LlamaFactory Releases
37 releases of hiyouga/LlamaFactory
- v0.5.0v0.5.0: Agent Tuning, Unsloth Integration
Celebrates 10k stars by adding agent tuning, Unsloth integration for speed, and function calling support.
Jan 20, 2024
- v0.4.0v0.4.0: Mixtral-8x7B, DPO-ftx, AutoGPTQ Integration
A core refactor introducing Mixtral support, AutoGPTQ integration, and several breaking changes to core arguments.
Dec 16, 2023
- v0.3.3v0.3.3: ModelScope Integration, Reward Server
Adds ModelScope Hub integration and a reward model server for API demos and PPO training.
Dec 3, 2023
- v0.3.2v0.3.2: Patch release
A patch release adding support for training GPTQ quantized models and resuming reward model training.
Nov 21, 2023
- v0.3.0v0.3.0: Full-Parameter RLHF
Initial full-parameter RLHF release featuring core refactoring and support for full-parameter training.
Nov 16, 2023
- v0.2.2v0.2.2: Patch release
Minimal release notes: "### Bug fix - Fix the OOM issue in PPO training by @mmbwf in #424 - Fix fine-tuning arguments by @yyq in #1454 - "
Nov 13, 2023
- v0.2.1v0.2.1: Variant Models, NEFTune Trick
Adds NEFTune support, a wide variety of new model variants, and dataset loading improvements including ShareGPT formats and caching.
Nov 9, 2023
- v0.2.0v0.2.0: Web UI Refactor, LongLoRA
Major refactor introducing LongLoRA, support for training large models (Qwen-14B, InternLM-20B), Ascend NPU support, and benchmark integration.
Oct 15, 2023
- v0.1.8v0.1.8: FlashAttention-2 and Baichuan2
Focuses on performance optimization with FlashAttention-2 and adds Baichuan2 training capabilities with automatic LoRA module selection.
Sep 11, 2023
- v0.1.7v0.1.7: Script Preview and RoPE Scaling
Introduces training script previews, checkpoint resumption, and RoPE scaling techniques for LLaMA models.
Aug 18, 2023
- v0.1.6v0.1.6: DPO Training and Qwen-7B
Adds Direct Preference Optimization (DPO) training and expands model support to Qwen and XVERSE.
Aug 11, 2023
- v0.1.5v0.1.5: Patch release
Minimal release notes: "- Fix LLaMA-2 template #307 - Fix bug in preprocessing 968ce0dcce6bfef582ce37aea6566a65f5aac811 - Fix #294 #296"
Aug 2, 2023
- v0.1.4v0.1.4: Dataset Streaming
Minimal release notes: "- Support [dataset streaming](https://huggingface.co/docs/datasets/stream) - Fix LLaMA-2 #268 - Fix DeepSpeed ZeRO-3 "
Aug 1, 2023
- v0.1.3v0.1.3: Patch release
Jul 21, 2023
- v0.1.2v0.1.2: LLaMA-2 Models
Adds LLaMA-2 support and fixes critical configuration issues including ZeRO-3 and API bugs.
Jul 20, 2023
- v0.1.1
Initial Web UI features including configuration options and a demo, alongside bug fixes for the reward mechanism.
Jul 18, 2023
- v0.1.0v0.1.0: All-in-one Web UI
Minimal release notes: "- Fix gradient accumulation in PPO Trainer https://github.com/hiyouga/ChatGLM-Efficient-Tuning/issues/299 - All-in-one "
Jul 17, 2023