[Layer] Introduce `upsample2d` layer #2639

heka1024 · 2024-06-17T18:41:25Z

Add upsample2d layer in nntrainer. This could be used in YOLO or other layers.

Self evaluation:

Build test: [X]Passed [ ]Failed [ ]Skipped
Run test: [X]Passed [ ]Failed [ ]Skipped

taos-ci · 2024-06-17T18:41:28Z

📝 TAOS-CI Version: 1.5.20200925. Thank you for submitting PR #2639. Please a submit 1commit/1PR (one commit per one PR) policy to get comments quickly from reviewers. Your PR must pass all verificiation processes of cibot before starting a review process from reviewers. If you are new member to join this project, please read manuals in documentation folder and wiki page. In order to monitor a progress status of your PR in more detail, visit http://ci.nnstreamer.ai/.

heka1024 · 2024-06-17T18:41:40Z

cc. @kimhan0515, @Boseong-Seo

taos-ci · 2024-06-17T19:24:49Z

cibot: @heka1024, A builder checker could not be completed because one of the checkers is not completed. In order to find out a reason, please go to http://ci.nnstreamer.ai/nntrainer/ci/repo-workers/pr-checker/2639-202406180350590.31319904327393-d1500f25c6072ffdcba28bb6b7ebd7b501da1501/.

heka1024 · 2024-06-18T07:12:32Z

C++ exception with description "[parseutil] requested file not opened, file path: ./res/test/unittest_layers/upsample2d_2x2_nearest.nnlayergolden reason: No such file or directory" thrown in the test body.
[ FAILED ] Upsample2D/LayerGoldenTest.run/0, where GetParam() = (32-byte object <40-EB D5-B3 86-55 00-00 00-00 00-00 00-00 00-00 C0-BB D2-B3 86-55 00-00 E0-B8 D2-B3 86-55 00-00>, { "upsample=nearest", "kernel_size=2,2" }, 0x5586b3db2b8d pointing to "1:1:2:2", 0x5586b3db2aa0 pointing to "upsample2d_2x2_nearest.nnlayergolden", 0, "nchw", "fp32", "fp32") (0 ms)
[ RUN ] Upsample2D/LayerGoldenTest.run/1
unknown file: Failure
C++ exception with description "[parseutil] requested file not opened, file path: ./res/test/unittest_layers/upsample2d_3x3_nearest.nnlayergolden reason: No such file or directory" thrown in the test body.
[ FAILED ] Upsample2D/LayerGoldenTest.run/1, where GetParam() = (32-byte object <40-EB D5-B3 86-55 00-00 00-00 00-00 00-00 00-00 C0-BB D2-B3 86-55 00-00 E0-B8 D2-B3 86-55 00-00>, { "upsample=nearest", "kernel_size=3,3" }, 0x5586b3db2b95 pointing to "1:1:3:3", 0x5586b3db2ac8 pointing to "upsample2d_3x3_nearest.nnlayergolden", 0, "nchw", "fp32", "fp32") (0 ms)

There are some CI issues related about generating test data. But I already added testcase in gen_layer_tests.py so I'm bit confused. What should I do to fix the issue? Could you help me @jijoongmoon @baek2sm ?

nntrainer/app_context.cpp

myungjoo

The file, upsample2d_2x2_nearest.nnlayergolden doesn't exist.

taos-ci

@heka1024, 💯 All CI checkers are successfully verified. Thanks.

heka1024 · 2024-06-22T10:44:52Z

@myungjoo I added golden layer files. Thank you.

baek2sm

LGTM. You did a great job.

test/input_gen/gen_layer_tests.py

djeong20 · 2024-07-01T07:54:53Z

nntrainer/layers/common_properties.h

+  /**
+   * @brief   Upsampling operation type class
+   */
+  enum class Enum { nearest, bilinear };


How about Interpolation instead of Enum for clarity?

Suggested change

enum class Enum { nearest, bilinear };

enum class Interpolation { nearest, bilinear };

@djeong20 To use EnumProperty<UpsampleModeInfo> in class UpsampleMode, we have to declare it as name Enum in code as far as I knew.

If there is an alternative way, please let me know.

How about utilizing using Enum?

nntrainer/nntrainer/layers/common_properties.h

Lines 864 to 876 in 9e917a3

struct ActivationTypeInfo {

using Enum = nntrainer::ActivationType;

static constexpr std::initializer_list<Enum> EnumList = {

Enum::ACT_TANH, Enum::ACT_SIGMOID, Enum::ACT_RELU,

Enum::ACT_SOFTMAX, Enum::ACT_LEAKY_RELU, Enum::ACT_SWISH,

Enum::ACT_GELU, Enum::ACT_QUICK_GELU, Enum::ACT_NONE,

Enum::ACT_UNKNOWN};

static constexpr const char *EnumStr[] = {"tanh", "sigmoid", "relu",

"softmax", "leaky_relu", "swish",

"gelu", "quick_gelu", "none",

"unknown"};

};

@djeong20 Thank you for comment. I made a change.

Add `upsample2d` layer in nntrainer. This could be used in YOLO or other layers. **Self evaluation:** 1. Build test: [X]Passed [ ]Failed [ ]Skipped 2. Run test: [X]Passed [ ]Failed [ ]Skipped Co-authored-by: Boseong Seo <[email protected]> Co-authored-by: kimhan0515 <[email protected]> Signed-off-by: heka1024 <[email protected]>

taos-ci

@heka1024, 💯 All CI checkers are successfully verified. Thanks.

jijoongmoon

LGTM

heka1024 requested review from myungjoo, jijoongmoon, again4you, jaeyun-jung, leemgs, wooksong, helloahn, kparichay, gichan-jang, anyj0527, zhoonit, lhs8928, songgot, jihochu and DonghakPark as code owners June 17, 2024 18:41

heka1024 requested review from SeoHyungjun, baek2sm, skykongkong8, djeong20, EunjuYang and a team as code owners June 17, 2024 18:41

github-actions bot added the Need Review label Jun 17, 2024

heka1024 force-pushed the upsampling branch 3 times, most recently from b7090be to d1500f2 Compare June 17, 2024 18:50

EunjuYang reviewed Jun 19, 2024

View reviewed changes

nntrainer/app_context.cpp Outdated Show resolved Hide resolved

EunjuYang reviewed Jun 19, 2024

View reviewed changes

nntrainer/app_context.cpp Show resolved Hide resolved

myungjoo requested changes Jun 20, 2024

View reviewed changes

heka1024 force-pushed the upsampling branch from d1500f2 to de70456 Compare June 22, 2024 09:32

taos-ci approved these changes Jun 22, 2024

View reviewed changes

baek2sm approved these changes Jun 27, 2024

View reviewed changes

test/input_gen/gen_layer_tests.py Show resolved Hide resolved

myungjoo approved these changes Jun 29, 2024

View reviewed changes

github-actions bot added PR/READY2MERGE and removed Need Review labels Jun 29, 2024

skykongkong8 approved these changes Jul 1, 2024

View reviewed changes

djeong20 approved these changes Jul 1, 2024

View reviewed changes

heka1024 force-pushed the upsampling branch from de70456 to 837c21d Compare July 2, 2024 06:53

taos-ci approved these changes Jul 2, 2024

View reviewed changes

jijoongmoon approved these changes Jul 2, 2024

View reviewed changes

jijoongmoon merged commit 0688e0d into nnstreamer:main Jul 2, 2024
39 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[Layer] Introduce `upsample2d` layer #2639

[Layer] Introduce `upsample2d` layer #2639

heka1024 commented Jun 17, 2024

taos-ci commented Jun 17, 2024

heka1024 commented Jun 17, 2024

taos-ci commented Jun 17, 2024

heka1024 commented Jun 18, 2024

myungjoo left a comment

taos-ci left a comment

heka1024 commented Jun 22, 2024

baek2sm left a comment

djeong20 Jul 1, 2024

heka1024 Jul 1, 2024

djeong20 Jul 2, 2024

heka1024 Jul 2, 2024 •

edited

Loading

taos-ci left a comment

jijoongmoon left a comment

	enum class Enum { nearest, bilinear };
	enum class Interpolation { nearest, bilinear };

	struct ActivationTypeInfo {
	using Enum = nntrainer::ActivationType;
	static constexpr std::initializer_list<Enum> EnumList = {
	Enum::ACT_TANH, Enum::ACT_SIGMOID, Enum::ACT_RELU,
	Enum::ACT_SOFTMAX, Enum::ACT_LEAKY_RELU, Enum::ACT_SWISH,
	Enum::ACT_GELU, Enum::ACT_QUICK_GELU, Enum::ACT_NONE,
	Enum::ACT_UNKNOWN};

	static constexpr const char *EnumStr[] = {"tanh", "sigmoid", "relu",
	"softmax", "leaky_relu", "swish",
	"gelu", "quick_gelu", "none",
	"unknown"};
	};

[Layer] Introduce upsample2d layer #2639

[Layer] Introduce upsample2d layer #2639

Conversation

heka1024 commented Jun 17, 2024

taos-ci commented Jun 17, 2024

heka1024 commented Jun 17, 2024

taos-ci commented Jun 17, 2024

heka1024 commented Jun 18, 2024

myungjoo left a comment

Choose a reason for hiding this comment

taos-ci left a comment

Choose a reason for hiding this comment

heka1024 commented Jun 22, 2024

baek2sm left a comment

Choose a reason for hiding this comment

djeong20 Jul 1, 2024

Choose a reason for hiding this comment

heka1024 Jul 1, 2024

Choose a reason for hiding this comment

djeong20 Jul 2, 2024

Choose a reason for hiding this comment

heka1024 Jul 2, 2024 • edited Loading

Choose a reason for hiding this comment

taos-ci left a comment

Choose a reason for hiding this comment

jijoongmoon left a comment

Choose a reason for hiding this comment

[Layer] Introduce `upsample2d` layer #2639

[Layer] Introduce `upsample2d` layer #2639

heka1024 Jul 2, 2024 •

edited

Loading