frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-21 05:34:18 +08:00
Code Issues Packages Projects Releases Wiki Activity

Changes to transition to generic API for ops with weight prepacking (#35010)

Browse Source 
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/35010

semantics.

This PR moves all the xnnpack specific interfces to a generic interface.
Accordingly removes xnnpac specific reference from API and some variable
names.
What has not yet changed:

TODO:
USE_XNNPACK is still used. This can be removed where no XNNPACK
specific things are done. e.g., RegisterOpContext.cpp and
xnnpack_rewrite.cpp.
Also the filename and structure also remains. Some of the generic class
definition can be moved non-XNNPACK specific folder.

Test Plan:
python test/test_xnnpack_integration.py

Imported from OSS

Differential Revision: D20526416

fbshipit-source-id: 2e1725345c44bbb26bdc448097a7384eca121387
This commit is contained in:
Kimish Patel
2020-03-22 08:24:42 -07:00
committed by Facebook GitHub Bot
parent 1ff5d9c557
commit e1c092fe3a
14 changed files with 436 additions and 374 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
torch/csrc/jit/python/init.cpp
View File

@ -460,19 +460,19 @@ void initJITBindings(PyObject* module) {
return debugGetFusedKernelCode(g, inps);
})
.def(
"_jit_pass_insert_xnnpack_ops",
"_jit_pass_insert_prepacked_ops",
[](std::shared_ptr<Graph>& graph) {
return insertXNNPACKOps(graph);
return insertPrePackedOps(graph);
})
.def(
"_jit_pass_insert_xnnpack_ops",
"_jit_pass_insert_prepacked_ops",
[](script::Module& module) {
return insertXNNPACKOps(module);
return insertPrePackedOps(module);
})
.def(
"_jit_pass_fold_xnnpack_prepack_ops",
"_jit_pass_fold_prepacking_ops",
[](script::Module& module) {
return FoldXNNPACKPrePackingOps(module);
return FoldPrePackingOps(module);
})
.def(
"_jit_pass_onnx_unpack_quantized_weights",