Bill-Orz

Results 5 comments of Bill-Orz

[BUG] RuntimeError: inflight params error when using DeepSpeed for Reinforcement Learning

facing the same issue

[BUG] RuntimeError: inflight params error when using DeepSpeed for Reinforcement Learning

I update to the latest version, but when i use zero stage 3, i'm hanging with the log "[2023-07-11 16:01:28,870] [INFO] [partition_parameters.py:326:__exit__] finished initializing model with 6.74B parameters". ![image](https://github.com/microsoft/DeepSpeed/assets/122064954/496da155-81cf-45a9-8c8c-caf992e34f98) ![image](https://github.com/microsoft/DeepSpeed/assets/122064954/a04abd2a-11c2-4ae0-9be8-3d73e0f7ee33)...

element 0 of tensors does not require grad and does not have a grad_fn

我跟原始deepspeed-chat代码对比了一下，看起来是一致的

element 0 of tensors does not require grad and does not have a grad_fn

谢谢，请问你这边step3是用了4个7B的llama来完成的吗？我这边80G A100，不开gradient checkpointing，一直会OOM

element 0 of tensors does not require grad and does not have a grad_fn

我用了4张卡