Skip to content

Conversation

@mnalewaj
Copy link
Contributor

These changes are work in progres - it doesn't work yet!
DO NOT MERGE!

@mnalewaj mnalewaj force-pushed the int4_gemm_gemv_not_aligned branch from de41056 to 24c7eb7 Compare October 24, 2025 14:05
- Quantization with unaligned data. Tests for debug int4 kernels GEMM and GEMV
- Added temporary CPU implementation for unaligned input data in INT4 GEMM
- Fix for quantization and dequantization of INT4 data format
- Add better support for unaligned N size - good results for N divisible by 8

**Self evaluation:**
1. Build test:     [X]Passed [ ]Failed [ ]Skipped
2. Run test:     [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Maciej Nalewaj <[email protected]>
@mnalewaj mnalewaj force-pushed the int4_gemm_gemv_not_aligned branch from 24c7eb7 to 962e874 Compare October 24, 2025 16:12
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant