Dataset: dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6614
cosine_accuracy@3	0.7986
cosine_accuracy@5	0.8443
cosine_accuracy@10	0.8814
cosine_precision@1	0.6614
cosine_precision@3	0.2662
cosine_precision@5	0.1689
cosine_precision@10	0.0881
cosine_recall@1	0.6614
cosine_recall@3	0.7986
cosine_recall@5	0.8443
cosine_recall@10	0.8814
cosine_ndcg@10	0.7729
cosine_mrr@10	0.7378
cosine_map@100	0.7418

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: 2 tokens mean: 45.87 tokens max: 272 tokens	min: 2 tokens mean: 20.43 tokens max: 41 tokens

• Samples:

  positive	anchor
  Significant judgment is required in evaluating our tax positions and during the ordinary course of business, there are many transactions and calculations for which the ultimate tax settlement is uncertain. As a result, we recognize the effect of this uncertainty on our tax attributes or taxes payable based on our estimates of the eventual outcome.	Why might the company's tax settlements vary?
  OPSUMIT is used for the treatment of pediatric pulmonary arterial hypertension.	What medical condition does OPSUMIT treat?
  Tangible equity ratios and tangible book value per share of common stock are non-GAAP financial measures. For more information on these ratios and corresponding reconciliations to GAAP financial measures, see Supplemental Financial Data and Non-GAAP Reconciliations.	What is the tangible equity ratio considered according to standard financial measures?

• 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
• gradient_accumulation_steps: 16
• learning_rate: 2e-05
• num_train_epochs: 2
• lr_scheduler_type: cosine
• warmup_ratio: 0.1
• tf32: False
• 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: 8
• per_device_eval_batch_size: 8
• 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: 2
• 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: False
• fp16: False
• fp16_opt_level: O1
• half_precision_backend: auto
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: False
• 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@100	dim_256_cosine_map@100	dim_512_cosine_map@100	dim_64_cosine_map@100	dim_768_cosine_map@100
0.2030	10	0.7168	-	-	-	-	-
0.4061	20	0.3345	-	-	-	-	-
0.6091	30	0.2234	-	-	-	-	-
0.8122	40	0.2126	-	-	-	-	-
0.9949	49	-	0.7796	0.7844	0.7905	0.7293	0.7973
1.0152	50	0.2301	-	-	-	-	-
1.2183	60	0.1595	-	-	-	-	-
1.4213	70	0.1082	-	-	-	-	-
1.6244	80	0.0911	-	-	-	-	-
1.8274	90	0.1068	-	-	-	-	-
1.9898	98	-	0.7762	0.7921	0.7927	0.7418	0.7999

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.0.1
• Transformers: 4.41.2
• PyTorch: 2.1.2+cu121
• Accelerate: 0.31.0
• 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}
}