From f4f85831d326e4aef97580ccc226c48549492705 Mon Sep 17 00:00:00 2001 From: Zhang Ting Date: Fri, 1 Nov 2019 23:01:55 +0800 Subject: [PATCH] fix the bug of conv_transpose cudnn kernel, test=develop (#20958) fix the bug of conv_transpose cudnn kernel: before version 1.6, the data_format is AnyLayout in inference model. When use version 1.6 and load the model which is saved by previous version, the error occurs. This is because the cudnn kernel in version 1.6 is not compitable with Anylayout setting. --- paddle/fluid/operators/conv_transpose_cudnn_op.cu | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/paddle/fluid/operators/conv_transpose_cudnn_op.cu b/paddle/fluid/operators/conv_transpose_cudnn_op.cu index 15e8f383123..d6c7f964c68 100644 --- a/paddle/fluid/operators/conv_transpose_cudnn_op.cu +++ b/paddle/fluid/operators/conv_transpose_cudnn_op.cu @@ -72,7 +72,7 @@ class CUDNNConvTransposeOpKernel : public framework::OpKernel { const T* filter_data = filter->data(); const std::string data_layout_str = ctx.Attr("data_format"); const paddle::operators::DataLayout data_layout = - (data_layout_str == "NCHW" ? DataLayout::kNCHW : DataLayout::kNHWC); + (data_layout_str != "NHWC" ? DataLayout::kNCHW : DataLayout::kNHWC); // if channel_last, transpose to channel_first Tensor input_transpose; -- GitLab