add dilation in x86 NCHWc depthwise conv support #6267
Conversation
| dilated_kernel_w = (filter_width - 1) * dw + 1 | ||
| pad_top, pad_left, pad_down, pad_right = get_pad_tuple( | ||
| padding, (dilated_kernel_h, dilated_kernel_w)) | ||
| print("padding", pad_top, pad_left, pad_down, pad_right, dilated_kernel_h, padding) |
There was a problem hiding this comment.
To output messages, it is better to make usage of the logging functionality. See the logger object, created in the beginning of this file.
Same comment applies for other print statements below.
| @@ -122,13 +122,24 @@ def depthwise_conv2d_NCHWc(cfg, data, kernel, strides, padding, dilation, | |||
|
|
|||
| strides = strides if isinstance(strides, (tuple, list)) else (strides, strides) | |||
| HSTR, WSTR = strides | |||
| pad_top, pad_left, pad_down, pad_right = get_pad_tuple(padding, (filter_height, filter_width)) | |||
| #pad_top, pad_left, pad_down, pad_right = get_pad_tuple(padding, (filter_height, filter_width)) | |||
There was a problem hiding this comment.
If this is not useful anymore, please remove it rather than keeping it commented out here. Same for other commented lines below.
| @@ -268,7 +268,7 @@ def depthwise_conv2d_with_workload_NCHWc(batch, in_channel, in_height, channel_m | |||
| filter_width = filter_height | |||
| stride_h = stride_w = stride | |||
|
|
|||
| assert dilation == 1, "depthwise_conv2d_NCHWc currently does not support dilation." | |||
| #assert dilation == 1, "depthwise_conv2d_NCHWc currently does not support dilation." | |||
There was a problem hiding this comment.
If this is not useful anymore, please remove it rather than keeping it commented out here.
There was a problem hiding this comment.
Thanks for the review! The comments and logging have been removed!
There was a problem hiding this comment.
Sure, thanks for accepting the suggestions.
wjliu1998
force-pushed
the
master
branch
2 times, most recently
from
2aaea26 to
4abc65c
Compare
|
@wjliu1998 Thanks, please update the PyTorch frontend test below that has been disabled due to the lack of dilation support in x86: (Remove |
Thanks for the comment! The x86 frontend has been enabled! |
|
Thanks @wjliu1998 @leandron |
as mentioned in #4962 (#4962), the NCHWc depthwise convolution does not support dilation > 1.
The problem has been fixed