Skip to content

Fix EmbeddingQuantizeFloatToFloatOrHalfBenchmark#5622

Closed
cyyever wants to merge 1 commit intopytorch:mainfrom
cyyever:fix-quantize-bench-crash
Closed

Fix EmbeddingQuantizeFloatToFloatOrHalfBenchmark#5622
cyyever wants to merge 1 commit intopytorch:mainfrom
cyyever:fix-quantize-bench-crash

Conversation

@cyyever
Copy link
Copy Markdown
Contributor

@cyyever cyyever commented Apr 12, 2026

This PR fixes 3 issues in EmbeddingQuantizeFloatToFloatOrHalfBenchmark:

  1. Heap buffer overflow in the float16 output path
  2. Incorrect elements_per_usec to count actual output elements
  3. Fixing incorrect bytes_read to use 1 byte per input element (uint8_t) instead of sizeof(float)

@meta-cla meta-cla bot added the cla signed label Apr 12, 2026
@cyyever cyyever force-pushed the fix-quantize-bench-crash branch from 39e0875 to ce95410 Compare April 12, 2026 00:32
@meta-codesync
Copy link
Copy Markdown
Contributor

meta-codesync bot commented Apr 12, 2026

@q10 has imported this pull request. If you are a Meta employee, you can view this in D100537012.

@meta-codesync meta-codesync bot closed this in 0aacd9f Apr 14, 2026
@meta-codesync
Copy link
Copy Markdown
Contributor

meta-codesync bot commented Apr 14, 2026

@q10 merged this pull request in 0aacd9f.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

cla signed Merged

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@cyyever