Skip to content

disable mixed precision in bn fusion & clone #93

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
EikanWang merged 2 commits into from
Jun 24, 2020
Merged

disable mixed precision in bn fusion & clone #93

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
d7e8833
disable mixed precision in bn fusion
pinzhenx Jun 23, 2020
878b578
disable mixed precision in dil_clone
pinzhenx Jun 24, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 6 additions & 0 deletions intel_pytorch_extension_py/ops/jit_script.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,5 @@
import torch
import intel_pytorch_extension as ipex
import _torch_ipex as core
from torch.jit._recursive import wrap_cpp_module

Expand All @@ -13,7 +14,12 @@ def script_(obj, optimize=None, _frames_up=0, _rcb=None):
torch.jit.script = script_

if core.get_jit_opt():
# Disable mix precision in model fusion, since mixed precision cannot
# bring any benefits for inference, but will lead to loss of accuracy
orig_mixed_type = ipex.get_auto_mix_precision()
ipex.enable_auto_mix_precision(None)
jit_m = wrap_cpp_module(torch._C._jit_pass_fold_convbn(jit_m._c))
ipex.enable_auto_mix_precision(orig_mixed_type)

return jit_m

Expand Down
1 change: 0 additions & 1 deletion torch_ipex/csrc/cpu/DevOPs.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1351,7 +1351,6 @@ at::Tensor AtenIpexCPUDev::dil_clone(const at::Tensor& self, c10::optional<c10::
DEBUG("AtenIpexCPUDev::dil_clone\n");
CHECK_DNNL_OP_PRE_COND(self);

dbl::comm::reorder_to_bf16_for_mix_prec(self);
dil::tensor src = dbl::comm::try_gen_dil_tensor(self);
dil::tensor dst;
dil::direct_copy::compute(src, dst);
Expand Down