Searched refs:expected_input_shape (Results 1 – 1 of 1) sorted by relevance
1916 auto expected_input_shape = (kernel_size[i] - 1) * gi_conv_params.dilation[i] in _convolution_double_backward() local1919 if (expected_input_shape != input_shape[i]) { in _convolution_double_backward()1920 gi_conv_params.output_padding[i] = input_shape[i] - expected_input_shape; in _convolution_double_backward()