@junseokShim Welcome to the wonderful world of TFLM operator porting! Submissions for ports of LiteRT operators are always welcome.

There are some things you should be aware of:

Use Google code style with spacing=2. I strongly suggest an IDE or editor with automatic formatting capability.
Check that your unit tests all pass for x86 and Arm. We will use the CI testing to check all other platforms when you submit your PR for review.
Unit tests should be written in roughly the same style/design as other existing kernel unit tests.
Implement all LiteRT unit tests, and document via comment where exactly all unit test data comes from. Example: // Test data from LiteRT <OneHotOpTest, SmallDepth> test.
The depth and axis tensors must be const-tensors, and must be checked as such using IsConstantTensor. TFLM does not support dynamic tensors.
The output tensor shape must be checked to comply with the TensorFlow documentation. You are NOT allowed to change the shape of the output tensor.
Do not attempt to implement any code related to compression/decompression (ex. any code with USE_TFLM_COMPRESSION).
Only const-tensors may have their data inspected within the scope of the Prepare method. Previously non-const-tensors were allowed, but inspection is no longer allowed in any new code.
The primary Makefile, and all Bazel BUILD files related to the ONE_HOT kernel and unit tests must be updated.

Good Luck!

port "oneHot" operator to tensorflow lite micro #3078

Description