I don’t believe there is a way to continue your last build incrementally, since it probably wasn’t using one of the above methods. If you got a killed message at some point (or in the dmesg kernel log), it means your board probably ran out-of-memory and that you need to mount more SWAP.
@3332190067 that’s great that it continued the build for you! The patches, I typically apply “by hand” (meaning just looking the diff file and manually making the changes in the associated PyTorch files) - this is to account for any differences in the version of PyTorch you’re building versus the version of PyTorch that I originally made the patch for. Yes, you can use the patch tool or git apply and it will overcome some minor differences, however I prefer to “know” what has changed as to not encounter strange/cryptic errors later on. Plus, the patches are short to apply, and you only need to apply it once (not everytime you run setup.py)
Hi @3332190067, assuming that you already made edits manually to pytorch source files, try running git status and git diff from your pytorch directory, and it will show you a diff of the changes that you made.