FX pass: fuse_sparse_matmul_add (#69340)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/69340
- An FX pass to fuse ops resulting from addmm(a, b.t())
- Used to enable structured sparsity using TRT
Reviewed By: 842974287
Differential Revision: D32456684
fbshipit-source-id: 601826af216cea314ee85ed522d5c54a5151d720
Author
Marat Subkhankulov