abinj30/bge-base-financial-matryoshka

BGE base Financial Matryoshka

This is a sentence-transformers model finetuned from BAAI/bge-base-en-v1.5 on the json dataset. 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
• Training Dataset:
  • json
• 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("abinj30/bge-base-financial-matryoshka")
# Run inference
sentences = [
    'Management assessed the effectiveness of the company’s internal control over financial reporting as of December 31, 2023. In making this assessment, we used the criteria set forth by the Committee of Sponsoring Organizations of the Treadway Commission (COSO) in Internal Control—Integrated Framework (2013).',
    'What criteria did Caterpillar Inc. use to assess the effectiveness of its internal control over financial reporting as of December 31, 2023?',
    'What are the primary components of U.S. sales volumes for Ford?',
]
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	0.69
cosine_accuracy@3	0.8229
cosine_accuracy@5	0.86
cosine_accuracy@10	0.9057
cosine_precision@1	0.69
cosine_precision@3	0.2743
cosine_precision@5	0.172
cosine_precision@10	0.0906
cosine_recall@1	0.69
cosine_recall@3	0.8229
cosine_recall@5	0.86
cosine_recall@10	0.9057
cosine_ndcg@10	0.799
cosine_mrr@10	0.7647
cosine_map@100	0.7683

Information Retrieval

• Dataset: dim_512
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6843
cosine_accuracy@3	0.8186
cosine_accuracy@5	0.8643
cosine_accuracy@10	0.91
cosine_precision@1	0.6843
cosine_precision@3	0.2729
cosine_precision@5	0.1729
cosine_precision@10	0.091
cosine_recall@1	0.6843
cosine_recall@3	0.8186
cosine_recall@5	0.8643
cosine_recall@10	0.91
cosine_ndcg@10	0.797
cosine_mrr@10	0.7608
cosine_map@100	0.7639

Information Retrieval

• Dataset: dim_256
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6871
cosine_accuracy@3	0.8171
cosine_accuracy@5	0.8571
cosine_accuracy@10	0.8957
cosine_precision@1	0.6871
cosine_precision@3	0.2724
cosine_precision@5	0.1714
cosine_precision@10	0.0896
cosine_recall@1	0.6871
cosine_recall@3	0.8171
cosine_recall@5	0.8571
cosine_recall@10	0.8957
cosine_ndcg@10	0.7921
cosine_mrr@10	0.7588
cosine_map@100	0.7626

Information Retrieval

• Dataset: dim_128
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6657
cosine_accuracy@3	0.8057
cosine_accuracy@5	0.8429
cosine_accuracy@10	0.8786
cosine_precision@1	0.6657
cosine_precision@3	0.2686
cosine_precision@5	0.1686
cosine_precision@10	0.0879
cosine_recall@1	0.6657
cosine_recall@3	0.8057
cosine_recall@5	0.8429
cosine_recall@10	0.8786
cosine_ndcg@10	0.7744
cosine_mrr@10	0.7407
cosine_map@100	0.745

Information Retrieval

• Dataset: dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6429
cosine_accuracy@3	0.7786
cosine_accuracy@5	0.8171
cosine_accuracy@10	0.8586
cosine_precision@1	0.6429
cosine_precision@3	0.2595
cosine_precision@5	0.1634
cosine_precision@10	0.0859
cosine_recall@1	0.6429
cosine_recall@3	0.7786
cosine_recall@5	0.8171
cosine_recall@10	0.8586
cosine_ndcg@10	0.7517
cosine_mrr@10	0.7174
cosine_map@100	0.7223

Training Details

Training Dataset

json

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

  	positive	anchor
  type	string	string
  details	min: 8 tokens mean: 44.33 tokens max: 289 tokens	min: 9 tokens mean: 20.43 tokens max: 46 tokens

• Samples:

  positive	anchor
  The Company defines fair value as the price received to transfer an asset or paid to transfer a liability in an orderly transaction between market participants at the measurement date. In accordance with ASC 820, Fair Value Measurements and Disclosures, the Company uses the fair value hierarchy which prioritizes the inputs used to measure fair value. The hierarchy gives the highest priority to unadjusted quoted prices in active markets for identical assets or liabilities (Level 1), observable inputs other than quoted prices (Level 2), and unobservable inputs (Level 3).	What is the role of Level 1, Level 2, and Level 3 inputs in the fair value hierarchy according to ASC 820?
  In the event of conversion of the Notes, if shares are delivered to the Company under the Capped Call Transactions, they will offset the dilutive effect of the shares that the Company would issue under the Notes.	What happens to the dilutive effect of shares issued under the Notes if shares are delivered to the Company under the Capped Call Transactions during the conversion?
  Marketing expenses increased $48.8 million to $759.2 million in the year ended December 31, 2023 compared to the year ended December 31, 2022.	How much did the marketing expenses increase 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
• 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_768_cosine_map@100	dim_512_cosine_map@100	dim_256_cosine_map@100	dim_128_cosine_map@100	dim_64_cosine_map@100
0.8122	10	1.5604	-	-	-	-	-
0.9746	12	-	0.7550	0.7549	0.7477	0.7288	0.6928
1.6244	20	0.6617	-	-	-	-	-
1.9492	24	-	0.7653	0.7628	0.7576	0.7425	0.7206
2.4365	30	0.4579	-	-	-	-	-
2.9239	36	-	0.77	0.7637	0.7636	0.7462	0.723
3.2487	40	0.3997	-	-	-	-	-
3.8985	48	-	0.7683	0.7639	0.7626	0.7450	0.7223

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.2.0
• Transformers: 4.41.2
• PyTorch: 2.2.0a0+6a974be
• Accelerate: 0.27.0
• 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}
}