DungHugging/bge-m3-banking-v2

SentenceTransformer based on BAAI/bge-m3

This is a sentence-transformers model finetuned from BAAI/bge-m3. 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: 512 tokens
• Output Dimensionality: 1024 dimensions
• Similarity Function: Cosine Similarity

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, 'architecture': 'PeftModelForFeatureExtraction'})
  (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.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("DungHugging/bge-m3-banking-v2")
# Run inference
sentences = [
    'không tích điểm thưởng cho giao dịch rút tiền mặt',
    'giao dịch rút tiền tại ATM bị loại trừ khỏi chương trình ưu đãi',
    'Salmonoid chuyên phục vụ các món hải sản, không có món thịt bò trong menu.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.8485, 0.6857],
#         [0.8485, 1.0000, 0.6427],
#         [0.6857, 0.6427, 1.0000]])

Evaluation

Metrics

Semantic Similarity

• Datasets: spec_sim and final_evaluation
• Evaluated with EmbeddingSimilarityEvaluator

Metric	spec_sim	final_evaluation
pearson_cosine	0.4669	0.4669
spearman_cosine	0.4602	0.4602

Binary Classification

• Dataset: spec_bin
• Evaluated with BinaryClassificationEvaluator

Metric	Value
cosine_accuracy	0.7297
cosine_accuracy_threshold	0.8031
cosine_f1	0.7561
cosine_f1_threshold	0.8026
cosine_precision	0.7209
cosine_recall	0.7949
cosine_ap	0.7603
cosine_mcc	0.4574

Training Details

Training Dataset

Unnamed Dataset

• Size: 2,665 training samples
• Columns: sentence_0, sentence_1, and label
• Approximate statistics based on the first 1000 samples:

  	sentence_0	sentence_1	label
  type	string	string	float
  details	min: 6 tokens mean: 14.98 tokens max: 28 tokens	min: 8 tokens mean: 18.38 tokens max: 55 tokens	min: 0.0 mean: 0.5 max: 1.0

• Samples:

  sentence_0	sentence_1	label
  HSSV từ 20 tuổi dùng Combo Hi-Tek được miễn phí nếu số dư bình quân đạt 500.000 VND.	Điều kiện miễn phí cho sinh viên trên 20 tuổi là duy trì số dư bình quân từ 500.000 VND.	1.0
  vay tín chấp dựa trên lịch sử tín dụng (CIC)	vay cầm cố dựa trên số dư sổ tiết kiệm	0.0
  tất toán sổ tiết kiệm từng phần linh hoạt	rút gốc từng phần sẽ làm mất toàn bộ lãi suất của sổ	0.0

• Loss: OnlineContrastiveLoss

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• num_train_epochs: 10
• multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 1
• eval_accumulation_steps: None
• torch_empty_cache_steps: None
• learning_rate: 5e-05
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1
• num_train_epochs: 10
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.0
• 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
• bf16: False
• 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: False
• 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}
• parallelism_config: 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
• project: huggingface
• trackio_space_id: trackio
• 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
• hub_revision: None
• 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
• include_tokens_per_second: False
• include_num_input_tokens_seen: no
• neftune_noise_alpha: None
• optim_target_modules: None
• batch_eval_metrics: False
• eval_on_start: False
• use_liger_kernel: False
• liger_kernel_config: None
• eval_use_gather_object: False
• average_tokens_across_devices: True
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: round_robin
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	spec_sim_spearman_cosine	spec_bin_cosine_ap	final_evaluation_spearman_cosine
0.4970	83	-	-0.1803	0.4536	-
0.9940	166	-	-0.0670	0.4908	-
1.0	167	-	-0.0659	0.4914	-
1.4910	249	-	0.0642	0.5423	-
1.9880	332	-	0.1309	0.5742	-
2.0	334	-	0.1322	0.5750	-
2.4850	415	-	0.2133	0.6127	-
2.9820	498	-	0.2749	0.6459	-
2.9940	500	0.9824	-	-	-
3.0	501	-	0.2777	0.6468	-
3.4790	581	-	0.3197	0.6726	-
3.9760	664	-	0.3543	0.6954	-
4.0	668	-	0.3554	0.6960	-
4.4731	747	-	0.3734	0.7081	-
4.9701	830	-	0.3955	0.7203	-
5.0	835	-	0.3963	0.7207	-
5.4671	913	-	0.4037	0.7250	-
5.9641	996	-	0.4241	0.7372	-
5.9880	1000	0.6612	-	-	-
6.0	1002	-	0.4259	0.7381	-
6.4611	1079	-	0.4328	0.7414	-
6.9581	1162	-	0.4393	0.7452	-
7.0	1169	-	0.4398	0.7454	-
7.4551	1245	-	0.4466	0.7530	-
7.9521	1328	-	0.4520	0.7562	-
8.0	1336	-	0.4527	0.7566	-
8.4491	1411	-	0.4573	0.7592	-
8.9461	1494	-	0.4590	0.7599	-
8.9820	1500	0.5747	-	-	-
9.0	1503	-	0.4586	0.7596	-
9.4431	1577	-	0.4597	0.7600	-
9.9401	1660	-	0.4602	0.7603	-
10.0	1670	-	0.4602	0.7603	-
-1	-1	-	-	-	0.4602

Framework Versions

• Python: 3.12.12
• Sentence Transformers: 5.1.1
• Transformers: 4.57.1
• PyTorch: 2.8.0+cu126
• Accelerate: 1.11.0
• Datasets: 4.4.2
• Tokenizers: 0.22.1

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",
}