dpokhrel/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("dpokhrel/bge-base-financial-matryoshka")
# Run inference
sentences = [
    'CMS made significant changes to the structure of the hierarchical condition category model in version 28, which may impact risk adjustment factor scores for a larger percentage of Medicare Advantage beneficiaries and could result in changes to beneficiary RAF scores with or without a change in the patient’s health status.',
    'What significant regulatory change did CMS make to the hierarchical condition category model in its version 28?',
    'What is the primary method by which the company manages its cash, cash equivalents, and marketable securities?',
]
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	0.8443
cosine_accuracy@5	0.8814
cosine_accuracy@10	0.9271
cosine_precision@1	0.6986
cosine_precision@3	0.2814
cosine_precision@5	0.1763
cosine_precision@10	0.0927
cosine_recall@1	0.6986
cosine_recall@3	0.8443
cosine_recall@5	0.8814
cosine_recall@10	0.9271
cosine_ndcg@10	0.8157
cosine_mrr@10	0.7796
cosine_map@100	0.7822

Information Retrieval

• Dataset: dim_512
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.71
cosine_accuracy@3	0.8457
cosine_accuracy@5	0.8786
cosine_accuracy@10	0.9271
cosine_precision@1	0.71
cosine_precision@3	0.2819
cosine_precision@5	0.1757
cosine_precision@10	0.0927
cosine_recall@1	0.71
cosine_recall@3	0.8457
cosine_recall@5	0.8786
cosine_recall@10	0.9271
cosine_ndcg@10	0.8195
cosine_mrr@10	0.7849
cosine_map@100	0.7873

Information Retrieval

• Dataset: dim_256
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.7086
cosine_accuracy@3	0.8343
cosine_accuracy@5	0.8643
cosine_accuracy@10	0.9143
cosine_precision@1	0.7086
cosine_precision@3	0.2781
cosine_precision@5	0.1729
cosine_precision@10	0.0914
cosine_recall@1	0.7086
cosine_recall@3	0.8343
cosine_recall@5	0.8643
cosine_recall@10	0.9143
cosine_ndcg@10	0.8116
cosine_mrr@10	0.7788
cosine_map@100	0.7821

Information Retrieval

• Dataset: dim_128
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.69
cosine_accuracy@3	0.8271
cosine_accuracy@5	0.86
cosine_accuracy@10	0.91
cosine_precision@1	0.69
cosine_precision@3	0.2757
cosine_precision@5	0.172
cosine_precision@10	0.091
cosine_recall@1	0.69
cosine_recall@3	0.8271
cosine_recall@5	0.86
cosine_recall@10	0.91
cosine_ndcg@10	0.8014
cosine_mrr@10	0.7665
cosine_map@100	0.7699

Information Retrieval

• Dataset: dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6657
cosine_accuracy@3	0.79
cosine_accuracy@5	0.8286
cosine_accuracy@10	0.8857
cosine_precision@1	0.6657
cosine_precision@3	0.2633
cosine_precision@5	0.1657
cosine_precision@10	0.0886
cosine_recall@1	0.6657
cosine_recall@3	0.79
cosine_recall@5	0.8286
cosine_recall@10	0.8857
cosine_ndcg@10	0.7733
cosine_mrr@10	0.7375
cosine_map@100	0.7417

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: 10 tokens mean: 46.37 tokens max: 248 tokens	min: 7 tokens mean: 20.57 tokens max: 51 tokens

• Samples:

  positive	anchor
  Scenario analysis is used to quantify the impact of a specified event, including how the event impacts multiple risk factors simultaneously. For example, for sovereign stress testing, it calculates potential exposure related to sovereign positions as well as the corresponding debt, equity, and currency exposures that may be impacted by sovereign distress.	How does Goldman Sachs utilize scenario analysis in its risk management strategy?
  The company is involved in various other legal proceedings incidental to the conduct of our business, including, but not limited to, claims and allegations related to wage and hour violations, unlawful termination, employment practices, product liability, privacy and cybersecurity, environmental matters, and intellectual property rights or regulatory compliance.	What types of legal proceedings is the company currently involved in?
  In 2023, $505 million was utilized for common stock repurchases.	How much cash was utilized for common stock repurchases in the year ended December 31, 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
• half_precision_backend: cpu_amp
• 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
• 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: 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: cpu_amp
• 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_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
• eval_on_start: False
• eval_use_gather_object: 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
0.8122	10	1.5241	-	-	-	-	-
0.9746	12	-	0.7486	0.7656	0.7662	0.7108	0.7679
1.6244	20	0.658	-	-	-	-	-
1.9492	24	-	0.7656	0.7793	0.7843	0.7348	0.7798
2.4365	30	0.4743	-	-	-	-	-
2.9239	36	-	0.7683	0.7814	0.7859	0.7400	0.7812
3.2487	40	0.4241	-	-	-	-	-
3.8985	48	-	0.7699	0.7821	0.7873	0.7417	0.7822

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.11.5
• Sentence Transformers: 3.0.1
• Transformers: 4.43.4
• PyTorch: 2.4.0.dev20240607+cu118
• Accelerate: 0.32.0
• Datasets: 2.20.0
• 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}
}