Abinaya/snowflake-arctic-embed-financial-matryoshka

BGE base Financial Matryoshka

This is a sentence-transformers model finetuned from Snowflake/snowflake-arctic-embed-m-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: Snowflake/snowflake-arctic-embed-m-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': False}) 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("Abinaya/snowflake-arctic-embed-financial-matryoshka")
# Run inference
sentences = [
    'As of December 31, 2023, Bank of America reported gross derivative assets and liabilities totaling $290.3 billion and $301.2 billion, respectively. After accounting for legally enforceable master netting agreements and cash collateral, these figures were adjusted to $39.3 billion in assets and $43.4 billion in liabilities.',
    'What were the total derivative assets and liabilities at Bank of America as of December 31, 2023, after adjusting for master netting agreements and cash collateral?',
    'By what percentage did HIV product sales increase in 2023 compared to the previous year?',
]
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	0.7543	0.7543	0.7557	0.7443	0.7
cosine_accuracy@3	0.8614	0.8614	0.86	0.85	0.8157
cosine_accuracy@5	0.8914	0.8914	0.8943	0.8871	0.8571
cosine_accuracy@10	0.9329	0.93	0.9286	0.9143	0.9071
cosine_precision@1	0.7543	0.7543	0.7557	0.7443	0.7
cosine_precision@3	0.2871	0.2871	0.2867	0.2833	0.2719
cosine_precision@5	0.1783	0.1783	0.1789	0.1774	0.1714
cosine_precision@10	0.0933	0.093	0.0929	0.0914	0.0907
cosine_recall@1	0.7543	0.7543	0.7557	0.7443	0.7
cosine_recall@3	0.8614	0.8614	0.86	0.85	0.8157
cosine_recall@5	0.8914	0.8914	0.8943	0.8871	0.8571
cosine_recall@10	0.9329	0.93	0.9286	0.9143	0.9071
cosine_ndcg@10	0.8431	0.8409	0.8409	0.8298	0.8023
cosine_mrr@10	0.8144	0.8124	0.8129	0.8026	0.7689
cosine_map@100	0.8171	0.8153	0.8158	0.8061	0.7722

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: 6 tokens mean: 46.81 tokens max: 326 tokens	min: 9 tokens mean: 20.23 tokens max: 45 tokens

• Samples:

  positive	anchor
  Opioids Related Securities Class Actions and Derivative Litigation: Three derivative complaints and two securities class actions drawing heavily on the allegations of the DOJ complaint have been filed in Delaware naming the Company and various current and former directors and certain current and former officers as defendants. The plaintiffs in the derivative suits (in which the Company is a nominal defendant) allege, among other things, that the defendants breached their fidariety duties in connection with oversight of opioids dispensing and distribution and that the defendants violated Section 14(a) of the Securities Exchange Act of 1934, as amended (the 'Exchange Act()), and are liable for contribution under Section 10(b) of the Exchange Act in connection with the Company's disclosures about opioids.	What kind of claims are involved in the securities and derivative litigation against the Company listed in the document?
  Walmart's fintech venture, ONE, provides financial services such as money orders, prepaid access, money transfers, check cashing, bill payment, and certain types of installment lending.	What types of financial services are offered through Walmart's fintech venture, ONE?
  Juice and juice concentrate from various fruits, particularly orange juice and orange juice concentrate, are principal raw materials for juice and juice drink products, and milk is the principal raw material for dairy products managed through fairlife, LLC.	What are the primary raw materials for the company's juice and dairy products?

• 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	1.5521	-	-	-	-	-
1.0	13	-	0.8136	0.8108	0.8143	0.7949	0.7552
1.5685	20	0.4812	-	-	-	-	-
2.0	26	-	0.8405	0.8388	0.8384	0.8284	0.7990
2.3249	30	0.3585	-	-	-	-	-
3.0	39	-	0.8420	0.8409	0.8408	0.8290	0.8009
3.0812	40	0.3101	-	-	-	-	-
3.731	48	-	0.8431	0.8409	0.8409	0.8298	0.8023

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.15
• Sentence Transformers: 3.4.0
• Transformers: 4.47.1
• PyTorch: 2.5.1+cu124
• Accelerate: 1.0.1
• Datasets: 3.0.1
• 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}
}