NickyNicky/bge-base-financial-matryoshka_test_3

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("NickyNicky/bge-base-financial-matryoshka")
# Run inference
sentences = [
    'Information on legal proceedings is included in Contact Email  PRIOR HISTORY: None PLACEHOLDER FOR ARBITRATION.',
    'Where can information about legal proceedings be found in the financial statements?',
    'What remaining authorization amount was available for share repurchases as of January 28, 2023?',
]
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	0.71
cosine_accuracy@3	0.8429
cosine_accuracy@5	0.8771
cosine_accuracy@10	0.9143
cosine_precision@1	0.71
cosine_precision@3	0.281
cosine_precision@5	0.1754
cosine_precision@10	0.0914
cosine_recall@1	0.71
cosine_recall@3	0.8429
cosine_recall@5	0.8771
cosine_recall@10	0.9143
cosine_ndcg@10	0.8152
cosine_mrr@10	0.7832
cosine_map@100	0.7867

Information Retrieval

• Dataset: dim_512
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.7029
cosine_accuracy@3	0.8457
cosine_accuracy@5	0.88
cosine_accuracy@10	0.9157
cosine_precision@1	0.7029
cosine_precision@3	0.2819
cosine_precision@5	0.176
cosine_precision@10	0.0916
cosine_recall@1	0.7029
cosine_recall@3	0.8457
cosine_recall@5	0.88
cosine_recall@10	0.9157
cosine_ndcg@10	0.8132
cosine_mrr@10	0.78
cosine_map@100	0.7833

Information Retrieval

• Dataset: dim_256
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6986
cosine_accuracy@3	0.8457
cosine_accuracy@5	0.8786
cosine_accuracy@10	0.9071
cosine_precision@1	0.6986
cosine_precision@3	0.2819
cosine_precision@5	0.1757
cosine_precision@10	0.0907
cosine_recall@1	0.6986
cosine_recall@3	0.8457
cosine_recall@5	0.8786
cosine_recall@10	0.9071
cosine_ndcg@10	0.8072
cosine_mrr@10	0.7746
cosine_map@100	0.7782

Information Retrieval

• Dataset: dim_128
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6914
cosine_accuracy@3	0.8429
cosine_accuracy@5	0.8714
cosine_accuracy@10	0.9057
cosine_precision@1	0.6914
cosine_precision@3	0.281
cosine_precision@5	0.1743
cosine_precision@10	0.0906
cosine_recall@1	0.6914
cosine_recall@3	0.8429
cosine_recall@5	0.8714
cosine_recall@10	0.9057
cosine_ndcg@10	0.8053
cosine_mrr@10	0.7726
cosine_map@100	0.7764

Information Retrieval

• Dataset: dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6757
cosine_accuracy@3	0.8114
cosine_accuracy@5	0.85
cosine_accuracy@10	0.8843
cosine_precision@1	0.6757
cosine_precision@3	0.2705
cosine_precision@5	0.17
cosine_precision@10	0.0884
cosine_recall@1	0.6757
cosine_recall@3	0.8114
cosine_recall@5	0.85
cosine_recall@10	0.8843
cosine_ndcg@10	0.7836
cosine_mrr@10	0.7509
cosine_map@100	0.7558

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: 4 tokens mean: 47.19 tokens max: 512 tokens	min: 7 tokens mean: 20.59 tokens max: 41 tokens

• Samples:

  positive	anchor
  For the year ended December 31, 2023, $305 million was recorded as a distribution against retained earnings for dividends.	How much in dividends was recorded against retained earnings in 2023?
  In February 2023, we announced a 10% increase in our quarterly cash dividend to $2.09 per share.	By how much did the company increase its quarterly cash dividend in February 2023?
  Depreciation and amortization totaled $4,856 as recorded in the financial statements.	How much did depreciation and amortization total to in the financial statements?

• 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: 40
• per_device_eval_batch_size: 16
• gradient_accumulation_steps: 16
• learning_rate: 2e-05
• num_train_epochs: 20
• lr_scheduler_type: cosine
• warmup_ratio: 0.1
• bf16: True
• tf32: 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: 40
• 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: 20
• 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: False
• 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
0.9114	9	-	0.7124	0.7361	0.7366	0.6672	0.7443
1.0127	10	2.0952	-	-	-	-	-
1.9241	19	-	0.7437	0.7561	0.7628	0.7172	0.7653
2.0253	20	1.1175	-	-	-	-	-
2.9367	29	-	0.7623	0.7733	0.7694	0.7288	0.7723
3.0380	30	0.6104	-	-	-	-	-
3.9494	39	-	0.7723	0.7746	0.7804	0.7405	0.7789
4.0506	40	0.4106	-	-	-	-	-
4.9620	49	-	0.7777	0.7759	0.7820	0.7475	0.7842
5.0633	50	0.314	-	-	-	-	-
5.9747	59	-	0.7802	0.7796	0.7856	0.7548	0.7839
6.0759	60	0.2423	-	-	-	-	-
6.9873	69	-	0.7756	0.7772	0.7834	0.7535	0.7818
7.0886	70	0.1962	-	-	-	-	-
8.0	79	-	0.7741	0.7774	0.7841	0.7551	0.7822
8.1013	80	0.1627	-	-	-	-	-
8.9114	88	-	0.7724	0.7752	0.7796	0.7528	0.7816
9.1139	90	0.1379	-	-	-	-	-
9.9241	98	-	0.7691	0.7782	0.7834	0.7559	0.7836
10.1266	100	0.1249	-	-	-	-	-
10.9367	108	-	0.7728	0.7802	0.7831	0.7536	0.7848
11.1392	110	0.1105	-	-	-	-	-
11.9494	118	-	0.7748	0.7785	0.7814	0.7558	0.7851
12.1519	120	0.1147	-	-	-	-	-
12.9620	128	-	0.7756	0.7788	0.7839	0.7550	0.7864
13.1646	130	0.098	-	-	-	-	-
13.9747	138	-	0.7767	0.7792	0.7828	0.7557	0.7873
14.1772	140	0.0927	-	-	-	-	-
14.9873	148	-	0.7758	0.7804	0.7847	0.7569	0.7892
15.1899	150	0.0921	-	-	-	-	-
16.0	158	-	0.7760	0.7794	0.7831	0.7551	0.7873
16.2025	160	0.0896	-	-	-	-	-
16.9114	167	-	0.7753	0.7799	0.7841	0.7570	0.7888
17.2152	170	0.0881	-	-	-	-	-
17.9241	177	-	0.7763	0.7787	0.7842	0.7561	0.7867
18.2278	180	0.0884	0.7764	0.7782	0.7833	0.7558	0.7867

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.0.1
• Transformers: 4.41.2
• PyTorch: 2.2.0+cu121
• Accelerate: 0.31.0
• 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}
}