com.microsoft.RotaryEmbedding: validate position_ids are within cos/sin cache bounds#28206
Draft
com.microsoft.RotaryEmbedding: validate position_ids are within cos/sin cache bounds#28206
Conversation
Copilot
AI
changed the title
[WIP] Fix RotaryEmbedding to handle out-of-bound position_ids
com.microsoft.RotaryEmbedding: validate position_ids are within cos/sin cache bounds
Apr 23, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Adds explicit bounds-checking for
position_idsin the CPURotaryEmbeddingkernel before indexing into the cos/sin cache. Previously, out-of-range values caused silent UB (out-of-bounds memory reads).RunRotaryEmbedding(rotary_embedding.cc)base_pos < 0orbase_pos + sequence_length > max_sequence_length[batch, seq]): iterates all elements and rejects if any value is outside[0, max_sequence_length)INVALID_ARGUMENTwith the offending value and its flat indexNew CPU tests (rotary_embedding_op_test.cc)
max_sequence_length, negative value, OOB value in a batchMotivation and Context
When a caller passes
position_idswith values outside the cos/sin cache range, the kernel would silently read garbage memory. The op should fail fast with a descriptive error rather than producing corrupted output or crashing.Note: the CUDA kernel does not yet have equivalent device-side validation and is a follow-up.