This is a sentence-transformers model finetuned from huudan123/stage1. 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: huudan123/stage1
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/stage2")
# Run inference
sentences = [
    'bạn tiếp_tục nhập thông_tin cơ_sở dữ_liệu',
    'bạn mọi thứ bạn bắt_đầu_từ',
    'bạn tiếp_tục bạn nhập mọi thứ',
]
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-dev
Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.7133
spearman_cosine	0.714
pearson_manhattan	0.6924
spearman_manhattan	0.6987
pearson_euclidean	0.6928
spearman_euclidean	0.6988
pearson_dot	0.6562
spearman_dot	0.6553
pearson_max	0.7133
spearman_max	0.714

Training Details

Training Dataset

Unnamed Dataset

Size: 254,546 training samples
Columns: anchor, positive, and negative
Approximate statistics based on the first 1000 samples:
anchor positive negative
type string string string
details
min: 3 tokens
mean: 14.78 tokens
max: 110 tokens
min: 3 tokens
mean: 14.78 tokens
max: 110 tokens
min: 3 tokens
mean: 10.19 tokens
max: 29 tokens

Samples:

anchor	positive	negative
`conceptualy kem skiming hai kích_thước cơ_bản sản_phẩm địa_lý`	`sản_phẩm địa_lý làm kem skiming làm_việc`	`kem skiming hai tập_trung sản_phẩm địa_lý`
`sản_phẩm địa_lý làm kem skiming làm_việc`	`conceptualy kem skiming hai kích_thước cơ_bản sản_phẩm địa_lý`	`kem skiming hai tập_trung sản_phẩm địa_lý`
`bạn biết trong mùa giải tôi đoán ở mức_độ bạn bạn mất chúng đến mức_độ tiếp_theo họ quyết_định nhớ đội_ngũ cha_mẹ chiến_binh quyết_định gọi nhớ một người ba a một người đàn_ông đi đến thay_thế anh ta một người đàn_ông nào đi thay_thế anh ta`	`recals thực_hiện thứ sáu`	`anh mất mọi thứ ở mức_độ người dân nhớ`

Loss: MultipleNegativesRankingLoss with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "cos_sim" 4}

Evaluation Dataset

Unnamed Dataset

Size: 1,660 evaluation samples
Columns: anchor, positive, and negative
Approximate statistics based on the first 1000 samples:
anchor positive negative
type string string string
details
min: 4 tokens
mean: 13.54 tokens
max: 51 tokens
min: 4 tokens
mean: 13.54 tokens
max: 51 tokens
min: 3 tokens
mean: 8.78 tokens
max: 22 tokens

Samples:

anchor	positive	negative
`anh ấy nói mẹ con về nhà`	`xuống xe_buýt trường anh ấy gọi mẹ`	`anh nói mẹ anh về nhà`
`xuống xe_buýt trường anh ấy gọi mẹ`	`anh ấy nói mẹ con về nhà`	`anh nói mẹ anh về nhà`
`tôi biết mình hướng tới mục_đích báo_cáo một địa_chỉ ở washington`	`tôi bao_giờ đến washington tôi chỉ_định ở tôi lạc cố_gắng tìm`	`tôi hoàn_toàn chắc_chắn tôi làm tôi đi đến washington tôi giao báo_cáo`

Loss: MultipleNegativesRankingLoss with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "cos_sim" 4}

Training Hyperparameters

Non-Default Hyperparameters

overwrite_output_dir: True
eval_strategy: epoch
per_device_train_batch_size: 256
per_device_eval_batch_size: 256
num_train_epochs: 20
lr_scheduler_type: cosine
warmup_ratio: 0.05
fp16: True
load_best_model_at_end: True
gradient_checkpointing: True

All Hyperparameters

Click to expand

overwrite_output_dir: True
do_predict: False
eval_strategy: epoch
prediction_loss_only: True
per_device_train_batch_size: 256
per_device_eval_batch_size: 256
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 1
eval_accumulation_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: 1.0
num_train_epochs: 20
max_steps: -1
lr_scheduler_type: cosine
lr_scheduler_kwargs: {}
warmup_ratio: 0.05
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: batch_sampler
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	loss	sts-dev_spearman_cosine
0	0	-	-	0.5307
0.0503	50	9.1742	-	-
0.1005	100	5.9716	-	-
0.1508	150	4.6737	-	-
0.2010	200	3.2819	-	-
0.2513	250	2.8832	-	-
0.3015	300	2.7327	-	-
0.3518	350	2.6305	-	-
0.4020	400	2.6239	-	-
0.4523	450	2.5527	-	-
0.5025	500	2.5271	-	-
0.5528	550	2.4904	-	-
0.6030	600	2.4987	-	-
0.6533	650	2.4009	-	-
0.7035	700	2.3944	-	-
0.7538	750	2.5054	-	-
0.8040	800	2.3989	-	-
0.8543	850	2.4019	-	-
0.9045	900	2.3638	-	-
0.9548	950	2.3478	-	-
1.0	995	-	3.0169	0.7322
1.0050	1000	2.4424	-	-
1.0553	1050	2.2478	-	-
1.1055	1100	2.2448	-	-
1.1558	1150	2.205	-	-
1.2060	1200	2.1811	-	-
1.2563	1250	2.1794	-	-
1.3065	1300	2.1495	-	-
1.3568	1350	2.1548	-	-
1.4070	1400	2.1299	-	-
1.4573	1450	2.1335	-	-
1.5075	1500	2.1388	-	-
1.5578	1550	2.0999	-	-
1.6080	1600	2.0859	-	-
1.6583	1650	2.0959	-	-
1.7085	1700	2.0334	-	-
1.7588	1750	2.0647	-	-
1.8090	1800	2.0261	-	-
1.8593	1850	2.0133	-	-
1.9095	1900	2.0517	-	-
1.9598	1950	2.0152	-	-
2.0	1990	-	3.1210	0.7187
2.0101	2000	1.924	-	-
2.0603	2050	1.7472	-	-
2.1106	2100	1.7485	-	-
2.1608	2150	1.7536	-	-
2.2111	2200	1.751	-	-
2.2613	2250	1.7172	-	-
2.3116	2300	1.7269	-	-
2.3618	2350	1.7352	-	-
2.4121	2400	1.7019	-	-
2.4623	2450	1.7278	-	-
2.5126	2500	1.7046	-	-
2.5628	2550	1.6962	-	-
2.6131	2600	1.6881	-	-
2.6633	2650	1.6806	-	-
2.7136	2700	1.6614	-	-
2.7638	2750	1.6918	-	-
2.8141	2800	1.6794	-	-
2.8643	2850	1.6708	-	-
2.9146	2900	1.6531	-	-
2.9648	2950	1.6236	-	-
3.0	2985	-	3.2556	0.7140

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: 0.32.1
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",
}

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	negative
type	string	string	string
details	min: 3 tokens mean: 14.78 tokens max: 110 tokens	min: 3 tokens mean: 14.78 tokens max: 110 tokens	min: 3 tokens mean: 10.19 tokens max: 29 tokens