This is a sentence-transformers model finetuned from vinai/phobert-base-v2. 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: vinai/phobert-base-v2
Maximum Sequence Length: 512 tokens
Output Dimensionality: 768 tokens
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}) with Transformer model: RobertaModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
)

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("huudan123/stag_123")
# Run inference
sentences = [
    'Câu trả lời đơn giản là có, chồi hoa trên rau diếp là một dấu hiệu chắc chắn của việc bắt vít.',
    'Có vẻ như nó đã bắt đầu bắt đầu.',
    'Hai người đàn ông đang đợi một chuyến đi bên lề đường đất.',
]
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

Semantic Similarity

Dataset: sts-evaluator
Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.5793
spearman_cosine	0.5985
pearson_manhattan	0.7081
spearman_manhattan	0.7154
pearson_euclidean	0.4588
spearman_euclidean	0.529
pearson_dot	0.3239
spearman_dot	0.5079
pearson_max	0.7081
spearman_max	0.7154

Training Details

Training Hyperparameters

Non-Default Hyperparameters

overwrite_output_dir: True
eval_strategy: steps
per_device_train_batch_size: 128
per_device_eval_batch_size: 128
gradient_accumulation_steps: 2
learning_rate: 1e-05
num_train_epochs: 15
lr_scheduler_type: cosine_with_restarts
warmup_ratio: 0.1
fp16: True
load_best_model_at_end: True
gradient_checkpointing: True
batch_sampler: no_duplicates

All Hyperparameters

Click to expand

overwrite_output_dir: True
do_predict: False
eval_strategy: steps
prediction_loss_only: True
per_device_train_batch_size: 128
per_device_eval_batch_size: 128
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 2
eval_accumulation_steps: None
learning_rate: 1e-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: 15
max_steps: -1
lr_scheduler_type: cosine_with_restarts
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: False
fp16: True
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: False
hub_always_push: False
gradient_checkpointing: True
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
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional

Training Logs

Click to expand

Epoch	Step	Training Loss	stage1 loss	stage2 loss	stage3 loss	sts-evaluator_spearman_max
0	0	-	-	-	-	0.6643
0.0877	100	4.3054	-	-	-	-
0.1754	200	3.93	-	-	-	-
0.2632	300	3.585	-	-	-	-
0.3509	400	3.4482	-	-	-	-
0.4386	500	3.1858	4.3297	2.6006	0.1494	0.7527
0.5263	600	3.141	-	-	-	-
0.6140	700	2.9477	-	-	-	-
0.7018	800	2.6271	-	-	-	-
0.7895	900	2.6175	-	-	-	-
0.8772	1000	2.4931	2.9001	2.3487	0.1593	0.6907
0.9649	1100	2.4516	-	-	-	-
1.0526	1200	2.4662	-	-	-	-
1.1404	1300	2.5022	-	-	-	-
1.2281	1400	2.4325	-	-	-	-
1.3158	1500	2.4058	2.7163	2.1658	0.1392	0.7121
1.4035	1600	2.3305	-	-	-	-
1.4912	1700	2.2677	-	-	-	-
1.5789	1800	2.2555	-	-	-	-
1.6667	1900	2.2275	-	-	-	-
1.7544	2000	2.1846	2.5441	2.1172	0.1293	0.6781
1.8421	2100	2.2007	-	-	-	-
1.9298	2200	2.192	-	-	-	-
2.0175	2300	2.1491	-	-	-	-
2.1053	2400	2.2419	-	-	-	-
2.1930	2500	2.1822	2.4765	2.0476	0.1055	0.6893
2.2807	2600	2.1384	-	-	-	-
2.3684	2700	2.1379	-	-	-	-
2.4561	2800	2.0558	-	-	-	-
2.5439	2900	2.057	-	-	-	-
2.6316	3000	2.0263	2.4108	2.0751	0.0904	0.7016
2.7193	3100	1.9587	-	-	-	-
2.8070	3200	2.0702	-	-	-	-
2.8947	3300	2.0058	-	-	-	-
2.9825	3400	2.0093	-	-	-	-
3.0702	3500	2.0347	2.3948	1.9958	0.0937	0.7131
3.1579	3600	2.0071	-	-	-	-
3.2456	3700	1.9708	-	-	-	-
3.3333	3800	2.027	-	-	-	-
3.4211	3900	1.9432	-	-	-	-
3.5088	4000	1.9245	2.3858	2.0274	0.0831	0.7197
3.5965	4100	1.8814	-	-	-	-
3.6842	4200	1.8619	-	-	-	-
3.7719	4300	1.8987	-	-	-	-
3.8596	4400	1.8764	-	-	-	-
3.9474	4500	1.8908	2.3753	2.0066	0.0872	0.7052
4.0351	4600	1.8737	-	-	-	-
4.1228	4700	1.9289	-	-	-	-
4.2105	4800	1.8755	-	-	-	-
4.2982	4900	1.8542	-	-	-	-
4.3860	5000	1.8514	2.3731	2.0023	0.0824	0.7191
4.4737	5100	1.7939	-	-	-	-
4.5614	5200	1.8126	-	-	-	-
4.6491	5300	1.7662	-	-	-	-
4.7368	5400	1.7448	-	-	-	-
4.8246	5500	1.7736	2.3703	2.0038	0.0768	0.7044
4.9123	5600	1.7993	-	-	-	-
5.0	5700	1.7811	-	-	-	-
5.0877	5800	1.7905	-	-	-	-
5.1754	5900	1.7539	-	-	-	-
5.2632	6000	1.7393	2.3568	2.0173	0.0853	0.7263
5.3509	6100	1.7882	-	-	-	-
5.4386	6200	1.682	-	-	-	-
5.5263	6300	1.7175	-	-	-	-
5.6140	6400	1.6806	-	-	-	-
5.7018	6500	1.6243	2.3715	2.0202	0.0770	0.7085
5.7895	6600	1.7079	-	-	-	-
5.8772	6700	1.6743	-	-	-	-
5.9649	6800	1.6897	-	-	-	-
6.0526	6900	1.668	-	-	-	-
6.1404	7000	1.6806	2.3826	1.9925	0.0943	0.7072
6.2281	7100	1.6394	-	-	-	-
6.3158	7200	1.6738	-	-	-	-
6.4035	7300	1.6382	-	-	-	-
6.4912	7400	1.6109	-	-	-	-
6.5789	7500	1.5864	2.3849	2.0064	0.0831	0.7200
6.6667	7600	1.5838	-	-	-	-
6.7544	7700	1.5776	-	-	-	-
6.8421	7800	1.5904	-	-	-	-
6.9298	7900	1.6198	-	-	-	-
7.0175	8000	1.5661	2.3917	2.0038	0.0746	0.7131
7.1053	8100	1.6253	-	-	-	-
7.1930	8200	1.5564	-	-	-	-
7.2807	8300	1.5947	-	-	-	-
7.3684	8400	1.5982	-	-	-	-
7.4561	8500	1.53	2.3761	2.0162	0.0775	0.7189
7.5439	8600	1.5412	-	-	-	-
7.6316	8700	1.5287	-	-	-	-
7.7193	8800	1.4652	-	-	-	-
7.8070	8900	1.5611	-	-	-	-
7.8947	9000	1.5258	2.3870	1.9896	0.0828	0.7126
7.9825	9100	1.552	-	-	-	-
8.0702	9200	1.5287	-	-	-	-
8.1579	9300	1.4889	-	-	-	-
8.2456	9400	1.4893	-	-	-	-
8.3333	9500	1.5538	2.3810	1.9956	0.0772	0.7181
8.4211	9600	1.4863	-	-	-	-
8.5088	9700	1.4894	-	-	-	-
8.5965	9800	1.4516	-	-	-	-
8.6842	9900	1.4399	-	-	-	-
8.7719	10000	1.4699	2.3991	1.9760	0.0894	0.7122
8.8596	10100	1.4653	-	-	-	-
8.9474	10200	1.4849	-	-	-	-
9.0351	10300	1.4584	-	-	-	-
9.1228	10400	1.4672	-	-	-	-
9.2105	10500	1.4353	2.3906	2.0104	0.0760	0.7154

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.10.12
Sentence Transformers: 3.0.1
Transformers: 4.42.4
PyTorch: 2.3.1+cu121
Accelerate: 0.33.0
Datasets: 2.20.0
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",
}

TripletLoss

bibtex

@misc{hermans2017defense,
    title={In Defense of the Triplet Loss for Person Re-Identification}, 
    author={Alexander Hermans and Lucas Beyer and Bastian Leibe},
    year={2017},
    eprint={1703.07737},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}

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}
}