XTuner Release V0.1.6

What's Changed

• [Feature] Support ChatGLM3-6B-Base by @LZHgrla in #177
• bump version to 0.1.6 by @LZHgrla in #178

Full Changelog: v0.1.5...v0.1.6

XTuner Release V0.1.5

What's Changed

• [Fix] Rename internlm-chat-20b by @LZHgrla in #131
• [Fix] Fix CPU OOM during the merge step by @LZHgrla in #133
• [Fix] Add --offload-folder for merge and chat by @LZHgrla in #140
• [Feature] Support to remove history for chat script by @LZHgrla in #144
• [Docs] add conda env create by @KevinNuNu in #147
• [Fix] Fix activation checkpointing bug by @LZHgrla in #159
• [Refactor] Refactor the preprocess of dataset by @LZHgrla in #163
• [Feature] Support deepspeed for HF trainer by @LZHgrla in #164
• [Feature] Support the fine-tuning of MSAgent dataset by @LZHgrla in #156
• [Fix] Fix bugs on traverse_dict by @LZHgrla in #141
• [Doc] Update chat.md by @LZHgrla in #168
• bump version to 0.1.5 by @LZHgrla in #171

New Contributors

• @KevinNuNu made their first contribution in #147

Full Changelog: v0.1.4...v0.1.5

XTuner Release V0.1.4

What's Changed

• [Fix] fix merge command in README.md by @liuyanyi in #111
• [Doc] Add doc about using alpaca format custom dataset by @HIT-cwh in #114
• [Feature] Support InternLM-20B by @LZHgrla in #128
• bump version to 0.1.4 by @LZHgrla in #129

New Contributors

• @liuyanyi made their first contribution in #111

Full Changelog: v0.1.3...v0.1.4

XTuner Release V0.1.3

What's Changed

• [Feature] Add Baichuan2 7B-chat, 13B-base, 13B-chat by @LZHgrla in #103
• [Fix] Use token_id instead of token for encode_fn & Set eval mode before generate by @LZHgrla in #107
• [Feature] Support log processed dataset & Fix doc by @HIT-cwh in #101
• [Fix] move toy data by @HIT-cwh in #108
• bump version to 0.1.3 by @HIT-cwh in #109

Full Changelog: v0.1.2...v0.1.3

XTuner Release V0.1.2

What's Changed

• [Doc] Fix dataset docs by @HIT-cwh in #87
• [Doc] Fix readme by @HIT-cwh in #92
• [Improve] Add ZeRO2-offload configs by @LZHgrla in #94
• [Improve] Redesign convert tools by @LZHgrla in #96
• [Fix] fix generation config by @HIT-cwh in #98
• [Feature] Support Baichuan2 models by @LZHgrla in #102
• bump version to 0.1.2 by @LZHgrla in #100

Full Changelog: v0.1.1...v0.1.2

XTuner Release V0.1.1

What's Changed

• [Doc] Update WeChat image by @LZHgrla in #74
• [Doc] Modify install commands for DeepSpeed integration by @LZHgrla in #75
• Add bot: Create .owners.yml by @del-zhenwu in #81
• [Improve] Add several InternLM-7B full parameters fine-tuning configs by @LZHgrla in #84
• [Feature] Add starcoder example by @HIT-cwh in #83
• [Doc] Add data_prepare.md docs by @LZHgrla in #82
• bump version to 0.1.1 by @HIT-cwh in #85

New Contributors

• @del-zhenwu made their first contribution in #81

Full Changelog: v0.1.0...v0.1.1

XTuner Release V0.1.0

Changelog

v0.1.0 (2023.08.30)

XTuner is released! 🔥🔥🔥

Highlights

• XTuner supports LLM fine-tuning on consumer-grade GPUs. The minimum GPU memory required for 7B LLM fine-tuning is only 8GB.
• XTuner supports various LLMs, datasets, algorithms and training pipelines.
• Several fine-tuned adapters are released simultaneously, including various gameplays such as the colorist LLM, plugins-based LLM, and many more. For further details, please visit XTuner on HuggingFace!