nomic-embed-indonesian

This is a sentence-transformers model finetuned from nomic-ai/nomic-embed-text-v1.5 specifically for Indonesian language text embedding tasks. It maps Indonesian 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.

🚀 Quick Start

from sentence_transformers import SentenceTransformer

# Load the model (requires trust_remote_code=True)
model = SentenceTransformer("asmud/nomic-embed-indonesian", trust_remote_code=True)

# Indonesian text examples
texts = [
    "search_query: Apa itu kecerdasan buatan?",
    "search_document: Kecerdasan buatan adalah teknologi yang memungkinkan mesin belajar",
    "classification: Produk ini sangat berkualitas (sentimen: positif)"
]

# Generate embeddings
embeddings = model.encode(texts)
print(f"Embedding shape: {embeddings.shape}")  # (3, 768)

🇮🇩 Specialized for Indonesian Language

This model is optimized for Indonesian text understanding across multiple domains including:

• Technology (Teknologi) - AI, gadgets, digital innovation
• Politics (Politik) - Government, elections, public policy
• Law (Hukum) - Legal affairs, crime, justice
• Economy (Ekonomi) - Business, finance, trade
• Education (Pendidikan) - Academic, learning, research
• Health (Kesehatan) - Medical, wellness, healthcare
• Sports (Olahraga) - Athletics, competitions, fitness
• Culture (Budaya) - Literature, arts, traditions
• And more...

Model Details

Model Description

• Model Type: Sentence Transformer
• Base model: nomic-ai/nomic-embed-text-v1.5
• Maximum Sequence Length: 8192 tokens
• Output Dimensionality: 768 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': 8192, 'do_lower_case': False, 'architecture': 'NomicBertModel'})
  (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

⚠️ Important: This model requires trust_remote_code=True due to custom model architecture.

Then you can load this model and run inference.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("asmud/nomic-embed-indonesian", trust_remote_code=True)
# Run inference with Indonesian text
sentences = [
    'search_query: Apa itu kecerdasan buatan?',
    'search_document: Kecerdasan buatan adalah teknologi yang memungkinkan mesin belajar dari data',
    'classification: Produk ini sangat berkualitas dan sesuai harapan (sentimen: positif)',
    'clustering: makanan tradisional Indonesia seperti rendang dan gudeg',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.7154, 0.7378],
#         [0.7154, 1.0000, 0.6583],
#         [0.7378, 0.6583, 1.0000]])

Evaluation

Metrics

Semantic Similarity

• Dataset: indonesian-diversity-eval
• Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.4358
spearman_cosine	0.2857

Training Details

Training Dataset

Unnamed Dataset

• Size: 6,294 training samples
• Columns: sentence_0, sentence_1, and label
• Approximate statistics based on the first 1000 samples:

  	sentence_0	sentence_1	label
  type	string	string	float
  details	min: 8 tokens mean: 20.45 tokens max: 181 tokens	min: 7 tokens mean: 117.93 tokens max: 508 tokens	min: 0.0 mean: 0.51 max: 1.0

• Samples:

  sentence_0	sentence_1	label
  clustering: artikel berita Indonesia	clustering: Paris Saint - Germain gagal mempertahankan status tak terkalahkan di Ligue 1 Prancis , setelah dipaksa menelan kekalahan perdana musim ini kala menyambangi Strasbourg . Tanda - tanda kurang maksimalnya performa klub ibukota Prancis ini sudah terlihat di awal pertandingan . Lini belakang gagal mengantisipasi skema tendangan bebas Strasbourg sehingga umpan Dimitri Lienard diteruskan dengan mudah oleh Nuno Da Costa pada menit ke - 13 untuk mencetak gol pembuka . Skuat asuhan Unai Emery langsung bermain agresif untuk mengejar ketertinggalan , mengandalkan trio Neymar , Kylian Mbappe dan Angel Di Maria . Nama terakhir mendapat kesempatan pada menit ke - 39 usai menerima umpan terobosan dari Neymar , tetapi sayang sepakannya gagal menemui sasaran meski sudah tidak dapat diantisipasi kiper . Mbappe akhirnya yang sukses mencatatkan namanya di papan skor . Mantan pemain Monaco itu menyambar umpan tarik Rabiot di dalam kotak penalti pada menit ke - 42 untuk membuat skor sama kuat . B...	1.0
  search_query: KPK resmi menetapkan Ketua DPR Setya Novanto sebag	search_document: KPK resmi menetapkan Ketua DPR Setya Novanto sebagai tersangka kasus korupsi pengadaan proyek e - KTP . Penetapan status tersangka yang kedua kalinya ini disampaikan Wakil Ketua KPK Saut Situmorang . Novanto dijerat dengan Pasal 2 ayat 1 subsider Pasal 3 Undang-Undang Nomor 31 tahun 1999 sebagaimana diubah dengan Undang-Undang Nomor 20 tahun 2001 tentang Pemberantasan Korupsi juncto Pasal 55 ayat 1 ke - 1 KUHP .	1.0
  search_query: Google memperkenalkan laptop chromebook kelas atas	classification: ga da wifi d lantai 2,kamar mandi ga da gantungan handuk or baju,over all bagus,n recomended (sentimen: positif)	0.0

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim"
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 1
• per_device_eval_batch_size: 1
• num_train_epochs: 1
• 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: 1
• per_device_eval_batch_size: 1
• 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: 1
• 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
• use_ipex: False
• bf16: False
• fp16: False
• 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: 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: False
• 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: False
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: round_robin
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	indonesian-diversity-eval_spearman_cosine
0.0794	500	0.0	-
0.1589	1000	0.0	-
0.2383	1500	0.0	-
0.3178	2000	0.0	-
0.3972	2500	0.0	-
0.4766	3000	0.0	-
0.5561	3500	0.0	-
0.6355	4000	0.0	-
0.7150	4500	0.0	-
0.7944	5000	0.0	-
0.8738	5500	0.0	-
0.9533	6000	0.0	-
1.0	6294	-	0.2857

Framework Versions

• Python: 3.11.13
• Sentence Transformers: 5.0.0
• Transformers: 4.54.1
• PyTorch: 2.7.1
• Accelerate: 1.9.0
• Datasets: 4.0.0
• Tokenizers: 0.21.4

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",
}

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}
}