pytorch
f89f2d42 - Memory format propagation parser patch (#1675)

Commit

2 years ago

Memory format propagation parser patch (#1675) mark memory format on a few ops that does support memory format propagation (clamp/softplus/threshold) This causes us to output tensor in wrong shape, causing failure on torchvision models pytorch#76791

Author

jjsjann123

Parents

a18e1094

Files2

test
- test_jit_cuda_fuser.py
torch/csrc/jit/codegen/cuda
- parser.cpp

pytorch f89f2d42 - Memory format propagation parser patch (#1675)

pytorch
f89f2d42 - Memory format propagation parser patch (#1675)