Shun Zuo
Shun Zuo
> 感谢您的回复，我再去仔细检查一遍 请问你排查出问题了吗？我也遇到这个问题了
> 我刚刚将您的代码在voc2012跑起来了，这个应该是没影响的吧。 > > 我有一个小疑惑，我的卡是两张3090，所以我的bs应该是8，其它配置应该和您的是一样的，但是我在前两个epoch的有监督损失非常高，平均都在2.0左右，而我看您的日志中前两个epoch也是0.6左右，请问这是否和我的bs较小有关呢，此外，我是否应该调整一下学习率，请问您有什么好的建议吗，期待您的回答，感谢！ 怎么跑voc2012呢，能不能告诉我一下，修改代码的方式
> `RuntimeError: No executable batch size found, reached zero`这个错误请参考这两个issues：[issues/37](https://github.com/charent/ChatLM-mini-Chinese/issues/37)和[issues/40](https://github.com/charent/ChatLM-mini-Chinese/issues/40) 数据集去重时间太长了，能发布一下清洗后的数据吗
> Hi I've been following the instructions given but it takes very long to read subset for YFCC. I tried to run it on a single CPU and it has...
I encountered difficulties downloading the data cc12m and cc3m, and the vast majority of the images cannot be downloaded. Could you please publish these data to hugging_face and provide us...
感谢您能够回复我的issue！最近拜读的您文章，反复阅读了一周，每次阅读都能有新的体会，可视为论文教材。祝愿您的科研工作越来越好。再次感谢！！
hello,where is COCO_2017_train.json?
> Hi, thank you for your amazing work. Will codes for over clustering with PiCIE (PiCIE+H) released> hello，I can't find hello, where is COCO_2017_train.json, where is it, or how should...
> 有的，见[dpo_data_process.py](https://github.com/charent/ChatLM-mini-Chinese/blob/main/utils/dpo_data_process.py)和[raw_data_process.py](https://github.com/charent/ChatLM-mini-Chinese/blob/main/utils/raw_data_process.py) 数据去重太慢了怎么办
@Mollylulu I encountered difficulties downloading the data cc12m and cc3m, and the vast majority of the images cannot be downloaded. Could you please publish these data to hugging_face and provide...