TyrionZK

vc++中调用cygwin编译的崩溃

1

我在该项目cmakelists.txt中添加了新的dll，尝试将simple_vad.h中的函数导出，然后在vc++中调用。 根据网上资料编写了def，利用lib工具生成了lib，vc++可以编译通过，但是运行时崩溃，一直没搞定。

[BUG]: some param and grad of critic_optim are not in the same device after backward

1

### 🐛 Describe the bug After self.strategy.backward(critic_loss, self.critic, self.critic_optim), some param and grad of critic_optim are not in the same device . I don't know what happen ### Environment device:...

[BUG]: train_sft.py LLAMA-7B colossalai_zero2 OOM

4

### 🐛 Describe the bug I run train_sft.py to finetune LLAMA-7B with colossalai_zero2 batch_size=1 max_len=512，but OOM happen. theoretically, memory usage of single GPU is about (2+(2+12)/4)*7=38.5G, plus usage for model...