metadata
tags:
- sentence-transformers
- sentence-similarity
- feature-extraction
- generated_from_trainer
- dataset_size:22654
- loss:ContrastiveLoss
- loss:TripletLoss
- loss:CoSENTLoss
- loss:MultipleNegativesRankingLoss
base_model: sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2
widget:
- source_sentence: Network Operations Specialist yêu cầu tối ưu hóa mạng.
sentences:
- >-
Actor cần có kỹ năng biểu diễn sân khấu và hóa thân vào nhiều loại nhân
vật.
- Network Operations Specialist cần tối ưu hóa mạng.
- >-
Nhà tư vấn PR hỗ trợ doanh nghiệp trong việc phát triển hình ảnh công
chúng và xử lý khủng hoảng.
- source_sentence: Cybersecurity Specialist với kinh nghiệm bảo mật hệ thống 5 năm.
sentences:
- Kỹ sư cơ khí cần phát triển hệ thống sản xuất tự động hóa.
- Cybersecurity Engineer, yêu cầu tối thiểu 5 năm trong bảo mật.
- Data Scientist cần kỹ năng Machine Learning và Python.
- source_sentence: Tư vấn môi trường hỗ trợ kiểm soát ô nhiễm môi trường đô thị.
sentences:
- Quản lý chất thải có kinh nghiệm xử lý và tái chế nước.
- Tư vấn môi trường quản lý chất lượng môi trường đô thị.
- >-
Illustrator cần có khả năng minh họa cho sách giáo dục và tài liệu học
tập.
- source_sentence: Mobile Developer với kinh nghiệm phát triển ứng dụng iOS và Swift.
sentences:
- Tuyển iOS Developer có kỹ năng làm việc với Swift.
- Tuyển chuyên viên QA kiểm tra chất lượng phần mềm.
- Mobile Developer cần biết phát triển ứng dụng đa nền tảng.
- source_sentence: Mobile Developer, kinh nghiệm lập trình ứng dụng iOS với Swift.
sentences:
- Tuyển kỹ sư cơ khí giám sát dây chuyền sản xuất.
- >-
Công ty XYZ tuyển Data Scientist với tối thiểu 2 năm kinh nghiệm học
máy.
- Tuyển iOS Developer thành thạo Swift.
datasets:
- HZeroxium/job-cv-binary
- HZeroxium/cv-job-triplet
- HZeroxium/cv-job-similarity
- HZeroxium/job-paraphrase
- HZeroxium/cv-paraphrase
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
- cosine_accuracy
- cosine_accuracy_threshold
- cosine_f1
- cosine_f1_threshold
- cosine_precision
- cosine_recall
- cosine_ap
- pearson_cosine
- spearman_cosine
model-index:
- name: >-
SentenceTransformer based on
sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2
results:
- task:
type: binary-classification
name: Binary Classification
dataset:
name: Unknown
type: unknown
metrics:
- type: cosine_accuracy
value: 0.9755351681957186
name: Cosine Accuracy
- type: cosine_accuracy_threshold
value: 0.5808850526809692
name: Cosine Accuracy Threshold
- type: cosine_f1
value: 0.9779005524861878
name: Cosine F1
- type: cosine_f1_threshold
value: 0.5644330978393555
name: Cosine F1 Threshold
- type: cosine_precision
value: 0.9833333333333333
name: Cosine Precision
- type: cosine_recall
value: 0.9725274725274725
name: Cosine Recall
- type: cosine_ap
value: 0.9956042554162885
name: Cosine Ap
- type: cosine_accuracy
value: 0.9968051118210862
name: Cosine Accuracy
- type: cosine_accuracy_threshold
value: 0.7650139331817627
name: Cosine Accuracy Threshold
- type: cosine_f1
value: 0.9984
name: Cosine F1
- type: cosine_f1_threshold
value: 0.7650139331817627
name: Cosine F1 Threshold
- type: cosine_precision
value: 1
name: Cosine Precision
- type: cosine_recall
value: 0.9968051118210862
name: Cosine Recall
- type: cosine_ap
value: 0.9999999999999999
name: Cosine Ap
- type: cosine_accuracy
value: 0.9936305732484076
name: Cosine Accuracy
- type: cosine_accuracy_threshold
value: 0.8211346864700317
name: Cosine Accuracy Threshold
- type: cosine_f1
value: 0.9968051118210862
name: Cosine F1
- type: cosine_f1_threshold
value: 0.8211346864700317
name: Cosine F1 Threshold
- type: cosine_precision
value: 1
name: Cosine Precision
- type: cosine_recall
value: 0.9936305732484076
name: Cosine Recall
- type: cosine_ap
value: 1
name: Cosine Ap
- task:
type: triplet
name: Triplet
dataset:
name: Unknown
type: unknown
metrics:
- type: cosine_accuracy
value: 1
name: Cosine Accuracy
- task:
type: semantic-similarity
name: Semantic Similarity
dataset:
name: Unknown
type: unknown
metrics:
- type: pearson_cosine
value: 0.970012297655986
name: Pearson Cosine
- type: spearman_cosine
value: 0.9430534588122865
name: Spearman Cosine
SentenceTransformer based on sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2
This is a sentence-transformers model finetuned from sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2 on the binary, triplet, similarity, job_paraphrase and cv_paraphrase datasets. It maps sentences & paragraphs to a 384-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 Sources
Full Model Architecture
SentenceTransformer(
(0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: BertModel
(1): Pooling({'word_embedding_dimension': 384, '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.
from sentence_transformers import SentenceTransformer
model = SentenceTransformer("paraphrase-multilingual-MiniLM-L12-v2-job-cv-multi-dataset")
sentences = [
'Mobile Developer, kinh nghiệm lập trình ứng dụng iOS với Swift.',
'Tuyển iOS Developer thành thạo Swift.',
'Tuyển kỹ sư cơ khí giám sát dây chuyền sản xuất.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
Evaluation
Metrics
Binary Classification
Metric |
Value |
cosine_accuracy |
0.9755 |
cosine_accuracy_threshold |
0.5809 |
cosine_f1 |
0.9779 |
cosine_f1_threshold |
0.5644 |
cosine_precision |
0.9833 |
cosine_recall |
0.9725 |
cosine_ap |
0.9956 |
Triplet
Metric |
Value |
cosine_accuracy |
1.0 |
Semantic Similarity
Metric |
Value |
pearson_cosine |
0.97 |
spearman_cosine |
0.9431 |
Binary Classification
Metric |
Value |
cosine_accuracy |
0.9968 |
cosine_accuracy_threshold |
0.765 |
cosine_f1 |
0.9984 |
cosine_f1_threshold |
0.765 |
cosine_precision |
1.0 |
cosine_recall |
0.9968 |
cosine_ap |
1.0 |
Binary Classification
Metric |
Value |
cosine_accuracy |
0.9936 |
cosine_accuracy_threshold |
0.8211 |
cosine_f1 |
0.9968 |
cosine_f1_threshold |
0.8211 |
cosine_precision |
1.0 |
cosine_recall |
0.9936 |
cosine_ap |
1.0 |
Training Details
Training Datasets
binary
triplet
similarity
job_paraphrase
cv_paraphrase
Evaluation Datasets
binary
triplet
similarity
job_paraphrase
cv_paraphrase
Training Hyperparameters
Non-Default Hyperparameters
eval_strategy
: steps
per_device_train_batch_size
: 32
per_device_eval_batch_size
: 32
learning_rate
: 2e-05
num_train_epochs
: 5
warmup_ratio
: 0.1
fp16
: True
batch_sampler
: no_duplicates
All Hyperparameters
Click to expand
overwrite_output_dir
: False
do_predict
: False
eval_strategy
: steps
prediction_loss_only
: True
per_device_train_batch_size
: 32
per_device_eval_batch_size
: 32
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
: 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
: 5
max_steps
: -1
lr_scheduler_type
: linear
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
: 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}
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
: 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
Epoch |
Step |
Training Loss |
binary loss |
triplet loss |
similarity loss |
job paraphrase loss |
cv paraphrase loss |
cosine_ap |
cosine_accuracy |
spearman_cosine |
0 |
0 |
- |
- |
- |
- |
- |
- |
1.0 |
0.9682 |
0.5468 |
0.2817 |
200 |
2.401 |
- |
- |
- |
- |
- |
- |
- |
- |
0.5634 |
400 |
1.5659 |
- |
- |
- |
- |
- |
- |
- |
- |
0.7042 |
500 |
- |
0.0088 |
0.2391 |
6.9067 |
0.1746 |
0.2689 |
1.0 |
0.9936 |
0.9123 |
0.8451 |
600 |
1.8501 |
- |
- |
- |
- |
- |
- |
- |
- |
1.1268 |
800 |
1.7318 |
- |
- |
- |
- |
- |
- |
- |
- |
1.4085 |
1000 |
1.3758 |
0.0079 |
0.0367 |
6.2019 |
0.1665 |
0.2657 |
1.0 |
1.0 |
0.9238 |
1.6901 |
1200 |
1.3554 |
- |
- |
- |
- |
- |
- |
- |
- |
1.9718 |
1400 |
1.5119 |
- |
- |
- |
- |
- |
- |
- |
- |
2.1127 |
1500 |
- |
0.0081 |
0.0144 |
5.7135 |
0.1633 |
0.2295 |
1.0 |
1.0 |
0.9341 |
2.2535 |
1600 |
1.2886 |
- |
- |
- |
- |
- |
- |
- |
- |
2.5352 |
1800 |
1.1131 |
- |
- |
- |
- |
- |
- |
- |
- |
2.8169 |
2000 |
1.3962 |
0.0108 |
0.0191 |
6.0231 |
0.1540 |
0.2342 |
1.0 |
1.0 |
0.9396 |
3.0986 |
2200 |
1.2394 |
- |
- |
- |
- |
- |
- |
- |
- |
3.3803 |
2400 |
1.1392 |
- |
- |
- |
- |
- |
- |
- |
- |
3.5211 |
2500 |
- |
0.0097 |
0.0025 |
5.6361 |
0.1580 |
0.2212 |
1.0 |
1.0 |
0.9410 |
3.6620 |
2600 |
1.1614 |
- |
- |
- |
- |
- |
- |
- |
- |
3.9437 |
2800 |
1.2351 |
- |
- |
- |
- |
- |
- |
- |
- |
4.2254 |
3000 |
1.1862 |
0.0100 |
0.0107 |
5.5943 |
0.1517 |
0.2158 |
1.0 |
1.0 |
0.9420 |
4.5070 |
3200 |
0.9371 |
- |
- |
- |
- |
- |
- |
- |
- |
4.7887 |
3400 |
1.3572 |
- |
- |
- |
- |
- |
- |
- |
- |
4.9296 |
3500 |
- |
0.0104 |
0.0057 |
5.6213 |
0.1539 |
0.2141 |
1.0 |
1.0 |
0.9429 |
5.0 |
3550 |
- |
- |
- |
- |
- |
- |
1.0 |
1.0 |
0.9431 |
Framework Versions
- Python: 3.12.4
- Sentence Transformers: 3.3.0
- Transformers: 4.46.2
- PyTorch: 2.5.1+cu124
- Accelerate: 1.1.1
- Datasets: 3.1.0
- Tokenizers: 0.20.3
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",
}
ContrastiveLoss
@inproceedings{hadsell2006dimensionality,
author={Hadsell, R. and Chopra, S. and LeCun, Y.},
booktitle={2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)},
title={Dimensionality Reduction by Learning an Invariant Mapping},
year={2006},
volume={2},
number={},
pages={1735-1742},
doi={10.1109/CVPR.2006.100}
}
TripletLoss
@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}
}
CoSENTLoss
@online{kexuefm-8847,
title={CoSENT: A more efficient sentence vector scheme than Sentence-BERT},
author={Su Jianlin},
year={2022},
month={Jan},
url={https://kexue.fm/archives/8847},
}
MultipleNegativesRankingLoss
@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}
}