verl icon indicating copy to clipboard operation
verl copied to clipboard

RuntimeError: CUDA error: uncorrectable ECC error encountered!!!

Open asirgogogo opened this issue 11 months ago • 2 comments

训练了几个step后遇到这样的错误,是OOM了吗? Image

asirgogogo avatar Feb 27 '25 11:02 asirgogogo

训练机器:4x8H800 80G

asirgogogo avatar Feb 27 '25 11:02 asirgogogo

ecc error is machine problem.

vermouth1992 avatar Feb 27 '25 14:02 vermouth1992