v0.6.2: ORPO and Qwen1.5-32B

[hiyouga]

New features

• Support ORPO algorithm by @hiyouga in support ORPO #3066
• Support inferring BNB 4-bit models in multiple GPUs via the quantization_device_map argument
• Reorganize README files, move examples scripts to the examples folder
• Support saving & loading arguments quickly in LlamaBoard by @hiyouga and @marko1616 in train params save feature #3046
• Support load alpaca-format dataset from the hub without dataset_info.json by specifying --dataset_dir ONLINE
• Add a parameter moe_aux_loss_coef to control the coefficient of auxiliary loss in MoE models.

New models

• Base models
  • Breeze-7B-Base
  • Qwen1.5-MoE-A2.7B (14B)
  • Qwen1.5-32B
• Instruct/Chat models
  • Breeze-7B-Instruct
  • Qwen1.5-MoE-A2.7B-Chat (14B)
  • Qwen1.5-32B-Chat

Bug fix

• Fix pile dataset download config by @lealaxy in Fix pile dataset download config #3053
• Fix model generation config by @marko1616 in Fix Llama model save for full param train #3057
• Fix qwen1.5 models DPO training by @changingivan and @hiyouga in add support qwen1.5-based DPO for Flash Attention 2  #3083
• Support Qwen1.5-32B by @sliderSun in support Qwen1.5-32B #3160
• Support Breeze-7B by @codemayq in support Breeze-7B #3161
• Fix addtional_target in unsloth by @kno10 in Pass additional_target to unsloth #3201
• Fix Dear developers, please add a configuration loading function in the web interface! #2807 更换 template 导致 train_bash.py FAILED，貌似与 cuda 管理有关 #3022 rtx-4090多卡推理（模型为qlora微调后qwen72b）是否支持？通过FSDP+QLoRA，可以正常对qwen-72b的模型进行微调，想问一下，如何使用rxt-4090对其进行推理部署呢？  #3023 train params save feature #3046 api_demo 使用问题：使用工具时Chatmessage的格式问题 #3077 Question on MixtralMOE SFT Loss #3085 在进行ppo训练时发生下面错误 #3116 --additional_target not supported with unsloth? #3200 FileNotFoundError: [Errno 2] No such file ./accuracies.png' #3225

---

This discussion was created from the release v0.6.2: ORPO and Qwen1.5-32B.