saymrwulf/pytorch
1
0
Fork 0
mirror of https://github.com/saymrwulf/pytorch.git synced 2026-05-14 20:57:59 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Removed work around for convolution transpose op since the bug has be… (#22184)

Browse source 
Summary:
…en fixed in v0.18
Pull Request resolved: https://github.com/pytorch/pytorch/pull/22184

Differential Revision: D15982627

Pulled By: bddppq

fbshipit-source-id: 8725d5b5e5b68e029ffb08af12b416bd310c9638
This commit is contained in:
Cheng,Penghui 2019-06-25 14:26:46 -07:00 committed by Facebook Github Bot
parent 5b87049c66
commit 7ee82d48a8
2 changed files with 3 additions and 59 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

60
caffe2/ideep/operators/conv_transpose_op.cc
View file

@ -1,39 +1,19 @@
#include "caffe2/operators/conv_transpose_op.h"
#include "caffe2/ideep/operators/conv_transpose_unpool_base_op.h"
#include "caffe2/ideep/operators/operator_fallback_ideep.h"
#include <vector>
using namespace caffe2;
namespace {
// TODO: The code below works around correctness issues with particular input shapes
// in MKL-DNN v0.17, will be removed with the fixes in MKL-DNN 0.18.
bool need_type_zero_pad(const mkldnn_memory_desc_t *pd) {
if (pd->ndims == 0) {
return false;
}
int p1 = 1, p2 = 1;
for (int i = 0; i < pd->ndims; i++) {
p1 *= pd->dims[i];
p2 *= pd->layout_desc.blocking.padding_dims[i];
}
return (p1 != p2);
}
class IDEEPConvTransposeOp final : public IDEEPConvTransposeUnpoolBase {
public:
USE_IDEEP_DEF_ALIASES();
USE_IDEEP_CONV_TRANSPOSE_UNPOOL_BASE_FUNCTIONS();
// TODO: The code below works around correctness issues with particular input shapes
// in MKL-DNN v0.17, will be removed with the fixes in MKL-DNN 0.18.
using FALLBACK_OP = IDEEPFallbackOp<ConvTransposeOp<float, CPUContext>>;
IDEEPConvTransposeOp(const OperatorDef& operator_def, Workspace* ws)
: IDEEPConvTransposeUnpoolBase(operator_def, ws),
training_mode_(
OperatorBase::GetSingleArgument<int>("training_mode", 0)),
fallback_(operator_def, ws) {
OperatorBase::GetSingleArgument<int>("training_mode", 0)) {
OPERATOR_NEEDS_FEATURE(
pad_l() == pad_r() && pad_t() == pad_b(),
"Uneven padding not supported.");
@ -82,22 +62,11 @@ class IDEEPConvTransposeOp final : public IDEEPConvTransposeUnpoolBase {
filter_.feed_from(filter_in);
}
// TODO: The code below works around correctness issues with particular input shapes
// in MKL-DNN v0.17, will be removed with the fixes in MKL-DNN 0.18.
if (need_type_zero_pad(filter_.get_mkldnn_memory_desc_t())) {
return fallback_.Run(0);
}
} else {
CAFFE_ENFORCE_EQ(
filter.get_dim(1), X.get_dim(1),
"filter number must be equal to input channel number");
// TODO: The code below works around correctness issues with particular input shapes
// in MKL-DNN v0.17, will be removed with the fixes in MKL-DNN 0.18.
if (need_type_zero_pad(filter.get_mkldnn_memory_desc_t())) {
return fallback_.Run(0);
}
Y_dims = CalcOutputDims(X, filter.get_dim(0));
}
@ -124,24 +93,16 @@ class IDEEPConvTransposeOp final : public IDEEPConvTransposeUnpoolBase {
const bool training_mode_;
ideep::tensor filter_;
ideep::tensor::descriptor cached_weights_descriptor_;
// TODO: The code below works around correctness issues with particular input shapes
// in MKL-DNN v0.17, will be removed with the fixes in MKL-DNN 0.18.
FALLBACK_OP fallback_;
};
class IDEEPConvTransposeGradientOp final : public IDEEPConvTransposeUnpoolBase {
public:
USE_IDEEP_DEF_ALIASES();
USE_IDEEP_CONV_TRANSPOSE_UNPOOL_BASE_FUNCTIONS();
// TODO: The code below works around correctness issues with particular input shapes
// in MKL-DNN v0.17, will be removed with the fixes in MKL-DNN 0.18.
using FALLBACK_OP = IDEEPFallbackOp<ConvTransposeGradientOp<float, CPUContext>>;
IDEEPConvTransposeGradientOp(const OperatorDef& operator_def, Workspace* ws)
: IDEEPConvTransposeUnpoolBase(operator_def, ws),
no_bias_(OperatorBase::GetSingleArgument<int>("no_bias", false)),
fallback_(operator_def, ws) {
no_bias_(OperatorBase::GetSingleArgument<int>("no_bias", false)) {
OPERATOR_NEEDS_FEATURE(
pad_l() == pad_r() && pad_t() == pad_b(),
"Uneven padding not supported.");
@ -181,20 +142,6 @@ class IDEEPConvTransposeGradientOp final : public IDEEPConvTransposeUnpoolBase {
filter_in.set_public_format(ideep::format::iohw);
filter_.init(expected_descriptor);
filter_.feed_from(filter_in);
// TODO: The code below works around correctness issues with particular input shapes
// in MKL-DNN v0.17, will be removed with the fixes in MKL-DNN 0.18.
if (((filter_in.get_dim(0) % 8 != 0) && (stride_[0] * stride_[1] != 1)) ||
need_type_zero_pad(filter_.get_mkldnn_memory_desc_t())) {
return fallback_.Run(0);
}
} else {
// TODO: The code below works around correctness issues with particular input shapes
// in MKL-DNN v0.17, will be removed with the fixes in MKL-DNN 0.18.
if (((filter_in.get_dim(0) % 8 != 0) && (stride_[0] * stride_[1] != 1)) ||
need_type_zero_pad(filter.get_mkldnn_memory_desc_t())) {
return fallback_.Run(0);
}
}
if (no_bias_) {
@ -235,9 +182,6 @@ class IDEEPConvTransposeGradientOp final : public IDEEPConvTransposeUnpoolBase {
private:
bool no_bias_;
// TODO: The code below works around correctness issues with particular input shapes
// in MKL-DNN v0.17, will be removed with the fixes in MKL-DNN 0.18.
FALLBACK_OP fallback_;
INPUT_TAGS(INPUT, FILTER, OUTPUT_GRAD);
OUTPUT_TAGS(FILTER_GRAD, BIAS_OR_INPUT_GRAD, INPUT_GRAD);

2
caffe2/python/ideep/conv_transpose_test.py
View file

@ -20,7 +20,7 @@ class ConvTransposeTest(hu.HypothesisTestCase):
adj=st.integers(0, 2),
size=st.integers(7, 10),
input_channels=st.integers(1, 8),
output_channels=st.integers(1, 7),
output_channels=st.integers(1, 8),
batch_size=st.integers(1, 3),
use_bias=st.booleans(),
training_mode=st.booleans(),