amichelini/bge-base-financial-matryoshka

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 tokens
• 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("amichelini/bge-base-financial-matryoshka")
# Run inference
sentences = [
    '2023 highlights include net revenues of $5,003.3 million which decreased 15% from $5,856.7 million in 2022.',
    "How did Hasbro's net revenues in 2023 compare to the previous year?",
    'How much cash did continuing operating activities provide in 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	0.68
cosine_accuracy@3	0.81
cosine_accuracy@5	0.8514
cosine_accuracy@10	0.8943
cosine_precision@1	0.68
cosine_precision@3	0.27
cosine_precision@5	0.1703
cosine_precision@10	0.0894
cosine_recall@1	0.68
cosine_recall@3	0.81
cosine_recall@5	0.8514
cosine_recall@10	0.8943
cosine_ndcg@10	0.7882
cosine_mrr@10	0.7541
cosine_map@100	0.7585

Information Retrieval

• Dataset: dim_512
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.68
cosine_accuracy@3	0.8029
cosine_accuracy@5	0.8457
cosine_accuracy@10	0.8971
cosine_precision@1	0.68
cosine_precision@3	0.2676
cosine_precision@5	0.1691
cosine_precision@10	0.0897
cosine_recall@1	0.68
cosine_recall@3	0.8029
cosine_recall@5	0.8457
cosine_recall@10	0.8971
cosine_ndcg@10	0.7871
cosine_mrr@10	0.752
cosine_map@100	0.7559

Information Retrieval

• Dataset: dim_256
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6714
cosine_accuracy@3	0.7986
cosine_accuracy@5	0.8457
cosine_accuracy@10	0.8843
cosine_precision@1	0.6714
cosine_precision@3	0.2662
cosine_precision@5	0.1691
cosine_precision@10	0.0884
cosine_recall@1	0.6714
cosine_recall@3	0.7986
cosine_recall@5	0.8457
cosine_recall@10	0.8843
cosine_ndcg@10	0.7799
cosine_mrr@10	0.7462
cosine_map@100	0.7506

Information Retrieval

• Dataset: dim_128
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.66
cosine_accuracy@3	0.7914
cosine_accuracy@5	0.8286
cosine_accuracy@10	0.8814
cosine_precision@1	0.66
cosine_precision@3	0.2638
cosine_precision@5	0.1657
cosine_precision@10	0.0881
cosine_recall@1	0.66
cosine_recall@3	0.7914
cosine_recall@5	0.8286
cosine_recall@10	0.8814
cosine_ndcg@10	0.7707
cosine_mrr@10	0.7354
cosine_map@100	0.7396

Information Retrieval

• Dataset: dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.6271
cosine_accuracy@3	0.7543
cosine_accuracy@5	0.8014
cosine_accuracy@10	0.86
cosine_precision@1	0.6271
cosine_precision@3	0.2514
cosine_precision@5	0.1603
cosine_precision@10	0.086
cosine_recall@1	0.6271
cosine_recall@3	0.7543
cosine_recall@5	0.8014
cosine_recall@10	0.86
cosine_ndcg@10	0.7404
cosine_mrr@10	0.7026
cosine_map@100	0.7069

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: 46.33 tokens max: 326 tokens	min: 7 tokens mean: 20.38 tokens max: 43 tokens

• Samples:

  positive	anchor
  The data includes transaction and integration costs listed as follows for each year: $0, $0, $59, $0, $0, $0, $269, $91, $39, $269, $91, $98.	What were the values of transaction and integration costs for each of the years provided in the data?
  In 2023, Delta Air Lines announced an increase in remuneration from their partnership with American Express to $6.8 billion, with expected growth of 10% in 2024.	What was the remuneration from Delta Air Lines' partnership with American Express in 2023, and what is the growth expectation for 2024?
  On December 1, 2023, we advanced $10.0 billion under the ASR program and received approximately 215 million shares of common stock with a value of $6.8 billion, which were immediately retired.	What significant financial activity occurred on December 1, 2023, under the ASR program?

• 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: 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
• eval_on_start: False
• eval_use_gather_object: 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	0	-	0.6648	0.6922	0.6982	0.6028	0.7029
0.8122	10	1.5362	-	-	-	-	-
0.9746	12	-	0.7259	0.7402	0.7481	0.6913	0.7510
1.6244	20	0.6012	-	-	-	-	-
1.9492	24	-	0.7341	0.7503	0.7554	0.7051	0.7576
2.4365	30	0.4225	-	-	-	-	-
2.9239	36	-	0.7383	0.7522	0.7569	0.7063	0.7570
3.2487	40	0.358	-	-	-	-	-
3.8985	48	-	0.7396	0.7506	0.7559	0.7069	0.7585

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.1.1
• Transformers: 4.44.2
• PyTorch: 2.4.1+cu121
• Accelerate: 0.34.2
• Datasets: 3.0.0
• 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}
}