xiaofengzi/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("xiaofengzi/bge-base-financial-matryoshka")
# Run inference
sentences = [
    'What is basic earnings per share based on?',
    'How is basic net income per share calculated?',
    "How did NIKE's fiscal 2023 revenue compare to its fiscal 2022 revenue?",
]
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	0.6829
cosine_accuracy@3	0.82
cosine_accuracy@5	0.8629
cosine_accuracy@10	0.91
cosine_precision@1	0.6829
cosine_precision@3	0.2733
cosine_precision@5	0.1726
cosine_precision@10	0.091
cosine_recall@1	0.6829
cosine_recall@3	0.82
cosine_recall@5	0.8629
cosine_recall@10	0.91
cosine_ndcg@10	0.7971
cosine_mrr@10	0.7608
cosine_map@100	0.7644

Information Retrieval

• Dataset: dim_512
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.68
cosine_accuracy@3	0.8186
cosine_accuracy@5	0.8657
cosine_accuracy@10	0.9114
cosine_precision@1	0.68
cosine_precision@3	0.2729
cosine_precision@5	0.1731
cosine_precision@10	0.0911
cosine_recall@1	0.68
cosine_recall@3	0.8186
cosine_recall@5	0.8657
cosine_recall@10	0.9114
cosine_ndcg@10	0.7955
cosine_mrr@10	0.7584
cosine_map@100	0.7618

Information Retrieval

• Dataset: dim_256
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6786
cosine_accuracy@3	0.81
cosine_accuracy@5	0.8571
cosine_accuracy@10	0.9143
cosine_precision@1	0.6786
cosine_precision@3	0.27
cosine_precision@5	0.1714
cosine_precision@10	0.0914
cosine_recall@1	0.6786
cosine_recall@3	0.81
cosine_recall@5	0.8571
cosine_recall@10	0.9143
cosine_ndcg@10	0.7955
cosine_mrr@10	0.7578
cosine_map@100	0.7606

Information Retrieval

• Dataset: dim_128
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6571
cosine_accuracy@3	0.7957
cosine_accuracy@5	0.8471
cosine_accuracy@10	0.9
cosine_precision@1	0.6571
cosine_precision@3	0.2652
cosine_precision@5	0.1694
cosine_precision@10	0.09
cosine_recall@1	0.6571
cosine_recall@3	0.7957
cosine_recall@5	0.8471
cosine_recall@10	0.9
cosine_ndcg@10	0.7771
cosine_mrr@10	0.7379
cosine_map@100	0.742

Information Retrieval

• Dataset: dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6343
cosine_accuracy@3	0.7643
cosine_accuracy@5	0.8057
cosine_accuracy@10	0.8657
cosine_precision@1	0.6343
cosine_precision@3	0.2548
cosine_precision@5	0.1611
cosine_precision@10	0.0866
cosine_recall@1	0.6343
cosine_recall@3	0.7643
cosine_recall@5	0.8057
cosine_recall@10	0.8657
cosine_ndcg@10	0.7467
cosine_mrr@10	0.7091
cosine_map@100	0.7142

Training Details

Training Dataset

Unnamed Dataset

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

  	anchor	positive
  type	string	string
  details	min: 7 tokens mean: 20.44 tokens max: 51 tokens	min: 6 tokens mean: 47.22 tokens max: 512 tokens

• Samples:

  anchor	positive
  How did the Energy & Transportation segment's sales and profit change in 2023?	Energy & Transportation's total sales were $28.001 billion in 2023, an increase of $4.249 billion, or 18... and profit was $4.936 billion in 2023, an increase of $1.627 billion, or 49 percent...
  In which segments were acquisitions made in 2022?	During 2022, acquisitions occurred in Workforce Solutions and USIS operating segments, and the International segment.
  What are the contents found on pages 163 to 309 in the document?	The Consolidated Financial Statements, together with the Notes thereto and the report thereon dated February 16, 2024, of PricewaterhouseCoopers LLP, appear on pages 163–309.

• 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@100	dim_256_cosine_map@100	dim_512_cosine_map@100	dim_64_cosine_map@100	dim_768_cosine_map@100
0.8122	10	1.5644	-	-	-	-	-
0.9746	12	-	0.7186	0.7399	0.7414	0.6757	0.7445
1.6244	20	0.6502	-	-	-	-	-
1.9492	24	-	0.7379	0.7544	0.7573	0.7069	0.7600
2.4365	30	0.434	-	-	-	-	-
2.9239	36	-	0.7426	0.7614	0.7616	0.7134	0.7634
3.2487	40	0.3627	-	-	-	-	-
3.8985	48	-	0.7420	0.7606	0.7618	0.7142	0.7644

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.4
• Sentence Transformers: 3.0.0
• Transformers: 4.41.2
• PyTorch: 2.6.0+cu118
• Accelerate: 1.6.0
• 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}
}