You can deploy this model and start earning money today!

Views

No views yet

BGE base Financial Matryoshka

Name: bge-base-financial-matryoshka2
Availability: InStock

This is a sentence-transformers model finetuned from BAAI/bge-m3 on the json dataset. It maps sentences & paragraphs to a 1024-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-m3
Maximum Sequence Length: 8192 tokens
Output Dimensionality: 1024 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': 8192, 'do_lower_case': False}) with Transformer model: XLMRobertaModel 
  (1): Pooling({'word_embedding_dimension': 1024, '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.

python

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("Ram934/bge-base-financial-matryoshka2")
# Run inference
sentences = [
    'When can I set TE at 50% of PM',
    'The accounting professional is committed to producing high-quality work during the audit process. We will conduct all necessary procedures to ensure accuracy and provide detailed explanations of our findings. Our team is dedicated to assisting you and ensuring that you are fully informed throughout the entire audit process. Thank you for choosing us to perform the audit.Follow-up Questions:1) Can the accounting professional anticipate any challenges during the audit process?2) What are some examples of thorough audit procedures?3) How can the accounting professional provide detailed explanations of findings?',
    'The need for extending other substative procedures is not necessary if auditors follow Ernst & Young (EY) policies American Institute of Certified Public Accountants (AICPA) AU-C 330.20 and Public Company Accounting Oversight Board (PCAOB) Advisory Services (AS) 2310.35, which address the criteria for requesting confirmations and document the rationale for not performing confirmations if unable to provide them. These policies ensure thoroughness and reliability in the audit process, minimizing the potential for errors or misstatements. To ensure compliance with these policies, auditors should carefully consider their approach to confirmation requests and have a clear understanding of the criteria outlined by the policies. By adhering to these guidelines, auditors can ensure a high level of accuracy and reliability in the audit process.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

# 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	1.0	1.0	1.0	0.3333	0.3333
cosine_accuracy@3	1.0	1.0	1.0	0.6667	0.6667
cosine_accuracy@5	1.0	1.0	1.0	1.0	0.6667
cosine_accuracy@10	1.0	1.0	1.0	1.0	1.0
cosine_precision@1	1.0	1.0	1.0	0.3333	0.3333
cosine_precision@3	0.3333	0.3333	0.3333	0.2222	0.2222
cosine_precision@5	0.2	0.2	0.2	0.2	0.1333
cosine_precision@10	0.1	0.1	0.1	0.1	0.1
cosine_recall@1	1.0	1.0	1.0	0.3333	0.3333
cosine_recall@3	1.0	1.0	1.0	0.6667	0.6667
cosine_recall@5	1.0	1.0	1.0	1.0	0.6667
cosine_recall@10	1.0	1.0	1.0	1.0	1.0
cosine_ndcg@10	1.0	1.0	1.0	0.6726	0.6111
cosine_mrr@10	1.0	1.0	1.0	0.5667	0.4921
cosine_map@100	1.0	1.0	1.0	0.5667	0.4921

Training Details

Training Dataset

json

Dataset: json
Size: 27 training samples
Columns: anchor and positive
Approximate statistics based on the first 27 samples:
anchor positive
type string string
details
min: 9 tokens
mean: 18.41 tokens
max: 40 tokens
min: 21 tokens
mean: 171.26 tokens
max: 308 tokens

Samples:

anchor	positive
`what is purpose of conducting walk-throughs and test of controls in ICFR audit`	The purpose of conducting walk-throughs and test of controls in ICFR audit is to ensure the effectiveness of the internal control system, identify control weaknesses, verify control activities, and assess the entity's ability to satisfy its financial reporting requirements. Walk-throughs provide auditors with a comprehensive understanding of critical control processes and underlying controls to inform their assessment of the entity's overall control environment. In the event of any potential issues discovered during the walk-through process, further examination and testing of controls should be performed. By conducting these processes, auditors can help maintain the integrity of the financial reporting process and safeguard against fraudulent activities.
`Sampling risk to be considered as some in which situations`	Sampling risk refers to the possibility of making an incorrect decision due to using a sample instead of examining the entire population. It is considered in the following situations:1. When the sample selected is not representative of the population.2. When the sampling process is not done correctly, leading to misclassification or non-representativeness of the sample.3. When the sample size is too small to produce reliable results or when the level of confidence in the sample results is too low.It is important to consider sampling risk when conducting audits because it allows auditors to understand the limitations of their work and to adjust their findings accordingly. By identifying potential risks and taking steps to minimize them, auditors can ensure that their results are accurate and reliable.
`Do I need to do test of control top up testing?`	`Re-performing applicable transactions involves conducting testing for each processed transaction at a lower level of risk, while manually reperforming controls refers to the manual inspection and examination of specific elements within the control system. Three examples of controls that are non-configurable are:1. Security protocols and policies2. Access control mechanisms3. Data integrity checksENDINSTRUCTION`

Loss: MatryoshkaLoss with these parameters:
json
1{ 2 "loss": "MultipleNegativesRankingLoss", 3 "matryoshka_dims": [ 4 768, 5 512, 6 256, 7 128, 8 64 9 ], 10 "matryoshka_weights": [ 11 1, 12 1, 13 1, 14 1, 15 1 16 ], 17 "n_dims_per_step": -1 18}

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: False
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: False
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
prompts: None
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	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
1.0	1	1.0	1.0	0.8333	0.8102	0.6624
2.0	2	1.0	1.0	0.8333	0.7956	0.6548
3.0	3	1.0	1.0	1.0	0.6872	0.6111
4.0	4	1.0	1.0	1.0	0.6726	0.6111

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.10.14
Sentence Transformers: 3.3.1
Transformers: 4.44.2
PyTorch: 2.4.1+cu121
Accelerate: 1.1.1
Datasets: 2.19.0
Tokenizers: 0.19.1

Citation

BibTeX

Sentence Transformers

bibtex

@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

bibtex

@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

bibtex

@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}
}

	anchor	positive
type	string	string
details	min: 9 tokens mean: 18.41 tokens max: 40 tokens	min: 21 tokens mean: 171.26 tokens max: 308 tokens