frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-11-04 16:04:58 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Lite Interpreter] Support features from to_backend (#52870)

Browse Source 
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/52870

Add the missing parts to support to_backend modules by lite interpreter.
1. Add ISINSTANCE instruction support, which is used in to_backend for output type check.
2. Bypass lite interpreter's type parser by checking the qualified name. If it starts with "torch.jit", use the same type resolver as nn module (starting with "__torch__").

Tests
Mobile module is serialized and loaded in ```BackendTest.TestCompiler```. The results are compared to those from original torchscript module.

Test Plan: Imported from OSS

Reviewed By: raziel

Differential Revision: D26715351

Pulled By: iseeyuan

fbshipit-source-id: ad9d74ee81c6aa692ab9e5dd7a9003bae5d4f01f
This commit is contained in:
Martin Yuan
2021-03-01 17:53:50 -08:00
committed by Facebook GitHub Bot
parent 8467e5cad3
commit b5ae8e69a7
5 changed files with 34 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
test/cpp/jit/test_backend.cpp
View File

@ -2,6 +2,7 @@
#include <test/cpp/jit/test_utils.h>
#include <torch/csrc/jit/api/module.h>
#include <torch/csrc/jit/backends/backend_detail.h>
#include <torch/csrc/jit/mobile/import.h>
#include <torch/torch.h>
// Tests go in torch::jit
@ -100,6 +101,12 @@ TEST(BackendTest, TestCompiler) {
"backend_with_compiler_demo", m, compile_spec, any_dict_ty);
auto res = lm.forward(inputs);
AT_ASSERT(res.toTensor().equal(ref.toTensor()));
std::stringstream ss;
lm._save_for_mobile(ss);
auto mlm = _load_for_mobile(ss);
auto mres = mlm.forward(inputs);
AT_ASSERT(mres.toTensor().equal(ref.toTensor()));
}
TEST(BackendTest, TestCompilerNotSupport) {