along26/all-MiniLM-L6-v2_multilingual_malaysian-v8

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-v8")
# Run inference
sentences = [
    'To calculate the distance between the object and the lens, we can use the lens formula:\n\n1/f = 1/v - 1/u\n\nwhere f is the focal length of the lens, v is the distance between the lens and the image, and u is the distance between the object and the lens.\n\nGiven the focal length (f) is 20 cm and the distance between the lens and the image (v) is 25 cm, we can plug these values into the formula:\n\n1/20 = 1/25 - 1/u\n\nNow, we need to solve for u:\n\n1/u = 1/20 - 1/25\n1/u = (25 - 20) / (20 * 25)\n1/u = 5 / 500\n1/u = 1/100\n\nu = 100 cm\n\nSo, the distance between the object and the lens is 100 cm.',
    'Untuk mengira jarak antara objek dan kanta, kita boleh menggunakan formula kanta:\n\n1/f = 1/v - 1/u\n\ndengan f ialah panjang fokus kanta, v ialah jarak antara kanta dan imej, dan u ialah jarak antara objek dan kanta.\n\nMemandangkan panjang fokus (f) ialah 20 cm dan jarak antara kanta dan imej (v) ialah 25 cm, kita boleh memasukkan nilai ini ke dalam formula:\n\n1/20 = 1/25 - 1/u\n\nSekarang, kami perlu menyelesaikan untuk anda:\n\n1/u = 1/20 - 1/25\n1/u = (25 - 20) / (20 * 25)\n1/u = 5 / 500\n1/u = 1/100\n\nu = 100 cm\n\nJadi, jarak antara objek dan kanta ialah 100 cm.',
    "Najib Razak's trial for corruption and money laundering has had a significant negative impact on public perception of his leadership and the United Malays National Organization (UMNO) party. Najib is the former Prime Minister of Malaysia, and the charges against him are related to the 1MDB scandal, in which billions of dollars were allegedly misappropriated from a Malaysian state investment fund.\n\nThe trial has brought renewed attention to issues of corruption and cronyism in Malaysia, and many Malaysians feel that the UMNO party, which Najib led until his election defeat in 2018, is tainted by association. The party, which has long been a dominant force in Malaysian politics, has seen its popularity decline as a result of the scandal and the trial.\n\nThe trial has also raised questions about the effectiveness of Malaysia's anti-corruption efforts and the rule of law in the country. Many Malaysians are disappointed that it took so long for Najib to be charged and brought to trial, and some feel that the legal process has been politically motivated.\n\nOverall, the trial has contributed to a sense of disillusionment and mistrust among the Malaysian public, and has damaged the reputation of the UMNO party and its leaders. It remains to be seen how the trial will ultimately be resolved, and what the long-term impact will be on Malaysian politics and society.",
]
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, 0.4679, 0.9782],
#         [0.4679, 1.0000, 0.4707],
#         [0.9782, 0.4707, 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: 6 tokens mean: 153.22 tokens max: 512 tokens	min: 6 tokens mean: 182.56 tokens max: 512 tokens	min: 6 tokens mean: 165.2 tokens max: 512 tokens

• Samples:

  sentence_0	sentence_1	sentence_2
  Kenu ku iya, enti deka nyadi kaban RELA, orang nya patut beumur 18 taun ke atas lalu deka ditapis piak di Bukit Aman dikena nentuka orang ke ngerejista diri nyadi kaban RELA beresi ari sebarang pengawa ke enda menuku.	Kenu ku iya, enti deka nyadi kaban RELA, orang nya patut beumur 18 taun ke atas lalu deka ditapis piak di Bukit Aman dikena nentuka orang ke ngerejista diri nyadi kaban RELA beresi ari sebarang pengawa ke enda menuku.	"How can the study of non-perturbative gauge dynamics in the context of string theory shed light on the behaviors and interactions of particles in high-energy physics experiments?"
  Let A = {1,2,3}. Find the power set of A and show that its cardinality is larger than the cardinality of A itself.	Biarkan A = {1,2,3}. Cari set kuasa A dan tunjukkan bahawa kardinalitinya lebih besar daripada kardinaliti A itu sendiri.	How can the Malaysian housing market still be so unaffordable for the average citizen?
  What is the standard enthalpy change for the dissolution of sodium chloride (NaCl) in water, given that 5.00 g of NaCl is dissolved in 100.0 mL of water and the resulting solution has a final temperature of 25.0°C? The molar enthalpy of dissolution of NaCl is -3.9 kJ/mol.	Apakah perubahan entalpi standard untuk pembubaran natrium klorida (NaCl) dalam air, memandangkan 5.00 g NaCl dibubarkan dalam 100.0 ml air dan penyelesaian yang dihasilkan mempunyai suhu akhir 25.0 ° C? Entalpi molar pembubaran NaCl ialah -3.9 kJ/mol.	Why have some opposition politicians and activists been critical of the government's handling of the 1MDB scandal and Najib Razak's prosecution?

• 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
• 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: 3
• 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

Epoch	Step	Training Loss
0.0380	500	4.7496
0.0761	1000	2.0391
0.1141	1500	1.5157
0.1522	2000	1.3982
0.1902	2500	1.3333
0.2283	3000	1.311
0.2663	3500	1.3084
0.3043	4000	1.3112
0.3424	4500	1.3168
0.3804	5000	1.2842
0.4185	5500	1.3008
0.4565	6000	1.3239
0.4946	6500	1.3005
0.5326	7000	1.2905
0.5706	7500	1.2811
0.6087	8000	1.2178
0.6467	8500	1.1743
0.6848	9000	1.1273
0.7228	9500	1.0966
0.7609	10000	1.0909
0.7989	10500	1.0586
0.8369	11000	1.0047
0.8750	11500	0.9998
0.9130	12000	1.0508
0.9511	12500	1.0211
0.9891	13000	0.9711
1.0272	13500	0.961
1.0652	14000	0.9487
1.1032	14500	0.9381
1.1413	15000	0.9497
1.1793	15500	0.9295
1.2174	16000	0.9247
1.2554	16500	0.9079
1.2935	17000	0.8922
1.3315	17500	0.9216
1.3696	18000	0.9004
1.4076	18500	0.8797
1.4456	19000	0.8717
1.4837	19500	0.8594
1.5217	20000	0.8711
1.5598	20500	0.8664
1.5978	21000	0.8623
1.6359	21500	0.8599
1.6739	22000	0.8259
1.7119	22500	0.8739
1.7500	23000	0.8532
1.7880	23500	0.8567
1.8261	24000	0.8519
1.8641	24500	0.8309
1.9022	25000	0.8207
1.9402	25500	0.8312
1.9782	26000	0.8329
2.0163	26500	0.8022
2.0543	27000	0.7744
2.0924	27500	0.7795
2.1304	28000	0.7567
2.1685	28500	0.7797
2.2065	29000	0.7711
2.2445	29500	0.7691
2.2826	30000	0.7578
2.3206	30500	0.7783
2.3587	31000	0.7182
2.3967	31500	0.7639
2.4348	32000	0.7484
2.4728	32500	0.7674
2.5108	33000	0.7663
2.5489	33500	0.764
2.5869	34000	0.7376
2.6250	34500	0.7471
2.6630	35000	0.7437
2.7011	35500	0.7562
2.7391	36000	0.74
2.7771	36500	0.7208
2.8152	37000	0.7392
2.8532	37500	0.7336
2.8913	38000	0.7192
2.9293	38500	0.7383
2.9674	39000	0.7432

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}
}