juanpablomesa/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("juanpablomesa/bge-base-financial-matryoshka")
# Run inference
sentences = [
    'HTC called the Samsung Galaxy S4 “mainstream”.',
    'What did HTC announce about the Samsung Galaxy S4?',
    "What is the essential aspect of the vocation to marriage according to Benedict XVI's message on the 40th Anniversary of Humanae Vitae?",
]
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	0.9675
cosine_accuracy@3	0.9792
cosine_accuracy@5	0.9829
cosine_accuracy@10	0.9888
cosine_precision@1	0.9675
cosine_precision@3	0.3264
cosine_precision@5	0.1966
cosine_precision@10	0.0989
cosine_recall@1	0.9675
cosine_recall@3	0.9792
cosine_recall@5	0.9829
cosine_recall@10	0.9888
cosine_ndcg@10	0.9777
cosine_mrr@10	0.9742
cosine_map@100	0.9745

Information Retrieval

• Dataset: dim_512
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.9642
cosine_accuracy@3	0.9775
cosine_accuracy@5	0.9817
cosine_accuracy@10	0.9888
cosine_precision@1	0.9642
cosine_precision@3	0.3258
cosine_precision@5	0.1963
cosine_precision@10	0.0989
cosine_recall@1	0.9642
cosine_recall@3	0.9775
cosine_recall@5	0.9817
cosine_recall@10	0.9888
cosine_ndcg@10	0.9759
cosine_mrr@10	0.9718
cosine_map@100	0.972

Information Retrieval

• Dataset: dim_256
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.9621
cosine_accuracy@3	0.9742
cosine_accuracy@5	0.9804
cosine_accuracy@10	0.9862
cosine_precision@1	0.9621
cosine_precision@3	0.3247
cosine_precision@5	0.1961
cosine_precision@10	0.0986
cosine_recall@1	0.9621
cosine_recall@3	0.9742
cosine_recall@5	0.9804
cosine_recall@10	0.9862
cosine_ndcg@10	0.9738
cosine_mrr@10	0.9698
cosine_map@100	0.9702

Information Retrieval

• Dataset: dim_128
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.9554
cosine_accuracy@3	0.97
cosine_accuracy@5	0.9767
cosine_accuracy@10	0.9838
cosine_precision@1	0.9554
cosine_precision@3	0.3233
cosine_precision@5	0.1953
cosine_precision@10	0.0984
cosine_recall@1	0.9554
cosine_recall@3	0.97
cosine_recall@5	0.9767
cosine_recall@10	0.9838
cosine_ndcg@10	0.9693
cosine_mrr@10	0.9647
cosine_map@100	0.9652

Information Retrieval

• Dataset: dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.9392
cosine_accuracy@3	0.9617
cosine_accuracy@5	0.9667
cosine_accuracy@10	0.9758
cosine_precision@1	0.9392
cosine_precision@3	0.3206
cosine_precision@5	0.1933
cosine_precision@10	0.0976
cosine_recall@1	0.9392
cosine_recall@3	0.9617
cosine_recall@5	0.9667
cosine_recall@10	0.9758
cosine_ndcg@10	0.9577
cosine_mrr@10	0.9519
cosine_map@100	0.9525

Training Details

Training Dataset

Unnamed Dataset

• Size: 9,600 training samples
• Columns: positive and anchor
• Approximate statistics based on the first 1000 samples:

  	positive	anchor
  type	string	string
  details	min: 3 tokens mean: 50.19 tokens max: 435 tokens	min: 3 tokens mean: 18.66 tokens max: 43 tokens

• Samples:

  positive	anchor
  The Berry Export Summary 2028 is a dedicated export plan for the Australian strawberry, raspberry, and blackberry industries. It maps the sectors’ current position, where they want to be, high-opportunity markets, and next steps. The purpose of this plan is to grow their global presence over the next 10 years.	What is the Berry Export Summary 2028 and what is its purpose?
  Benefits reported from having access to Self-supply water sources include convenience, less time spent for fetching water and access to more and better quality water. In some areas, Self-supply sources offer important added values such as water for productive use, income generation, family safety and improved food security.	What are some of the benefits reported from having access to Self-supply water sources?
  The unique features of the Coolands for Twitter app include Real-Time updates without the need for a refresh button, Avatar Indicator which shows small avatars on the title bar for new messages, Direct Link for intuitive and convenient link opening, Smart Bookmark to easily return to previous reading position, and User Level Notification which allows customized notification settings for different users.	What are the unique features of the Coolands for Twitter app?

• 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
0.5333	10	0.6065	-	-	-	-	-
0.96	18	-	0.9583	0.9674	0.9695	0.9372	0.9708
1.0667	20	0.3313	-	-	-	-	-
1.6	30	0.144	-	-	-	-	-
1.9733	37	-	0.9630	0.9699	0.9716	0.9488	0.9745
2.1333	40	0.1317	-	-	-	-	-
2.6667	50	0.0749	-	-	-	-	-
2.9867	56	-	0.9650	0.9701	0.9721	0.9522	0.9747
3.2	60	0.088	-	-	-	-	-
3.7333	70	0.0598	-	-	-	-	-
3.84	72	-	0.9652	0.9702	0.972	0.9525	0.9745

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.11.5
• Sentence Transformers: 3.0.1
• Transformers: 4.41.2
• PyTorch: 2.1.2+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}
}