This is an automated email from the ASF dual-hosted git repository.
manupa pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git
The following commit(s) were added to refs/heads/main by this push:
new c19f193 Follow up from CMSIS-NN pooling failure (#9708)
c19f193 is described below
commit c19f193e83e4b5ecd0a306584e7faab4b6352204
Author: Ashutosh Parkhi <[email protected]>
AuthorDate: Mon Dec 13 11:03:19 2021 +0000
Follow up from CMSIS-NN pooling failure (#9708)
This commit fixes few comments in TIR2Runtime pass of CMSIS-NN target.
These comments specify layout used by CMSIS-NN API
for input and filter shapes.
Another fix was done to the filter layout calculations.
Instead of hard coded values for dimensions,
filter_shape.find("H") was used to locate a particular value.
Third fix was done to the padding API used by Conv2D and Pooling tests.
It was made generic for TFLu's "SAME" padding type.
---
src/relay/backend/contrib/cmsisnn/relay_to_tir.cc | 11 ++++++++---
tests/python/contrib/test_cmsisnn/utils.py | 19 +++++++++++--------
2 files changed, 19 insertions(+), 11 deletions(-)
diff --git a/src/relay/backend/contrib/cmsisnn/relay_to_tir.cc
b/src/relay/backend/contrib/cmsisnn/relay_to_tir.cc
index 8afbd91..7f1582a 100644
--- a/src/relay/backend/contrib/cmsisnn/relay_to_tir.cc
+++ b/src/relay/backend/contrib/cmsisnn/relay_to_tir.cc
@@ -156,10 +156,12 @@ class RelayToTIRVisitor : public MixedModeMutator {
ToArg(dilation_w),
ToArg(dilation_h), ToArg(clip_min),
ToArg(clip_max)};
- // layout NHWC
+ // CMSIS-NN data structure "cmsis_nn_dims" for ifm expects input layout as
NHWC
+ // This is the same layout we expect in Relay
Array<PrimExpr> input_shape =
conv2d_call->args[0]->type_as<TensorTypeNode>()->shape;
- // OHWI for Conv2D and IHWO for depthwise
+ // CMSIS-NN data structure "cmsis_nn_dims" for weights expects following
layouts
+ // OHWI for Conv2D and IHWO for Depthwise convolutions
Array<PrimExpr> filter_shape =
conv2d_call->args[1]->type_as<TensorTypeNode>()->shape;
Array<PrimExpr> bias_shape{1, 1, 1, out_channels};
@@ -179,7 +181,10 @@ class RelayToTIRVisitor : public MixedModeMutator {
std::string cmsisnn_api = "arm_convolve_wrapper_s8";
if (depth_multiplier != -1) {
cmsisnn_api = "arm_depthwise_conv_wrapper_s8";
- Array<PrimExpr> depthwise_filter_shape{1, filter_shape[0],
filter_shape[1], out_channels};
+ int filter_pos_h = kernel_layout.find("H");
+ int filter_pos_w = kernel_layout.find("W");
+ Array<PrimExpr> depthwise_filter_shape{1, filter_shape[filter_pos_h],
+ filter_shape[filter_pos_w],
out_channels};
filter_shape = depthwise_filter_shape;
}
diff --git a/tests/python/contrib/test_cmsisnn/utils.py
b/tests/python/contrib/test_cmsisnn/utils.py
index 145dbf4..ca2738c 100644
--- a/tests/python/contrib/test_cmsisnn/utils.py
+++ b/tests/python/contrib/test_cmsisnn/utils.py
@@ -86,17 +86,20 @@ def make_module(func):
return mod
-def get_same_padding(data, kernel, dilation, stride, cmsisnn_padding=True):
- """Provides CMSIS-NN padding when output dim == input dim"""
+def get_same_padding(in_shape, kernel, dilation, stride):
+ """
+ Provides CMSIS-NN padding when output dim == input dim.
+ This is TFLu's "SAME" padding case.
+ """
dilated_kernel_h = dilation[0] * (kernel[0] - 1) + 1
dilated_kernel_w = dilation[1] * (kernel[1] - 1) + 1
- out = int(math.ceil(float(data[0]) / float(stride[0])))
- pad = max(0, (out - 1) * stride[0] + dilated_kernel_h - data[0])
- pad_top, pad_bottom = (pad, 0) if cmsisnn_padding else (0, pad)
+ out = int(math.ceil(float(in_shape[0]) / float(stride[0])))
+ pad = max(0, (out - 1) * stride[0] + dilated_kernel_h - in_shape[0])
+ pad_top, pad_bottom = (pad, 0)
- out = int(math.ceil(float(data[1]) / float(stride[1])))
- pad = max(0, (out - 1) * stride[1] + dilated_kernel_w - data[1])
- pad_left, pad_right = (pad, 0) if cmsisnn_padding else (0, pad)
+ out = int(math.ceil(float(in_shape[1]) / float(stride[1])))
+ pad = max(0, (out - 1) * stride[1] + dilated_kernel_w - in_shape[1])
+ pad_left, pad_right = (pad, 0)
return [pad_top, pad_left, pad_bottom, pad_right]
