Regulatory 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("hshashank06/regulatory-model")
# Run inference
sentences = [
    ' Home Depot\'s stock closed at $135.39 while being above a "golden cross" on January 19, 2017.',
    ' In the given text passage, when did Home Depot\'s stock close at $135.39 while being above a "golden cross"? \n',
    ' According to Maley, where might the funds from potentially declining sectors like FANGs be directed towards? \n',
]
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	0.6027	0.5989	0.5881	0.5719	0.5443
cosine_accuracy@3	0.7349	0.7302	0.7223	0.7058	0.6759
cosine_accuracy@5	0.7676	0.7651	0.7568	0.7406	0.7139
cosine_accuracy@10	0.8058	0.8034	0.7947	0.7819	0.7583
cosine_precision@1	0.6027	0.5989	0.5881	0.5719	0.5443
cosine_precision@3	0.245	0.2434	0.2408	0.2353	0.2253
cosine_precision@5	0.1535	0.153	0.1514	0.1481	0.1428
cosine_precision@10	0.0806	0.0803	0.0795	0.0782	0.0758
cosine_recall@1	0.6027	0.5989	0.5881	0.5719	0.5443
cosine_recall@3	0.7349	0.7302	0.7223	0.7058	0.6759
cosine_recall@5	0.7676	0.7651	0.7568	0.7406	0.7139
cosine_recall@10	0.8058	0.8034	0.7947	0.7819	0.7583
cosine_ndcg@10	0.7073	0.704	0.6945	0.6793	0.6523
cosine_mrr@10	0.6755	0.6719	0.6622	0.6463	0.6182
cosine_map@100	0.6797	0.6761	0.6666	0.6508	0.6229

Training Details

Training Dataset

json

Dataset: json
Size: 185,814 training samples
Columns: positive and anchor
Approximate statistics based on the first 1000 samples:
positive anchor
type string string
details
min: 3 tokens
mean: 43.18 tokens
max: 200 tokens

min: 10 tokens
mean: 23.08 tokens
max: 63 tokens

	positive	anchor
type	string	string
details	min: 3 tokens mean: 43.18 tokens max: 200 tokens	min: 10 tokens mean: 23.08 tokens max: 63 tokens

Samples:

positive	anchor
`The BVPS (Book Value Per Share) is calculated by dividing a company's common equity value by its total number of shares outstanding. In the given example, if a company has a common equity value of $100 million and 10 million shares outstanding, its BVPS would be $10 ($100 million / 10 million). You can calculate a company's BVPS using Microsoft Excel by entering the values of common stock, retained earnings, and additional paid-in capital into cells A1 through A3.`	`What is the BVPS and how is it calculated?`
`They facilitate commodities trading using their resources, can take delivery of commodities if needed, provide advisory services for clients, and act as market makers by buying and selling futures contracts to add liquidity to the marketplace. The passage uses the example of a commercial baking firm to demonstrate how their impact can be seen in the market.`	`What role do eligible commercial entities play in commodities trading and market liquidity?`
`Naive diversification is a type of diversification strategy where an investor randomly selects different securities, hoping to lower the risk of the portfolio due to the varied nature of the chosen securities. It is less sophisticated than diversification methods using statistical modeling, but when guided by experience, careful security examination, and common sense, it remains an effective strategy for reducing portfolio risk.`	`What is the concept of naive diversification in investing and how does it compare to more sophisticated diversification methods?`

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
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: 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_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