tqchen commented on PR #18271:
URL: https://github.com/apache/tvm/pull/18271#issuecomment-3258216362
Thanks @yaoyaoding
> I place the cache at ~/.cache/tvm-ffi. Does this folder sounds good? Do we
need to store the modules in a subfolder of the tvm-ffi cache?
We can go with your suggestion here
- TVM_FFI_CACHE_DIR/TVM_FFI_CUDA_ARCH_LIST: both sounds good to me, as I
think it is good to follow pytorch practice
> This PR uses <ffi-cache>/<name>_<hash> as the cache dir for this specific
module. Any alternative idea to orgnaize the cache?
I think this is great, seems aligns with existing practices
> To avoid introduce dependency like filelock, this PR adds a utility class
FileLock. Do we want to reuse any existing utility in tvm?
Our intention is to make tvm_ffi a standalone minimal lib, so this change
sounds good
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
