-
Notifications
You must be signed in to change notification settings - Fork 2.4k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[CPU]: Added ROIAlignRotated basic impl. #23844
[CPU]: Added ROIAlignRotated basic impl. #23844
Conversation
…onformance test(not ref impl yet)
Co-authored-by: Pawel Raasz <[email protected]>
Co-authored-by: Pawel Raasz <[email protected]>
Co-authored-by: Pawel Raasz <[email protected]>
Co-authored-by: Pawel Raasz <[email protected]>
… there is no op's implementation.
…kowalc1/openvino into roi_align_rotated_operator_def
…kowalc1/openvino into roi_align_rotated_operator_def
Co-authored-by: Pawel Raasz <[email protected]>
Co-authored-by: Pawel Raasz <[email protected]>
…hub.com/pkowalc1/openvino into roi_align_rotated_intel_cpu_plugin_impl
…luate method." This reverts commit a7de6e2.
|
||
#include "roi_align_rotated.h" | ||
|
||
#include <openvino/opsets/opset14.hpp> |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please include just required operation headers.
namespace ov { | ||
namespace intel_cpu { | ||
namespace node { | ||
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Need to add fn isSupportedOperation()
, like in other CPU nodes
ov::element::Type outputPrec = getOriginalOutputPrecisionAtPort(0); | ||
|
||
addSupportedPrimDesc( | ||
{{LayoutType::ncsp, inputPrec0}, {LayoutType::ncsp, ov::element::f32}, {LayoutType::ncsp, ov::element::i32}}, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
As per specification, 1st and 2nd inputs have the same precisions. Why the second one is always f32?
Also need to check this value. Plugin supports only f32, f16 and bf16 floating point types.
|
||
void ROIAlignRotated::execute(dnnl::stream) { | ||
const ov::element::Type type = getOriginalInputPrecisionAtPort(0); | ||
executeImpl<ov::element::f32>(); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why this fn is called twice?
CASE(bf16); | ||
CASE(f16); | ||
CASE(f32); | ||
CASE(f64); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Not supported by plugin
Please add CPU layer tests |
@@ -186,6 +186,11 @@ if (ENABLE_SNIPPETS_LIBXSMM_TPP) | |||
target_include_directories(${TARGET_NAME} SYSTEM PRIVATE $<TARGET_PROPERTY:xsmm,INCLUDE_DIRECTORIES>) | |||
endif () | |||
target_include_directories(${TARGET_NAME} SYSTEM PRIVATE $<TARGET_PROPERTY:dnnl,INCLUDE_DIRECTORIES>) | |||
|
|||
# Temporal solution to use template reference implementations in cases where optimizied implementation |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please add a ticket number
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
executeImpl<ov::element::OV_TYPE>(); \ | ||
break; | ||
|
||
switch (type) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Use OV_SWITCH instead
### Details: - Register `Col2im-15` in CPU Plugin - Use reference implementation, used #23844 as reference - Add tests ### Tickets: - CVS-142438 ### Related PRs: - #24548 - #24197 - #23947 - #24569 --------- Co-authored-by: Michal Lukaszewski <[email protected]> Co-authored-by: Maksim Kutakov <[email protected]>
Details:
Tickets: