Datasets:

izzako
/

balinese-pixelgpt

text_id int64	chunk_id int64	pixel_values list	grapheme_token_ids list	llama_token_ids list	text string
0	0	[[255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,25(...TRUNCATED)	[1,6,65,28,16,19,0,222,1280,41,38,41,673,16,8,53,8,4,7,49,6,10,230,11,162,238,149,7,8,53,44,53,113,8(...TRUNCATED)	[1,476,11989,14950,3304,1019,29891,5859,5653,29875,3772,29875,29911,381,941,29906,9753,29948,25364,1(...TRUNCATED)	"Kaca Utama ProyékWiki WikiTirta2 sané kalaksanayang ring warsa 2024 inggih punika parikrama indik(...TRUNCATED)
0	1	[[255,255,255,255,255,255,255,193,59,238,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255(...TRUNCATED)	[1,149,7,8,53,44,53,53,8,4,12735,199,0,73,25,16,666,41,0,18,14,0,359,183,92,4,38,41,673,16,8,53,8,4,(...TRUNCATED)	[1,1370,4977,29871,29906,29900,29906,29906,288,492,29882,7107,5441,294,7494,3384,18182,279,29889,377(...TRUNCATED)	"warsa 2022 olih Komunitas Wikimedia Denpasar. WikiTirta2 ngambil genah ring Kabupatén Badung, Taba(...TRUNCATED)
1	0	[[255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,122,111,163,216,144,139,211,25(...TRUNCATED)	[1,9,0,125,1142,0,2215,15433,0,125,1142,0,2215,76,172,29,25,6,21,199,191,82,450,27,7,49,6,57,12,0,19(...TRUNCATED)	[1,319,9481,309,29915,29879,13741,7420,319,9481,309,29915,29879,13741,7420,2348,3146,29882,6035,4106(...TRUNCATED)	"A Devil's Chaplain A Devil's Chaplain inggih punika silih tunggil buku sané kasurat olih Richard D(...TRUNCATED)
2	0	[[255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,240,255,255,255,201,164,96,255,69,255,(...TRUNCATED)	[1,9,0,6,0,0,0,0,6,0,0,0,26,7,15,0,49,21,14,0,9,0,6,0,10,16,0,4,13,172,29,25,6,21,199,191,82,1372,22(...TRUNCATED)	[1,319,22231,27326,397,319,22231,27326,397,313,6500,29874,1894,265,743,423,29901,319,22231,8365,2340(...TRUNCATED)	"Afrika Kelod Afrika Kelod (basa Indonésia: Afrika Selatan) inggih punika silih tunggil nagari ring(...TRUNCATED)
3	0	[[255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,25(...TRUNCATED)	[1,26,7,15,170,155,25,26,7,15,170,155,25,28,16,38,15,181,4,358,68,13,172,29,25,6,26,7,9,364,7,144,15(...TRUNCATED)	[1,4886,29874,379,20041,3270,4886,29874,379,20041,3270,318,941,4353,379,14108,29899,29965,29878,700,(...TRUNCATED)	"Basa Hindustani Basa Hindustani utawi Hindi-Urdu inggih punika basa asli saking India kalér, sané(...TRUNCATED)
4	0	[[255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,25(...TRUNCATED)	[1,57,6,23,47,17,4,82,14,378,57,6,23,47,0,9,230,12,26,30,0,57,6,23,47,0,4,13,172,29,25,6,21,199,191,(...TRUNCATED)	[1,2166,1335,29893,2219,29892,21569,8553,2166,1335,29893,2219,313,10327,2518,350,2606,29901,29871,22(...TRUNCATED)	"Sukawati, Gianyar Sukawati (aksara Bali: ᬲᬸᬓᬯᬢᬶ) inggih punika silih tunggil kecamatan (...TRUNCATED)
5	0	[[255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,25(...TRUNCATED)	[1,0,26,17,4,242,6,0,203,0,26,0,9,230,12,26,30,0,0,26,0,4,13,172,29,25,6,21,199,191,82,0,65,19,16,0,(...TRUNCATED)	[1,476,15330,29892,14320,1335,323,996,801,476,15330,313,29909,2039,2518,350,2606,29901,29871,228,175(...TRUNCATED)	"Koba, Bangka Tengah Koba (Aksara Bali: ᬓᭀᬩ) inggih punika silih tunggil kecamatan ring Kabupa(...TRUNCATED)
6	0	[[255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,215,102,255,69,255(...TRUNCATED)	[1,52,41,9498,123,0,4,157,6,1145,96,52,41,9498,123,0,9,230,12,26,30,0,52,41,32,15,123,0,4,13,172,29,(...TRUNCATED)	[1,350,2679,277,3159,273,29892,349,574,11311,12687,574,350,2679,277,3159,273,313,29909,2039,2518,350(...TRUNCATED)	"Bukit Intan, Pangkalpinang Bukit Intan (Aksara Bali: ᬩᬸᬓᬶᬢ᭄ᬳᬶᬦ᭄ᬢᬦ᭄) in(...TRUNCATED)
7	0	[[255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,215,102,255,69,255(...TRUNCATED)	[1,9,30,0,0,0,30,0,0,1653,172,29,25,6,21,199,191,82,1047,0,0,6,740,144,26,30,92,4,6,145,14,25,0,199,(...TRUNCATED)	[1,10785,264,10517,10785,264,10517,2348,3146,29882,6035,4106,269,2638,29882,260,686,29887,309,868,29(...TRUNCATED)	"Alien Child Alien Child inggih punika silih tunggil duo vokal saking Bali. Kakaryanin olih duo voka(...TRUNCATED)
8	0	[[255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,25(...TRUNCATED)	[1,28,11464,0,0,1969,8,105,44,106,8,4,0,4,28,533,30,0,4,0,1969,8,105,44,106,8,4,0,4,17,4,13,172,29,2(...TRUNCATED)	[1,17676,11143,313,7858,29899,29947,29900,29941,29897,17676,376,29931,1992,29908,313,7858,29899,2994(...TRUNCATED)	"USS Little (DD-803) USS \"Little\" (DD-803), inggih punika silih tunggil kapal parusak kelas Fletch(...TRUNCATED)

End of preview. Expand in Data Studio

Balinese PixelGPT Dataset

This dataset contains preprocessed Balinese text data for training PixelGPT models.

Dataset Statistics

Language: Balinese (bali)
Total samples: 54,467
Train samples: 54,017
Test samples: 450

Tokenizers

Grapheme tokenizer: izzako/javanese-llama-tokenizer
LLaMA tokenizer: ernie-research/DualGPT

Features

text_id: Document identifier
chunk_id: Chunk identifier within document
pixel_values: Rendered pixel representation of aksara text
grapheme_token_ids: Token IDs from grapheme-based tokenizer (with EOS token)
llama_token_ids: Token IDs from LLaMA tokenizer (with EOS token)
text: Original text chunk

Renderer Configuration

Renderer path: ../../renderers/m4_renderer

Usage

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("izzako/balinese-pixelgpt")

# Access train and test splits
train_data = dataset['train']
test_data = dataset['test']

# Example: Get first sample
sample = train_data[0]
print(sample['text'])

Citation

If you use this dataset, please cite:

@dataset{balinese_pixelgpt,
  title = {Balinese PixelGPT Dataset},
  author = {Musa Izzanardi Wijanarko},
  year = {2025},
  publisher = {Hugging Face},
  url = {https://huggingface.co/datasets/izzako/balinese-pixelgpt}
}

Downloads last month: 70

Size of downloaded dataset files:

323 MB

Size of the auto-converted Parquet files:

323 MB

Number of rows:

54,467