Skip to content

Commit a6ad6fe

Browse files
tastelikefeettastelikefeet
authored
Support DPO (#130)
1 parent 4d6ebeb commit a6ad6fe

File tree

27 files changed

+1801
-269
lines changed

27 files changed

+1801
-269
lines changed

‎README.md‎

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -101,6 +101,7 @@ Or use ModelScope's [official image](https://www.modelscope.cn/docs/intro/enviro
101101

102102
## Changelog
103103

104+
- 🎉2026-03-28 Support DPO training with both Transformers and Megatron backends. See [dpo_full.py](cookbook/rl/dpo_full.py) and [dpo_lora.py](cookbook/rl/dpo_lora.py).
104105
- 🎉2026-03-24 Twinkle Web site is now live at https://modelscope.github.io/twinkle-web/
105106
- 🎉2026-03-19 Support GKD training ，please refer to this [cookbook](cookbook/rl/gkd_on_policy.py).
106107
- 🎉2026-02-13 Initial version of Twinkle✨ released, including SFT/PT/RL support for text models.
@@ -137,8 +138,8 @@ supported on Twinkle✨ framework.
137138
| qwen3 series | [Qwen/Qwen3-14B-Base](https://modelscope.cn/models/Qwen/Qwen3-14B-Base) | 0.6B/1.7B/4B/8B/14B | transformers>=4.51 | ✔ | [Qwen/Qwen3-14B-Base](https://huggingface.co/Qwen/Qwen3-14B-Base) |
138139
| | [Qwen/Qwen3-32B](https://modelscope.cn/models/Qwen/Qwen3-32B) | 0.6B/1.7B/4B/8B/14B/32B | transformers>=4.51 | ✔ | [Qwen/Qwen3-32B](https://huggingface.co/Qwen/Qwen3-32B) |
139140
| qwen3_moe series | [Qwen/Qwen3-30B-A3B-Base](https://modelscope.cn/models/Qwen/Qwen3-30B-A3B-Base) | 30B-A3B/A3B-Base,235B-A22B | transformers>=4.51 | ✔ | [Qwen/Qwen3-30B-A3B-Base](https://huggingface.co/Qwen/Qwen3-30B-A3B-Base) |
140-
| qwen3.5 moe series | [Qwen/Qwen3.5-35B-A3B](https://www.modelscope.cn/models/Qwen/Qwen3.5-35B-A3B) | 35B-A3B,122B-A10B, etc. | transformers>=5.20 | ✔ | [Qwen/Qwen3.5-35B-A3B](https://huggingface.co/Qwen/Qwen3.5-35B-A3B) |
141-
| qwen3.5 series | [Qwen/Qwen3.5-9B](https://www.modelscope.cn/models/Qwen/Qwen3.5-9B) | 2B ~ 27B | transformers>=5.20 | ✔ | [Qwen/Qwen3.5-9B](https://huggingface.co/Qwen/Qwen3.5-9B) |
141+
| qwen3.5 moe series | [Qwen/Qwen3.5-35B-A3B](https://www.modelscope.cn/models/Qwen/Qwen3.5-35B-A3B) | 35B-A3B,122B-A10B, etc. | transformers>=5.2.0 | ✔ | [Qwen/Qwen3.5-35B-A3B](https://huggingface.co/Qwen/Qwen3.5-35B-A3B) |
142+
| qwen3.5 series | [Qwen/Qwen3.5-9B](https://www.modelscope.cn/models/Qwen/Qwen3.5-9B) | 2B ~ 27B | transformers>=5.2.0 | ✔ | [Qwen/Qwen3.5-9B](https://huggingface.co/Qwen/Qwen3.5-9B) |
142143
| qwen2 series | [Qwen/Qwen2-0.5B-Instruct](https://modelscope.cn/models/Qwen/Qwen2-0.5B-Instruct) | 0.5B/1.5B/7B/72B | transformers>=4.37 | ✔ | [Qwen/Qwen2-0.5B-Instruct](https://huggingface.co/Qwen/Qwen2-0.5B-Instruct) |
143144
| | [Qwen/Qwen2-1.5B](https://modelscope.cn/models/Qwen/Qwen2-1.5B) | 0.5B/1.5B/7B/72B | transformers>=4.37 | ✔ | [Qwen/Qwen2-1.5B](https://huggingface.co/Qwen/Qwen2-1.5B) |
144145
| | [Qwen/Qwen2.5-1.5B-Instruct](https://modelscope.cn/models/Qwen/Qwen2.5-1.5B-Instruct) | 0.5B/1.5B/3B/7B/14B/32B/72B | transformers>=4.37 | ✔ | [Qwen/Qwen2.5-1.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-1.5B-Instruct) |

‎README_ZH.md‎

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -91,6 +91,7 @@ Twinkle✨支持相同的算法接口运行在单GPU、torchrun多机、Ray、Cl
9191

9292
## 更新日志
9393

94+
🎉2026-03-28 支持 DPO 训练，同时支持 Transformers 和 Megatron 后端。参考 [dpo_full.py](cookbook/rl/dpo_full.py) 和 [dpo_lora.py](cookbook/rl/dpo_lora.py)。
9495
🎉2026-03-24 Twinkle 站点上线，访问地址 https://modelscope.github.io/twinkle-web/
9596
🎉2026-03-19 支持GKD蒸馏能力，参考[cookbook](cookbook/rl/gkd_on_policy.py)。
9697
🎉2026-02-13 Twinkle✨ 初始版本发布，支持文本模型的SFT/PT/RL训练。我们还通过兼容Tinker的API，在魔搭社区上提供了无服务器训练功能。
@@ -120,8 +121,8 @@ Twinkle✨支持相同的算法接口运行在单GPU、torchrun多机、Ray、Cl
120121
| qwen3 全系列 | [Qwen/Qwen3-14B-Base](https://modelscope.cn/models/Qwen/Qwen3-14B-Base) | 0.6B/1.7B/4B/8B/14B | transformers>=4.51 | ✔ | [Qwen/Qwen3-14B-Base](https://huggingface.co/Qwen/Qwen3-14B-Base) |
121122
| | [Qwen/Qwen3-32B](https://modelscope.cn/models/Qwen/Qwen3-32B) | 0.6B/1.7B/4B/8B/14B/32B | transformers>=4.51 | ✔ | [Qwen/Qwen3-32B](https://huggingface.co/Qwen/Qwen3-32B) |
122123
| qwen3_moe 全系列 | [Qwen/Qwen3-30B-A3B-Base](https://modelscope.cn/models/Qwen/Qwen3-30B-A3B-Base) | 30B-A3B/A3B-Base，235B-A22B | transformers>=4.51 | ✔ | [Qwen/Qwen3-30B-A3B-Base](https://huggingface.co/Qwen/Qwen3-30B-A3B-Base) |
123-
| qwen3.5 moe 全系列 | [Qwen/Qwen3.5-35B-A3B](https://www.modelscope.cn/models/Qwen/Qwen3.5-35B-A3B) | 35B-A3B,122B-A10B, etc. | transformers>=5.20 | ✔ | [Qwen/Qwen3.5-35B-A3B](https://huggingface.co/Qwen/Qwen3.5-35B-A3B) |
124-
| qwen3.5 全系列 | [Qwen/Qwen3.5-9B](https://www.modelscope.cn/models/Qwen/Qwen3.5-9B) | 2B ~ 27B | transformers>=5.20 | ✔ | [Qwen/Qwen3.5-9B](https://huggingface.co/Qwen/Qwen3.5-9B) |
124+
| qwen3.5 moe 全系列 | [Qwen/Qwen3.5-35B-A3B](https://www.modelscope.cn/models/Qwen/Qwen3.5-35B-A3B) | 35B-A3B,122B-A10B, etc. | transformers>=5.2.0 | ✔ | [Qwen/Qwen3.5-35B-A3B](https://huggingface.co/Qwen/Qwen3.5-35B-A3B) |
125+
| qwen3.5 全系列 | [Qwen/Qwen3.5-9B](https://www.modelscope.cn/models/Qwen/Qwen3.5-9B) | 2B ~ 27B | transformers>=5.2.0 | ✔ | [Qwen/Qwen3.5-9B](https://huggingface.co/Qwen/Qwen3.5-9B) |
125126
| qwen2 全系列 | [Qwen/Qwen2-0.5B-Instruct](https://modelscope.cn/models/Qwen/Qwen2-0.5B-Instruct) | 0.5B/1.5B/7B/72B | transformers>=4.37 | ✔ | [Qwen/Qwen2-0.5B-Instruct](https://huggingface.co/Qwen/Qwen2-0.5B-Instruct) |
126127
| | [Qwen/Qwen2-1.5B](https://modelscope.cn/models/Qwen/Qwen2-1.5B) | 0.5B/1.5B/7B/72B | transformers>=4.37 | ✔ | [Qwen/Qwen2-1.5B](https://huggingface.co/Qwen/Qwen2-1.5B) |
127128
| | [Qwen/Qwen2.5-1.5B-Instruct](https://modelscope.cn/models/Qwen/Qwen2.5-1.5B-Instruct) | 0.5B/1.5B/3B/7B/14B/32B/72B | transformers>=4.37 | ✔ | [Qwen/Qwen2.5-1.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-1.5B-Instruct) |

0 commit comments

Comments
 (0)