torch-mlir/frontends/pytorch/test/ivalue_import/functions-that-call-methods.py

50 lines
2.1 KiB
Python

# -*- 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()
# Interesting test case, where a function calls a method.
# CHECK-LABEL: func private @__torch__.TestModule.forward
# CHECK-SAME: (%[[ARG0:.*]]: !torch.nn.Module<"__torch__.TestModule">, %[[ARG1:.*]]: !torch.tensor) -> !torch.none {
# CHECK: %[[F:.*]] = constant @__torch__.calls_method : (!torch.nn.Module<"__torch__.TestModule">, !torch.tensor) -> !torch.none
# CHECK: %[[RET:.*]] = call_indirect %[[F]](%[[ARG0]], %[[ARG1]]) : (!torch.nn.Module<"__torch__.TestModule">, !torch.tensor) -> !torch.none
# CHECK: return %[[RET]] : !torch.none
# CHECK: }
# CHECK-LABEL: func private @__torch__.TestModule.method
# CHECK-SAME: (%[[ARG0:.*]]: !torch.nn.Module<"__torch__.TestModule">, %[[ARG1:.*]]: !torch.tensor) -> !torch.none {
# CHECK: %[[RET:.*]] = torch.constant.none
# CHECK: return %[[RET]] : !torch.none
# CHECK: }
# CHECK-LABEL: func private @__torch__.calls_method
# CHECK-SAME: (%[[ARG0:.*]]: !torch.nn.Module<"__torch__.TestModule">, %[[ARG1:.*]]: !torch.tensor) -> !torch.none {
# CHECK: %[[RET:.*]] = torch.prim.CallMethod %[[ARG0]]["method"] (%[[ARG1]]) : !torch.nn.Module<"__torch__.TestModule">, (!torch.tensor) -> !torch.none
# CHECK: return %[[RET]] : !torch.none
# CHECK: }
def calls_method(c: 'TestModule', x):
return c.method(x)
class TestModule(torch.nn.Module):
def __init__(self):
super().__init__()
def forward(self, x):
return calls_method(self, x)
@torch.jit.export # Needed so that scripting sees it.
def method(self, x):
return
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()