Files
pytorch/test/cpp/jit
Zachary DeVito fb4517132f Allow 'Any' to appear as a type argument. (#26572)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/26572

Combined with isinstance specialization this allows a degree of polymorphic
functions to work without needing to use our weirder overload hacks.

We do not define any operators on Any, so the only thing you can do with it
is to put it in containers or type refine it using an isinstance check.
Any is restricted from appearing in non-argument position because we
cannot restore type tags if it ends up as a field in a class.

Test Plan: Imported from OSS

Differential Revision: D17530643

Pulled By: zdevito

fbshipit-source-id: f06f78ce84819f7773953a492f3d4c49219ee94c
2019-10-16 11:07:08 -07:00
..
2019-08-18 16:49:56 -07:00
2019-08-18 16:49:56 -07:00
2019-10-12 09:51:57 -07:00
2019-08-18 16:49:56 -07:00
2019-08-20 16:17:59 -07:00
2019-08-18 16:49:56 -07:00

JIT C++ Tests

How to add a new test

First, create a new test file. Test files should have be placed in this directory, with a name that starts with test_, like test_foo.cpp.

Here is an example test file you can copy-paste.

#include <test/cpp/jit/test_base.h>

// Tests go in torch::jit
namespace torch {
namespace jit {

// 1. Test cases are void() functions.
// 2. They start with the prefix `test`
void testCaseOne() {
    // ...
}

void testCaseTwo() {
    // ...
}
}
}

Then, register your test in tests.h:

// Add to TH_FORALL_TESTS_CUDA instead for CUDA-requiring tests
#define TH_FORALL_TESTS(_)             \
  _(ADFormulas)                        \
  _(Attributes)                        \
  ...
  _(CaseOne)  // note that the `test` prefix is omitted.
  _(CaseTwo)

We glob all the test files together in CMakeLists.txt so that you don't have to edit it every time you add a test. Unfortunately, this means that in order to get the build to pick up your new test file, you need to re-run cmake:

python setup.py build --cmake

Why do we have two different test runners?

We have two different ways of running our cpp tests:

  1. With gtest, from a standalone binary.
  2. With Python, from TestJit.test_cpp and TestJit.test_cpp_cuda (in test/test_jit.py)

We want both because we need to test things from a pure-C++ environment and with all our various Python patch-points enabled.

How do I run the tests?

The following commands assume you are in PyTorch root.

  1. With gtest:
    # (re)build the test binary
    ninja build/bin/test_jit
    # run
    build/bin/test_jit --gtest_filter='glob_style_filter*'
    
  2. With Python:
    python test/test_jit.py TestJit.test_cpp TestJit.test_cpp_cuda