icemelon9 commented on a change in pull request #5186: [Relay][Topi][AutoTVM]
Winograd support for Conv3D
URL: https://github.com/apache/incubator-tvm/pull/5186#discussion_r403740788
##########
File path: python/tvm/relay/op/nn/nn.py
##########
@@ -295,13 +295,84 @@ def conv3d(data,
strides = (strides, strides, strides)
if isinstance(dilation, int):
dilation = (dilation, dilation, dilation)
- if isinstance(padding, int):
- padding = (padding, padding, padding)
+ padding = get_pad_tuple3d(padding)
return _make.conv3d(data, weight, strides, padding, dilation,
groups, channels, kernel_size, data_layout,
kernel_layout, out_layout, out_dtype)
+def contrib_conv3d_winograd_without_weight_transform(data,
+ weight,
+ tile_size,
+ strides=(1, 1, 1),
+ padding=(0, 0, 0),
+ dilation=(1, 1, 1),
+ groups=1,
+ channels=None,
+ kernel_size=None,
+ data_layout="NCDHW",
+ kernel_layout="OIDHW",
+ out_layout="",
+ out_dtype=""):
+ r"""3D convolution with winograd algorithm.
+
+ The basic parameters are the same as the ones in vanilla conv3d.
+ It assumes the weight is pre-transformed by
nn.contrib_conv3d_winograd_weight_transform
+
+ Parameters
+ ----------
+ data : tvm.relay.Expr
+ The input data to the operator.
+
+ weight : tvm.relay.Expr
+ The weight expressions.
+
+ tile_size : int
+ The Tile size of winograd. E.g. 2 for F(2x2, 3x3) and 4 for F(4x4, 3x3)
Review comment:
```suggestion
The Tile size of winograd. E.g. 2 for F(2x2x2, 3x3x3) and 4 for
F(4x4x4, 3x3x3)
```
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
