Skip to content

PyThaiNLP/KhanomTanLLM

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

13 Commits
 
 
 
 
 
 
 
 

Repository files navigation

KhanomTanLLM

KhanomTan (Thai name is ขนมตาล) + LLM

Image gen from FLUX.1 [dev]

KhanomTan LLM is a bilingual language model trained in Thai and English from open source dataset by PyThaiNLP. We train the model from public dataset only. It is a fully open source model. We releses the dataset, training pipeline, and models.

Codename: numfa-v2

Blog Post (Thai): https://pythainlp.org/2024-09-12-khanomtanllm/

Instruct Models

We fine-turning model from wannaphong/KhanomTanLLM-Instruct-dataset. We doesn't have any safeguard, so use your risk.

To get the best result, we suggest the setting:

  • temperature: 2 - 4
  • min_p: > 0.6

Acknowledgements

Research supported with Cloud TPUs from Google's TPU Research Cloud (TRC). We use TPU4-64 for training model.

Thank you TPU Research Cloud and EasyLM project! We use EasyLM for pretraining model.