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: >-
      Consolidated Regulatory Capital - The capital requirements calculated
      under the FRB’s Capital Framework include the capital conservation buffer
      requirements, which are comprised of a 2.5% buffer (under the Advanced
      Capital Rules).
    sentences:
      - >-
        What was the effective income tax rate for the year ended December 31,
        2023?
      - >-
        What is the function of capital conservation buffer requirements in the
        FRB's Capital Framework for banks like Group Inc. in 2023?
      - >-
        What incentive does the Hawaiian Electric’s Battery Bonus grid services
        program offer?
  - source_sentence: >-
      Balance at beginning of year 2021 was $30 million and, after charge-offs,
      recoveries, and provisions for credit losses, the balance at end of year
      was $18 million.
    sentences:
      - >-
        Between what dates did CS&Co allegedly violate their duty to seek best
        execution as per the plaintiffs' allegations in the lawsuit involving
        UBS Securities LLC?
      - >-
        What were the balance at the beginning and the end of the year for
        credit loss balances in 2021?
      - >-
        How does the company handle leasehold improvements in terms of
        depreciation?
  - source_sentence: >-
      The Compute reporting unit has an excess of fair value over carrying value
      of 5% as of the annual test date.
    sentences:
      - >-
        What percent excess of fair value over carrying value did the Compute
        reporting unit have as of the annual test date in 2023?
      - >-
        What were the effective income tax rates for fiscal years 2023, 2022,
        and 2021, and how did specific tax events affect these rates?
      - >-
        When does the latest expiring European composition of matter patent
        (Supplementary Protection Certificate) for STELARA expire?
  - source_sentence: >-
      The net revenue decrease during 2023 in the Entertainment segment was
      driven by lower entertainment productions and deliveries, reflecting the
      impact of the several months-long strikes during 2023 by the Writers Guild
      of America and the American actors' union, SAG-AFTRA.
    sentences:
      - What was the operating income for Google Cloud in 2023?
      - >-
        How much did the company contribute to its pension and OPEB plans in
        2023?
      - >-
        What was the impact of the strikes by the Writers Guild of America and
        SAG-AFTRA on the Entertainment segment's net revenues in 2023?
  - source_sentence: >-
      As a REIT, future repatriation of incremental undistributed earnings of
      the company's foreign subsidiaries will not be subject to federal or state
      income tax, with the exception of foreign withholding taxes.
    sentences:
      - >-
        What tax implications apply to the future repatriation of incremental
        undistributed earnings by a REIT from its foreign subsidiaries?
      - >-
        What was the accrued liability for product recall related matters as of
        the end of the fiscal year on June 30, 2023?
      - >-
        What was the total amount of future interest payments associated with
        the Notes as of September 30, 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: 0.7128571428571429
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.8428571428571429
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.88
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.92
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.7128571428571429
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.28095238095238095
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.176
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.09199999999999998
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.7128571428571429
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.8428571428571429
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.88
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.92
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.8194470096208256
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.7869285714285713
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.7892168694112985
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 512
          type: dim_512
        metrics:
          - type: cosine_accuracy@1
            value: 0.7214285714285714
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.8471428571428572
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.8757142857142857
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.9185714285714286
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.7214285714285714
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.2823809523809524
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.17514285714285713
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.09185714285714286
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.7214285714285714
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.8471428571428572
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.8757142857142857
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.9185714285714286
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.8222551376922121
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.7912256235827663
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.7935743687249276
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 256
          type: dim_256
        metrics:
          - type: cosine_accuracy@1
            value: 0.7042857142857143
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.8342857142857143
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.8771428571428571
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.9157142857142857
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.7042857142857143
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.27809523809523806
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.1754285714285714
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.09157142857142857
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.7042857142857143
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.8342857142857143
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.8771428571428571
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.9157142857142857
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.813165438848782
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.7800498866213152
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.7822651539071127
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 128
          type: dim_128
        metrics:
          - type: cosine_accuracy@1
            value: 0.6971428571428572
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.8142857142857143
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.8557142857142858
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.9028571428571428
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.6971428571428572
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.2714285714285714
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.17114285714285712
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.09028571428571427
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.6971428571428572
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.8142857142857143
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.8557142857142858
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.9028571428571428
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.7996582219917312
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.7667329931972787
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.7700915959452638
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 64
          type: dim_64
        metrics:
          - type: cosine_accuracy@1
            value: 0.6742857142857143
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.7942857142857143
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.8257142857142857
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.8742857142857143
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.6742857142857143
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.26476190476190475
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.16514285714285712
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.08742857142857141
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.6742857142857143
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.7942857142857143
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.8257142857142857
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.8742857142857143
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.7742733360934079
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.7424053287981859
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.7463231326238146
            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("mezeidragos-lateral/bge-base-financial-matryoshka")
# Run inference
sentences = [
    "As a REIT, future repatriation of incremental undistributed earnings of the company's foreign subsidiaries will not be subject to federal or state income tax, with the exception of foreign withholding taxes.",
    'What tax implications apply to the future repatriation of incremental undistributed earnings by a REIT from its foreign subsidiaries?',
    'What was the accrued liability for product recall related matters as of the end of the fiscal year on June 30, 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

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	0.7129	0.7214	0.7043	0.6971	0.6743
cosine_accuracy@3	0.8429	0.8471	0.8343	0.8143	0.7943
cosine_accuracy@5	0.88	0.8757	0.8771	0.8557	0.8257
cosine_accuracy@10	0.92	0.9186	0.9157	0.9029	0.8743
cosine_precision@1	0.7129	0.7214	0.7043	0.6971	0.6743
cosine_precision@3	0.281	0.2824	0.2781	0.2714	0.2648
cosine_precision@5	0.176	0.1751	0.1754	0.1711	0.1651
cosine_precision@10	0.092	0.0919	0.0916	0.0903	0.0874
cosine_recall@1	0.7129	0.7214	0.7043	0.6971	0.6743
cosine_recall@3	0.8429	0.8471	0.8343	0.8143	0.7943
cosine_recall@5	0.88	0.8757	0.8771	0.8557	0.8257
cosine_recall@10	0.92	0.9186	0.9157	0.9029	0.8743
cosine_ndcg@10	0.8194	0.8223	0.8132	0.7997	0.7743
cosine_mrr@10	0.7869	0.7912	0.78	0.7667	0.7424
cosine_map@100	0.7892	0.7936	0.7823	0.7701	0.7463

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: 44.62 tokens
max: 301 tokens

min: 10 tokens
mean: 20.66 tokens
max: 45 tokens

	positive	anchor
type	string	string
details	min: 4 tokens mean: 44.62 tokens max: 301 tokens	min: 10 tokens mean: 20.66 tokens max: 45 tokens

Samples:

positive	anchor
`We provide transaction processing services (primarily authorization, clearing and settlement) to our financial institution and merchant clients through VisaNet, our proprietary advanced transaction processing network.`	`What are the primary transaction processing services provided by Visa through VisaNet?`
`Information about legal proceedings is included in Item 8 of the Annual Report on Form 10-K, as referenced in Item 3.`	`What item in the Annual Report on Form 10-K provides information about legal proceedings?`
`Investing activities used cash of $3.0 billion in 2022.`	`What was the net cash used by investing activities in 2022?`

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
load_best_model_at_end: True
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: None
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
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	1.5626	-	-	-	-	-
1.0	13	-	0.8071	0.8040	0.7933	0.7781	0.7478
1.5685	20	0.6111	-	-	-	-	-
2.0	26	-	0.8173	0.8192	0.8111	0.7961	0.7661
2.3249	30	0.4333	-	-	-	-	-
3.0	39	-	0.8193	0.8211	0.8127	0.7996	0.7729
3.0812	40	0.3465	-	-	-	-	-
3.731	48	-	0.8194	0.8223	0.8132	0.7997	0.7743

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.12.8
Sentence Transformers: 3.3.1
Transformers: 4.48.0
PyTorch: 2.2.2
Accelerate: 1.2.1
Datasets: 3.2.0
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}
}