🔥Released Models Version <1.0.0-alpha> (03/08/23)

🇹🇭 OpenThaiGPT 1.0.0-alpha (3 August 2023)

🇹🇭 OpenThaiGPT Version 1.0.0-alpha is the first Thai implementation of a 7B-parameter LLaMA v2 Chat model finetuned to follow Thai translated instructions and makes use of the Huggingface LLaMA implementation.

Web Demo:

Gradio

https://demo.openthaigpt.aieat.or.th

Colab Demo:

Google Colaboratory

https://colab.research.google.com/drive/1kDQidCtY9lDpk49i7P3JjLAcJM04lawu?usp=sharing

Change Logs

🇹🇭 Version 1.0.0-alpha (Facebook LLama V2 Model)

Release date: 3 August 2023

🇹🇭 OpenThaiGPT Version 1.0.0-alpha is the first Thai implementation of a 7B-parameter LLaMA v2 Chat model finetuned to follow Thai translated instructions and makes use of the Huggingface LLaMA implementation.

Changes

(1) Using Facebook LLama v2 model 7b chat as a base model which is pretrained on over 2 trillion token. (2) Context Length is upgrade from 2048 token to 4096 token (3) Allow research and commerical use.

License

Source Code: License Apache Software License 2.0. Weight: Research and commercial uses.

Code and Weight

Colab Demo: https://colab.research.google.com/drive/1kDQidCtY9lDpk49i7P3JjLAcJM04lawu?usp=sharing Finetune Code: https://github.com/OpenThaiGPT/openthaigpt-finetune-010beta (Same code as OpenThaiGPT 0.1.0-beta) Inference Library: https://github.com/OpenThaiGPT/openthaigpt Weight (Lora Adapter): https://huggingface.co/openthaigpt/openthaigpt-1.0.0-alpha-7b-chat Weight (Huggingface Checkpoint): https://huggingface.co/openthaigpt/openthaigpt-1.0.0-alpha-7b-chat-ckpt-hf

Authors

• Kobkrit Viriyayudhakorn ([e-mail])

• Sumeth Yuenyong ([e-mail])

• Thaweewat Rugsujarit ([e-mail])

• Jillaphat Jaroenkantasima ([e-mail])

• Norapat Buppodom ([e-mail])

• Koravich Sangkaew ([e-mail])

• Peerawat Rojratchadakorn ([e-mail])

• Surapon Nonesung ([e-mail])

• Chanon Utupon ([e-mail])

• Sadhis Wongprayoon ([e-mail])

• Nucharee Thongthungwong ([e-mail])

• Chawakorn Phiantham ([e-mail])

• Patteera Triamamornwooth ([e-mail])

• Nattarika Juntarapaoraya ([e-mail])

• Kriangkrai Saetan ([e-mail])

• Pitikorn Khlaisamniang ([e-mail])

---

Version 0.1.0-beta (Facebook LLama Model)

Release date: 16 May 2023

OpenThaiGPT Version 0.1.0-beta is a 7B-parameter LLaMA model finetuned to follow Thai translated instructions below and makes use of the Huggingface LLaMA implementation.

Statistics

Number of parameters: 7B Dimension: 4096 Context Length: 2048 n heads: 32 n layers: 32 n tokens: 1T

License

Source Code: License Apache Software License 2.0. Weight: For research use only (due to the Facebook LLama's Weight LICENSE). Note that: A commercial use license for OpenThaiGPT 0.1.0 weight will be released later soon!

Code and Weight

Finetune Code: https://github.com/OpenThaiGPT/openthaigpt-finetune-010beta Inference Library: https://github.com/OpenThaiGPT/openthaigpt Weight (Lora Adapter): https://huggingface.co/kobkrit/openthaigpt-0.1.0-beta

Authors

Kobkrit Viriyayudhakorn ([e-mail]), Sumeth Yuenyong ([e-mail]) and Thaweewat Rugsujarit ([e-mail]).

Trained Datasets

Dataset Name	Instruction Pairs	Descriptions
Thaweewat/alpaca-finance-43k-th	43,000	Alpaca Finance Instruction translated into Thai by Thaweewat Ruksujarit.
kobkrit/rd-taxqa	600	RD's Tax QA Chatbot Training set by ทรงวุฒิ บุรงค์
datasets/iapp_wiki_qa_squad	4,000	iApp Technology's Extractive QA Dataset in Thai language
Thaweewat/databricks-dolly-15k-th	15,000	Databrick's Dolly Instruction translated into Thai by Thaweewat Ruksujarit.
Thaweewat/instruction-wild-52k-th	52,000	Instruction Wild's translated into Thai by Thaweewat Ruksujarit.
Thaweewat/alpaca-cleaned-52k-th	51,000	Standford Alpaca's translated into Thai by Thaweewat Ruksujarit.
Thaweewat/gpteacher-20k-th	20,000	GPT Teacher's Instruction translated into Thai by Thaweewat Ruksujarit.
Thaweewat/onet-m6-social	600	ONET m6 Social Exam
datasets/Thaweewat/hc3-24k-th	24,000	Hello Simple AI Summary Dataset translated into Thai by Thaweewat Ruksujarit.
OpenThaiGPT Self Instruct (https://docs.google.com/spreadsheets/d/1BSHkpRyD5RH90E85tLWe4UzpgfDHZafE2rKxLincyWI/edit?usp=sharing)	5,000	Thai SelfInstruct Dataset (Automatic Generated) by OpenThaiGPT

---

Version 0.1.0-alpha (ByT5-XL Model)

Release date: 24 April 2023 PoC Testing Website: https://colab.research.google.com/drive/1Uds0ioOZSZrJ9m2FgW3DHlqVRFNHVRtu#scrollTo=qPJIpwuz4ltF Model and Weight: https://huggingface.co/kobkrit/openthaigpt-0.1.0-alpha PIP Installation Page: https://pypi.org/project/openthaigpt/ Code Example: https://colab.research.google.com/drive/1Uds0ioOZSZrJ9m2FgW3DHlqVRFNHVRtu#scrollTo=qPJIpwuz4ltF ----

OpenThaiGPT version 0.1.0-alpha

Thai First 3 billion params models

• First Thai Byte-Level Text-to-Text Transfer Transformer

• Support Instruction following

  • Translation to Thai

  • Explanation

  • Paraphase

• Zero-shot and Few-shot Learning

• Pretraining Model: ByT5-XL (3.74 billion params)

• InstructDataset: 50,000 Thai SelfInstruct

• RLHF: None

• Developer: Sumeth Yuenyong, Kobkrit Viriyayudhakorn ([e-mail])

PoC Version 0.0.4 (The Fourth PoC Version)

Release date: 12 March 2023 PoC Testing Website: https://colab.research.google.com/drive/13yLIifBRDQp82QO4ICs_aEvz0N8tqVPm?usp=sharin Model and Weight: https://huggingface.co/kobkrit/openthaigpt-gpt2-instructgpt-poc-0.0.4 PIP Installation Page: https://pypi.org/project/openthaigpt/ Code Example: https://github.com/OpenThaiGPT/openthaigpt-example ----

OpenThaiGPT version 0.0.4

The Fourth PoC Model

• ตอบคำถามได้ลงรายละเอียดมากขึ้น และตอบคำถามได้ดีขึ้นกว่า 0.0.3 เป็นส่วนมาก

• Pretraining Model: GPT-2 Thai-base

• InstructDataset: 300,000 Pantip + 5,000 Wiki QA => 12,920 Thai InstructGPT

• RLHF: None

• Developer: Kobkrit Viriyayudhakorn ([e-mail])

PoC Version 0.0.3 (The Third PoC Version)

Release date: 28 February 2023 Model and Weight: https://huggingface.co/kobkrit/openthaigpt-gpt2-instructgpt-poc-0.0.3 PIP Installation Page: https://pypi.org/project/openthaigpt/ Code Example: https://github.com/OpenThaiGPT/openthaigpt-example ----

OpenThaiGPT version 0.0.3

The Third PoC Model

• Pretraining Model: GPT-2 Thai-base

• InstructDataset: 300,000 Pantip + 5,000 Wiki QA => 7,000 Thai InstructGPT

• RLHF: None

• Developer: Kobkrit Viriyayudhakorn ([e-mail])

PoC Version 0.0.2 (The Second PoC Version)

Release date: 27 February 2023 Model and Weight: https://huggingface.co/kobkrit/openthaigpt-gpt2-instructgpt-poc-0.0.2 PIP Installation Page: {Coming Soon} Colab Example: {Coming Soon} ----

OpenThaiGPT version 0.0.2

The Second PoC Model

• Pretraining Model: GPT-2 Thai-base

• InstructDataset: 7,000 Thai InstructGPT

• RLHF: None

Developer: Kobkrit Viriyayudhakorn ([e-mail])

PoC Version 0.0.1 (Very First PoC Version)

Release date: 20 February 2023 Model and Weight: openthaigpt-gpt2-pantipwiki-poc PIP Installation Page: {Coming Soon} Colab Example: {Coming Soon} ----

The Very First PoC Model

• Pretraining Model: GPT-2 Thai-base

• InstructDataset: 298,678 QA Pairs getting from 70,000 Pantip katoos + Wikipedia QA by iApp

• RLHF: None

• Developer: Kobkrit Viriyayudhakorn ([e-mail])