moritzglnr/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("moritzglnr/bge-base-financial-matryoshka")
# Run inference
sentences = [
    'The remaining amount available for borrowing under the Revolving Credit Facility as of December 31, 2023, was $2,245.2 million.',
    'What was the total amount available for borrowing under the Revolving Credit Facility at Iron Mountain as of December 31, 2023?',
    'What type of information is included in Note 13 of the Annual Report on Form 10-K?',
]
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	0.8243
cosine_accuracy@5	0.8557
cosine_accuracy@10	0.9057
cosine_precision@1	0.6829
cosine_precision@3	0.2748
cosine_precision@5	0.1711
cosine_precision@10	0.0906
cosine_recall@1	0.6829
cosine_recall@3	0.8243
cosine_recall@5	0.8557
cosine_recall@10	0.9057
cosine_ndcg@10	0.7964
cosine_mrr@10	0.7613
cosine_map@100	0.7649

Information Retrieval

• Dataset: dim_512
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.68
cosine_accuracy@3	0.8157
cosine_accuracy@5	0.8543
cosine_accuracy@10	0.9
cosine_precision@1	0.68
cosine_precision@3	0.2719
cosine_precision@5	0.1709
cosine_precision@10	0.09
cosine_recall@1	0.68
cosine_recall@3	0.8157
cosine_recall@5	0.8543
cosine_recall@10	0.9
cosine_ndcg@10	0.7912
cosine_mrr@10	0.7562
cosine_map@100	0.7601

Information Retrieval

• Dataset: dim_256
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.68
cosine_accuracy@3	0.8114
cosine_accuracy@5	0.8486
cosine_accuracy@10	0.8971
cosine_precision@1	0.68
cosine_precision@3	0.2705
cosine_precision@5	0.1697
cosine_precision@10	0.0897
cosine_recall@1	0.68
cosine_recall@3	0.8114
cosine_recall@5	0.8486
cosine_recall@10	0.8971
cosine_ndcg@10	0.7889
cosine_mrr@10	0.7542
cosine_map@100	0.758

Information Retrieval

• Dataset: dim_128
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6571
cosine_accuracy@3	0.79
cosine_accuracy@5	0.8286
cosine_accuracy@10	0.8857
cosine_precision@1	0.6571
cosine_precision@3	0.2633
cosine_precision@5	0.1657
cosine_precision@10	0.0886
cosine_recall@1	0.6571
cosine_recall@3	0.79
cosine_recall@5	0.8286
cosine_recall@10	0.8857
cosine_ndcg@10	0.7704
cosine_mrr@10	0.7336
cosine_map@100	0.7379

Information Retrieval

• Dataset: dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.62
cosine_accuracy@3	0.77
cosine_accuracy@5	0.8029
cosine_accuracy@10	0.85
cosine_precision@1	0.62
cosine_precision@3	0.2567
cosine_precision@5	0.1606
cosine_precision@10	0.085
cosine_recall@1	0.62
cosine_recall@3	0.77
cosine_recall@5	0.8029
cosine_recall@10	0.85
cosine_ndcg@10	0.7378
cosine_mrr@10	0.7016
cosine_map@100	0.7074

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: 46.27 tokens max: 326 tokens	min: 2 tokens mean: 20.87 tokens max: 51 tokens

• Samples:

  positive	anchor
  We utilize a full yield curve approach in the estimation of service and interest costs by applying the specific spot rates along the yield curve used in the determination of the benefit obligation to the relevant projected cash flows. This approach provides a more precise measurement of service and interest costs by improving the correlation between the projected cash flows to the corresponding spot rates along the yield curve. This approach does not affect the measurement of our pension and other post-retirement benefit liabilities but generally results in lower benefit expense in periods when the yield curve is upward sloping.	How does the use of a full yield curve approach in estimating pension costs affect the measurement of liabilities and expenses?
  Ending	8,134
  The company's capital expenditures for 2024 are expected to be approximately $5.7 billion.	How much does the company expect to spend on capital expenditures in 2024?

• 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	1.5661	-	-	-	-	-
0.9746	12	-	0.7151	0.7378	0.7443	0.6680	0.7546
1.6244	20	0.6602	-	-	-	-	-
1.9492	24	-	0.7326	0.7533	0.7564	0.7037	0.7640
2.4365	30	0.4675	-	-	-	-	-
2.9239	36	-	0.7384	0.7575	0.7601	0.7086	0.7643
3.2487	40	0.3891	-	-	-	-	-
3.8985	48	-	0.7379	0.758	0.7601	0.7074	0.7649

• 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: 0.32.1
• 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}
}