Hi, I hoping to find a layer wise profiler which can profiler the forward time and backward time for each layer in pytorch / tensorflow. It seems that only this pyprof tool can achieve this target.
However, I found the document in nvidia website is missing and this tool can't tell whether a add kernel belongs to forward or backward.
Is there other links for the document of this tool? Or Is there other layer-wise profiler recommended?
Hi, I hoping to find a layer wise profiler which can profiler the forward time and backward time for each layer in pytorch / tensorflow. It seems that only this pyprof tool can achieve this target. However, I found the document in nvidia website is missing and this tool can't tell whether a add kernel belongs to forward or backward. Is there other links for the document of this tool? Or Is there other layer-wise profiler recommended?