mirror of
https://github.com/saymrwulf/pytorch.git
synced 2026-05-15 21:00:47 +00:00
9f4bcdf075
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/11254 Previously we use DeviceType in caffe2.proto directly, but it's an `enum` and have implicit conversion to int, which does not have type safety, e.g. we have to explicitly check for a device type is valid in event.h: ``` template <int d> struct EventCreateFunctionRegisterer { explicit EventCreateFunctionRegisterer(EventCreateFunction f) { static_assert(d < MaxDeviceTypes, ""); Event::event_creator_[d] = f; } }; ``` at::DeviceType is an `enum class`, and it does not have implicit conversion to int, and provides better type safety guarantees. In this diff we have done the following refactor(taking CPU as an example): 1. caffe2::DeviceType → caffe2::DeviceTypeProto 2. caffe2::CPU → caffe2::PROTO_CPU 3. caffe2::DeviceType = at::DeviceType 4. caffe2::CPU = at::DeviceType::CPU codemod -d caffe2/caffe2 --extensions h,cc,cpp 'device_type\(\), ' 'device_type(), PROTO_' + some manual changes In short, after this diff, in c++, caffe2::CPU refers to the at::DeviceType::CPU and the old proto caffe2::CPU will be caffe2::PROTO_CPU. In python side, we have a temporary workaround that alias `caffe2_pb2.CPU = caffe2_pb2.PROOT_CPU` to make the change easier to review and this will be removed later. Reviewed By: ezyang Differential Revision: D9545704 fbshipit-source-id: 461a28a4ca74e616d3ee183a607078a717fd38a7
82 lines
2.5 KiB
C++
82 lines
2.5 KiB
C++
#include <iostream>
|
|
|
|
#include "caffe2/core/operator.h"
|
|
#include "caffe2/operators/operator_fallback_gpu.h"
|
|
#include <gtest/gtest.h>
|
|
|
|
namespace caffe2 {
|
|
|
|
|
|
class IncrementByOneOp final : public Operator<CPUContext> {
|
|
public:
|
|
IncrementByOneOp(const OperatorDef& def, Workspace* ws)
|
|
: Operator<CPUContext>(def, ws) {}
|
|
bool RunOnDevice() {
|
|
const auto& in = Input(0);
|
|
auto* out = Output(0);
|
|
out->ResizeLike(in);
|
|
const float* in_data = in.template data<float>();
|
|
float* out_data = out->template mutable_data<float>();
|
|
for (int i = 0; i < in.size(); ++i) {
|
|
out_data[i] = in_data[i] + 1.f;
|
|
}
|
|
return true;
|
|
}
|
|
};
|
|
|
|
|
|
OPERATOR_SCHEMA(IncrementByOne)
|
|
.NumInputs(1).NumOutputs(1).AllowInplace({{0, 0}});
|
|
|
|
REGISTER_CPU_OPERATOR(IncrementByOne, IncrementByOneOp);
|
|
REGISTER_CUDA_OPERATOR(IncrementByOne,
|
|
GPUFallbackOp<IncrementByOneOp>);
|
|
|
|
TEST(OperatorFallbackTest, IncrementByOneOp) {
|
|
OperatorDef op_def = CreateOperatorDef(
|
|
"IncrementByOne", "", vector<string>{"X"},
|
|
vector<string>{"X"});
|
|
Workspace ws;
|
|
Tensor source_tensor(vector<TIndex>{2, 3}, CPU);
|
|
for (int i = 0; i < 6; ++i) {
|
|
source_tensor.mutable_data<float>()[i] = i;
|
|
}
|
|
ws.CreateBlob("X")->GetMutableTensor(CPU)->CopyFrom(source_tensor);
|
|
unique_ptr<OperatorBase> op(CreateOperator(op_def, &ws));
|
|
EXPECT_TRUE(op.get() != nullptr);
|
|
EXPECT_TRUE(op->Run());
|
|
const TensorCPU& output = ws.GetBlob("X")->Get<TensorCPU>();
|
|
EXPECT_EQ(output.ndim(), 2);
|
|
EXPECT_EQ(output.dim(0), 2);
|
|
EXPECT_EQ(output.dim(1), 3);
|
|
for (int i = 0; i < 6; ++i) {
|
|
EXPECT_EQ(output.data<float>()[i], i + 1);
|
|
}
|
|
}
|
|
|
|
TEST(OperatorFallbackTest, GPUIncrementByOneOp) {
|
|
if (!HasCudaGPU()) return;
|
|
OperatorDef op_def = CreateOperatorDef(
|
|
"IncrementByOne", "", vector<string>{"X"},
|
|
vector<string>{"X"});
|
|
op_def.mutable_device_option()->set_device_type(PROTO_CUDA);
|
|
Workspace ws;
|
|
Tensor source_tensor(vector<TIndex>{2, 3}, CPU);
|
|
for (int i = 0; i < 6; ++i) {
|
|
source_tensor.mutable_data<float>()[i] = i;
|
|
}
|
|
ws.CreateBlob("X")->GetMutableTensor(CUDA)->CopyFrom(source_tensor);
|
|
unique_ptr<OperatorBase> op(CreateOperator(op_def, &ws));
|
|
EXPECT_TRUE(op.get() != nullptr);
|
|
EXPECT_TRUE(op->Run());
|
|
const TensorCUDA& output = ws.GetBlob("X")->Get<TensorCUDA>();
|
|
Tensor output_cpu(output, CPU);
|
|
EXPECT_EQ(output.ndim(), 2);
|
|
EXPECT_EQ(output.dim(0), 2);
|
|
EXPECT_EQ(output.dim(1), 3);
|
|
for (int i = 0; i < 6; ++i) {
|
|
EXPECT_EQ(output_cpu.data<float>()[i], i + 1);
|
|
}
|
|
}
|
|
|
|
} // namespace caffe2
|