SentenceTransformer based on sentence-transformers/all-MiniLM-L6-v2

This is a sentence-transformers model finetuned from sentence-transformers/all-MiniLM-L6-v2. 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 Type: Sentence Transformer
Base model: sentence-transformers/all-MiniLM-L6-v2
Maximum Sequence Length: 512 tokens
Output Dimensionality: 384 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': '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

# Download from the 🤗 Hub
model = SentenceTransformer("along26/all-MiniLM-L6-v2_multilingual_malaysian-v9")
# Run inference
sentences = [
    'According to Bintulu police chief ACP Zailanni Amit, the late Bermau Bagu, 70, was in the garden before being shot by the suspect, his brother-in-law, who was also hunting in the garden.',
    'Nitih ku Ketuai Polis Pelilih Bintulu, ACP Zailanni Amit, rambau penusah nya nyadi, niang ti benama Bermau Bagu, 70 taun benung ba kebun nya sebedau kena timbak suspek, ipar niang empu, ke bela ngasu dalam kandang kebun nya.',
    'What is finance generally?',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.1941, 0.9593],
#         [0.1941, 1.0000, 0.1978],
#         [0.9593, 0.1978, 1.0000]])

Training Details

Training Dataset

Unnamed Dataset

Size: 420,570 training samples
Columns: sentence_0, sentence_1, and sentence_2

Approximate statistics based on the first 1000 samples:

	sentence_0	sentence_1	sentence_2
type	string	string	string
details	min: 7 tokens mean: 164.03 tokens max: 512 tokens	min: 6 tokens mean: 197.0 tokens max: 512 tokens	min: 6 tokens mean: 178.02 tokens max: 512 tokens

Samples:

sentence_0	sentence_1	sentence_2
`Why have some analysts suggested that the outcome of Najib Razak's corruption trial could have significant implications for Malaysia's political landscape and the future of its democracy?`	`Mengapa sesetengah penganalisis mencadangkan bahawa keputusan perbicaraan rasuah Najib Razak boleh memberi implikasi yang besar kepada landskap politik Malaysia dan masa depan demokrasinya?`	A warming climate can significantly affect the emergence times of certain insect species, as many insects are ectothermic and rely on external environmental conditions to regulate their body temperature. This means that their development, reproduction, and behavior are closely linked to temperature and other climatic factors. As global temperatures rise, these changes can lead to shifts in the timing of insect emergence, which can have cascading effects on ecosystem interactions and services. 1. Phenological shifts: Warmer temperatures can lead to earlier emergence of insects, a phenomenon known as phenological shifts. This can result in a mismatch between the timing of insect emergence and the availability of their food resources, such as plants or other prey species. This mismatch can negatively impact the survival and reproduction of insects, as well as the species that depend on them for food. 2. Altered species interactions: Shifts in insect emergence times can also affect speci...
Corruption can have a significant impact on economic development and social inequality in Malaysia. Economic Development: Corruption can hinder economic development by discouraging investment, distorting markets, and undermining the rule of law. When businesses and individuals perceive a high level of corruption in a country, they may be less likely to invest or start businesses there, as they see it as a risky and unpredictable environment. This can limit the creation of jobs and the growth of industries, ultimately hindering economic development. Corruption can also distort markets by giving certain individuals or companies an unfair advantage. For example, if a company is able to secure a government contract through bribery or cronyism, it can stifle competition and lead to lower quality goods and services at higher prices. This can discourage innovation and entrepreneurship, further hindering economic development. Social Inequality: Corruption can also contribute to social inequ...	Rasuah boleh memberi kesan yang ketara kepada pembangunan ekonomi dan ketidaksamaan sosial di Malaysia. Pembangunan Ekonomi: Rasuah boleh menghalang pembangunan ekonomi dengan menggalakkan pelaburan, memutarbelitkan pasaran, dan menjejaskan kedaulatan undang-undang. Apabila perniagaan dan individu melihat tahap rasuah yang tinggi di sesebuah negara, mereka mungkin kurang berkemungkinan untuk melabur atau memulakan perniagaan di sana, kerana mereka melihatnya sebagai persekitaran yang berisiko dan tidak dapat diramalkan. Ini boleh mengehadkan penciptaan pekerjaan dan pertumbuhan industri, akhirnya menghalang pembangunan ekonomi. Rasuah juga boleh memutarbelitkan pasaran dengan memberikan individu atau syarikat tertentu kelebihan yang tidak adil. Sebagai contoh, jika syarikat mampu mendapatkan kontrak kerajaan melalui rasuah atau kronisme, ia boleh menyekat persaingan dan membawa kepada barangan dan perkhidmatan berkualiti rendah pada harga yang lebih tinggi. Ini boleh menghalang inova...	`"What are the specific mechanisms through which immunoglobulins act to neutralize antigens and prevent infections?"`
`He, who is also Minister of Public Health, Housing and Local Government Councils, said there were 302,243 Sarawakians at risk or recipients who should be more than 60 years old and had the second dose of COVID-19 on April`	`Iya ti mega Menteri Pengerai Mensia Mayuh, Pengawa Berumah enggau Kaunsil Kandang Menua madahka, bisi 302,243 rayat Sarawak ti bisi risiko tauka penerima ke patut beumur lebih 60 taun merima tuchuk kedua dos penyungkak kedua COVID-19 berengkah kena 12 April tu tadi.`	`What are the basic skills required to be a good programmer?`

Loss: TripletLoss with these parameters:

{
    "distance_metric": "TripletDistanceMetric.EUCLIDEAN",
    "triplet_margin": 5
}

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 32
per_device_eval_batch_size: 32
num_train_epochs: 4
fp16: True
multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: no
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: 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: 4
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: 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}
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

Click to expand

Epoch	Step	Training Loss
0.0380	500	4.7509
0.0761	1000	2.033
0.1141	1500	1.5296
0.1522	2000	1.3832
0.1902	2500	1.3084
0.2283	3000	1.3416
0.2663	3500	1.3118
0.3043	4000	1.3306
0.3424	4500	1.2771
0.3804	5000	1.2593
0.4185	5500	1.2278
0.4565	6000	1.207
0.4946	6500	1.1735
0.5326	7000	1.1842
0.5706	7500	1.1501
0.6087	8000	1.1562
0.6467	8500	1.1422
0.6848	9000	1.1229
0.7228	9500	1.0865
0.7609	10000	1.1094
0.7989	10500	1.0848
0.8369	11000	1.0957
0.8750	11500	1.0564
0.9130	12000	1.0688
0.9511	12500	0.9947
0.9891	13000	1.048
1.0272	13500	1.0183
1.0652	14000	1.0139
1.1032	14500	1.0291
1.1413	15000	1.001
1.1793	15500	0.9803
1.2174	16000	0.9874
1.2554	16500	0.9895
1.2935	17000	0.9721
1.3315	17500	0.9689
1.3696	18000	0.9622
1.4076	18500	0.9234
1.4456	19000	0.9039
1.4837	19500	0.9223
1.5217	20000	0.9091
1.5598	20500	0.9377
1.5978	21000	0.9174
1.6359	21500	0.9039
1.6739	22000	0.9009
1.7119	22500	0.8912
1.7500	23000	0.9378
1.7880	23500	0.9056
1.8261	24000	0.8748
1.8641	24500	0.8869
1.9022	25000	0.8972
1.9402	25500	0.8856
1.9782	26000	0.87
2.0163	26500	0.869
2.0543	27000	0.8255
2.0924	27500	0.8421
2.1304	28000	0.8196
2.1685	28500	0.8292
2.2065	29000	0.8374
2.2445	29500	0.8101
2.2826	30000	0.8329
2.3206	30500	0.8073
2.3587	31000	0.8015
2.3967	31500	0.8221
2.4348	32000	0.7914
2.4728	32500	0.7768
2.5108	33000	0.8036
2.5489	33500	0.7825
2.5869	34000	0.7981
2.6250	34500	0.779
2.6630	35000	0.7965
2.7011	35500	0.783
2.7391	36000	0.7748
2.7771	36500	0.7962
2.8152	37000	0.7782
2.8532	37500	0.7611
2.8913	38000	0.7877
2.9293	38500	0.757
2.9674	39000	0.7789
3.0054	39500	0.7745
3.0434	40000	0.7471
3.0815	40500	0.7299
3.1195	41000	0.7119
3.1576	41500	0.7199
3.1956	42000	0.7318
3.2337	42500	0.7446
3.2717	43000	0.7316
3.3097	43500	0.7534
3.3478	44000	0.704
3.3858	44500	0.7005
3.4239	45000	0.713
3.4619	45500	0.7492
3.5000	46000	0.7337
3.5380	46500	0.7025
3.5760	47000	0.753
3.6141	47500	0.7378
3.6521	48000	0.7242
3.6902	48500	0.7123
3.7282	49000	0.7277
3.7663	49500	0.7272
3.8043	50000	0.7094
3.8423	50500	0.7074
3.8804	51000	0.7162
3.9184	51500	0.6984
3.9565	52000	0.693
3.9945	52500	0.7026

Framework Versions

Python: 3.12.12
Sentence Transformers: 5.1.2
Transformers: 4.57.1
PyTorch: 2.9.0+cu126
Accelerate: 1.11.0
Datasets: 4.0.0
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",
}

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}
}