LilMoo - a Polygl0t Collection

Polygl0t 's Collections

ViTucano-v1 (Portuguese)

Tucano (Portuguese)

TeenyTinyLlama (Portuguese)

LilMoo

updated 22 days ago

A 0.6-billion-parameter Hindi language models trained entirely from scratch.

Raising Bars, Not Parameters: LilMoo Compact Language Model for Hindi

Paper • 2603.03508 • Published Mar 3 • 4
Polygl0t/LilMoo-v0.1

Text Generation • 0.7B • Updated Mar 5 • 9

Note 🧱 Base model pretrained only with Hindi text.
Polygl0t/LilMoo-v0.2

Text Generation • 0.7B • Updated Mar 5 • 7

Note 🧱 Base model pretrained with a Hindi+ English mixture.
Polygl0t/gigalekh-v1

Viewer • Updated 24 days ago • 83.6M • 293

Note 📚 Pretraining dataset.
Polygl0t/GigaLekh-ablation-EDU-1.5B

Text Generation • 2B • Updated about 1 month ago • 2

Note 🔬 Ablation Experiment (Edu)
Polygl0t/hindi-toxicity-qwen-annotations

Viewer • Updated Mar 5 • 354k • 91

Note 📚 Annotations to train classifiers/filters (Toxicity).
Polygl0t/hindi-roberta-toxicity-classifier

Text Classification • 0.3B • Updated Mar 5 • 7

Note 🎯 Quality Filter (Toxicity)
Polygl0t/hindi-edu-qwen-annotations

Viewer • Updated Mar 5 • 400k • 17

Note 📚 Annotations to train classifiers/filters (Educational).
Polygl0t/hindi-roberta-edu-classifier

Text Classification • 0.3B • Updated Mar 5 • 3

Note 🎯 Quality Filter (Educational)
Polygl0t/GigaLekh-ablation-NonEDU-1.5B

Text Generation • 2B • Updated about 1 month ago • 3

Note 🔬 Ablation Experiment (NonEdu)
Polygl0t/tokenizers

Viewer • Updated Mar 5 • 8.98M • 507

Note 📚 Data used to train the LilMoo tokenizer.