Liang Zhang

Results 2 comments of Liang Zhang

typo in docker command - documentation

thats the reason why i failed, thanks very much.

[BUG]: fp32 param and grad have different shape torch.Size([5064704]) vs torch.Size([128000]) when use lora_rank=4 at stage 1

i have a similar issue too ``` │ 153 │ def optimizer_step(self, optimizer: optim.Optimizer, **kwargs) -> None: │ │ ❱ 154 │ │ optimizer.step() │ │ 155 │ │ │...