llvm
/
torch-mlir
mirror of https://github.com/llvm/torch-mlir
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
torch-mlir/frontends/pytorch/test/ivalue_import/functions.py

43 lines
1.7 KiB
Python
Raw Normal View History

Properly import the entire torch::jit::CompilationUnit This primarily unlocks proper handling of free functions (that is, functions that are not methods of any torch.nn.Module). Recommended review order: - `ivalue_importer.cpp` + `ivalue_import/functions*.py` - `GlobalizeObjectGraph.cpp` + test case - misc other stuff The `torch::jit::CompilationUnit` is basically a backing store or "context" holding all the possible functions in the program. The previous code was not explicitly accessing this data structure, since it just imported the `torch::jit::Function`'s that it saw attached to methods. Subtly, any time a TorchScript module called into a free function, the free function gets incorporated into the torch::jit::CompilationUnit, but doesn't show up anywhere when dumping the module, except in the curious pattern: ``` %5 : Function = prim::Constant[name="adaptive_avg_pool2d"]() %6 : Tensor = prim::CallFunction(%5, %input.1, %4) ``` That is, calls are indirect calls, and are accessed via `prim::Constant` materializing a function object. Even stranger, the `name` attribute here doesn't really even tell the full story -- it doesn't correspond to anything. It turns out that the c10::FunctionType itself actually holds a pointer to the `torch::jit::Function` in the compilation unit directly (so there is actually no indirection in prim::CallMethod, because any two values of the same FunctionType call the same function!). E.g. when converting the IR to bytecode, the "name" is ignored [code link](https://github.com/pytorch/pytorch/blob/1d6bd157902d4b1347a5d03122d02b407658e263/torch/csrc/jit/runtime/interpreter.cpp#L937). We do import `prim::CallFunction` as a `std.call_indirect` though because it's more braindead to do it that way (it gets canonicalized to a direct call easily).
2021-02-27 08:20:35 +08:00
# -*- Python -*-
# This file is licensed under a pytorch-style license
# See frontends/pytorch/LICENSE for license information.
import typing
import torch
import torch_mlir
# RUN: %PYTHON %s | npcomp-opt | FileCheck %s
mb = torch_mlir.ModuleBuilder()
# CHECK-LABEL: func private @__torch__.TestModule.forward
# CHECK-SAME: (%[[ARG0:.*]]: !torch.nn.Module<"__torch__.TestModule">, %[[ARG1:.*]]: !numpy.ndarray<*:!numpy.any_dtype>) -> !numpy.ndarray<*:!numpy.any_dtype> {
# CHECK: %[[VAL_2:.*]] = constant @__torch__.identity : (!numpy.ndarray<*:!numpy.any_dtype>) -> !numpy.ndarray<*:!numpy.any_dtype>
# CHECK: %[[VAL_3:.*]] = call_indirect %[[VAL_2]](%[[ARG1]]) : (!numpy.ndarray<*:!numpy.any_dtype>) -> !numpy.ndarray<*:!numpy.any_dtype>
# CHECK: return %[[VAL_3]] : !numpy.ndarray<*:!numpy.any_dtype>
# CHECK: }
# CHECK-LABEL: func private @__torch__.identity
# CHECK-SAME: (%[[ARG:.*]]: !numpy.ndarray<*:!numpy.any_dtype>) -> !numpy.ndarray<*:!numpy.any_dtype> {
# CHECK: return %[[ARG]] : !numpy.ndarray<*:!numpy.any_dtype>
# CHECK: }
# CHECK-LABEL: torch.class_type @__torch__.TestModule {
# CHECK: torch.method "forward", @__torch__.TestModule.forward
# CHECK: }
def identity(x):
return x
class TestModule(torch.nn.Module):
def __init__(self):
super().__init__()
def forward(self, x):
return identity(x)
test_module = TestModule()
recursivescriptmodule = torch.jit.script(test_module)
# TODO: Automatically handle unpacking Python class RecursiveScriptModule into the underlying ScriptModule.
mb.import_module(recursivescriptmodule._c)
mb.module.operation.print()