在3090上使用collie微调moss7B,flash_attn报错
Closed this issue · 2 comments
ShacklesLay commented
RuntimeError: FlashAttention backward for head dim > 64 requires A100 or H100 GPUs as the implementation needs a large amount of shared memory.
KaiLv69 commented
你好,moss 7B的head dim是128,flash-attn在3090上不支持,可以设置config.use_flash=False
ZiboZ commented
你好 FlashAttention不是根据 L1缓存的大小来划分块大小的嘛 理论上V100应该也可以支持的吧