README.md

metadata

language:
  - en
license: apache-2.0
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:6300
  - loss:MatryoshkaLoss
  - loss:MultipleNegativesRankingLoss
base_model: BAAI/bge-base-en-v1.5
widget:
  - source_sentence: >-
      The Innovative Medicine segment is focused on the following therapeutic
      areas: Immunology, Infectious diseases, Neuroscience, Oncology, Pulmonary
      Hypertension, and Cardiovascular and Metabolic diseases.
    sentences:
      - >-
        What was the primary reason for the decrease in adjusted operating
        income in 2023?
      - >-
        What therapeutic areas does the Innovative Medicine segment of Johnson &
        Johnson focus on?
      - >-
        What was the remaining budget for the September 2022 Repurchase Program
        as of January 28, 2023?
  - source_sentence: >-
      It may be necessary in the future to seek or renew licenses relating to
      various aspects of the Company’s products, processes and services. While
      the Company has generally been able to obtain such licenses on
      commercially reasonable terms in the past, there is no guarantee that such
      licenses could be obtained in the future on reasonable terms or at all.
    sentences:
      - >-
        What was the percentage change in total earning assets from the previous
        year as reported in 2023?
      - >-
        What is Apple's approach to licenses for intellectual property owned by
        third parties used in its products and services?
      - >-
        Why did the Ontario class action related to the 2017 cybersecurity
        incident progress differently than other cases?
  - source_sentence: >-
      Assets and liabilities measured at fair value on a nonrecurring basis in
      the consolidated financial statements include items such as property,
      plant and equipment, ROU assets, goodwill and other intangible assets,
      equity and other investments and other assets. These are measured at fair
      value if determined to be impaired.
    sentences:
      - >-
        How are assets and liabilities that are measured at fair value on a
        nonrecurring basis identified in the financial statements?
      - >-
        How is goodwill reviewed for impairment in a company, and what methods
        are used to determine the fair value of reporting units?
      - >-
        What diversity and inclusion goals has Goldman Sachs set for its
        workforce by 2025?
  - source_sentence: >-
      Based on management’s allocation decision, the portion of the Credit
      Facility available to ME&T as of December 31, 2023 was $2.75 billion.
    sentences:
      - >-
        What were the components of the increase in costs related to operating
        channels in 2023?
      - >-
        What are the goals of American Express’s balance sheet management
        strategy?
      - >-
        How much of the Credit Facility was available to ME&T as of December 31,
        2023?
  - source_sentence: Item 8 is labeled 'Financial Statements and Supplementary Data.'
    sentences:
      - What section of the document is labeled 'Item 8'?
      - >-
        For comprehensive information on a company's legal matters, which part
        of the financial statement should one consult?
      - What was the return on average common stockholders’ equity for 2023?
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
  - cosine_accuracy@1
  - cosine_accuracy@3
  - cosine_accuracy@5
  - cosine_accuracy@10
  - cosine_precision@1
  - cosine_precision@3
  - cosine_precision@5
  - cosine_precision@10
  - cosine_recall@1
  - cosine_recall@3
  - cosine_recall@5
  - cosine_recall@10
  - cosine_ndcg@10
  - cosine_mrr@10
  - cosine_map@100
model-index:
  - name: BGE base Financial Matryoshka
    results:
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 768
          type: dim_768
        metrics:
          - type: cosine_accuracy@1
            value: 0.6928571428571428
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.8257142857142857
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.8671428571428571
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.9071428571428571
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.6928571428571428
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.2752380952380953
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.1734285714285714
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.0907142857142857
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.6928571428571428
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.8257142857142857
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.8671428571428571
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.9071428571428571
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.8015678007585516
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.7675442176870747
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.7711683558124478
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 512
          type: dim_512
        metrics:
          - type: cosine_accuracy@1
            value: 0.6914285714285714
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.8257142857142857
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.8671428571428571
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.9071428571428571
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.6914285714285714
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.2752380952380952
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.1734285714285714
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.0907142857142857
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.6914285714285714
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.8257142857142857
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.8671428571428571
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.9071428571428571
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.8009375601369785
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.7666672335600906
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.7701113420260945
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 256
          type: dim_256
        metrics:
          - type: cosine_accuracy@1
            value: 0.6871428571428572
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.8242857142857143
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.8585714285714285
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.9028571428571428
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.6871428571428572
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.2747619047619047
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.1717142857142857
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.09028571428571427
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.6871428571428572
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.8242857142857143
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.8585714285714285
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.9028571428571428
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.7965630325935761
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.7623344671201813
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.7659656636117955
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 128
          type: dim_128
        metrics:
          - type: cosine_accuracy@1
            value: 0.6728571428571428
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.8085714285714286
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.8485714285714285
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.8871428571428571
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.6728571428571428
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.26952380952380944
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.16971428571428568
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.0887142857142857
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.6728571428571428
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.8085714285714286
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.8485714285714285
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.8871428571428571
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.7820355932651222
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.7480856009070294
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.7523134135641188
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 64
          type: dim_64
        metrics:
          - type: cosine_accuracy@1
            value: 0.6357142857142857
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.7685714285714286
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.8128571428571428
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.86
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.6357142857142857
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.2561904761904762
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.16257142857142853
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.086
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.6357142857142857
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.7685714285714286
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.8128571428571428
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.86
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.7472648621107045
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.7111729024943308
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.7168773691247933
            name: Cosine Map@100

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 dimensions
• 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("schawla2/bge-base-financial-matryoshka")
# Run inference
sentences = [
    "Item 8 is labeled 'Financial Statements and Supplementary Data.'",
    "What section of the document is labeled 'Item 8'?",
    "For comprehensive information on a company's legal matters, which part of the financial statement should one consult?",
]
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

• Datasets: dim_768, dim_512, dim_256, dim_128 and dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	dim_768	dim_512	dim_256	dim_128	dim_64
cosine_accuracy@1	0.6929	0.6914	0.6871	0.6729	0.6357
cosine_accuracy@3	0.8257	0.8257	0.8243	0.8086	0.7686
cosine_accuracy@5	0.8671	0.8671	0.8586	0.8486	0.8129
cosine_accuracy@10	0.9071	0.9071	0.9029	0.8871	0.86
cosine_precision@1	0.6929	0.6914	0.6871	0.6729	0.6357
cosine_precision@3	0.2752	0.2752	0.2748	0.2695	0.2562
cosine_precision@5	0.1734	0.1734	0.1717	0.1697	0.1626
cosine_precision@10	0.0907	0.0907	0.0903	0.0887	0.086
cosine_recall@1	0.6929	0.6914	0.6871	0.6729	0.6357
cosine_recall@3	0.8257	0.8257	0.8243	0.8086	0.7686
cosine_recall@5	0.8671	0.8671	0.8586	0.8486	0.8129
cosine_recall@10	0.9071	0.9071	0.9029	0.8871	0.86
cosine_ndcg@10	0.8016	0.8009	0.7966	0.782	0.7473
cosine_mrr@10	0.7675	0.7667	0.7623	0.7481	0.7112
cosine_map@100	0.7712	0.7701	0.766	0.7523	0.7169

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: 4 tokens mean: 45.97 tokens max: 326 tokens	min: 7 tokens mean: 20.57 tokens max: 46 tokens

• Samples:

  positive	anchor
  In 2023, Delta took delivery of 43 aircraft.	How many new aircraft did Delta Air Lines take delivery of in 2023?
  Item 8 incorporates pages 44 through 121 of IBM’s 2023 Annual Report to Stockholders by reference.	What sections of IBM's 2023 Annual Report are incorporated into Item 8 of the Form 10-K?
  Total borrowings at the end of 2023 were $29.3 billion.	What was the total amount of debt the Company had at the end of 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
• 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: 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: None
• hub_always_push: False
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• include_inputs_for_metrics: False
• include_for_metrics: []
• 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
• use_liger_kernel: False
• eval_use_gather_object: False
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	dim_768_cosine_ndcg@10	dim_512_cosine_ndcg@10	dim_256_cosine_ndcg@10	dim_128_cosine_ndcg@10	dim_64_cosine_ndcg@10
0.8122	10	24.5441	-	-	-	-	-
1.0	13	-	0.7920	0.7932	0.7878	0.7699	0.7305
1.5685	20	10.1942	-	-	-	-	-
2.0	26	-	0.7998	0.7988	0.7968	0.7818	0.7443
2.3249	30	6.8787	-	-	-	-	-
3.0	39	-	0.8014	0.8015	0.7976	0.7825	0.7503
3.0812	40	6.1782	-	-	-	-	-
3.7310	48	-	0.8016	0.8009	0.7966	0.7820	0.7473

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.16
• Sentence Transformers: 3.3.1
• Transformers: 4.48.1
• PyTorch: 2.5.1+cu124
• Accelerate: 1.3.0
• Datasets: 3.3.2
• Tokenizers: 0.21.0

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}
}