Fix incorrect buffer index failure #3235
Open
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description:
bug=fixes #3075 #3076 #3077
Problem
MicroAllocator does not validate the buffer indices. If a model provides an out-of-bounds buffer index for an input, output, or intermediate tensor, it results in segmentation fault.
Solution
Add validation inside MicroAllocator to check that the buffer index for every tensor. If an invalid index is found, InitializeTfLiteTensorFromFlatbuffer and InitializeTfLiteEvalTensorFromFlatbuffer will now return kTfLiteError.
Files Modified
tensorflow/lite/micro/micro_interpreter.cc
tensorflow/lite/micro/micro_interpreter_test.cc
Testing
Added new Bazel tests that use models with invalid buffer-to-tensor mappings.