zenlm/engine
1
0
Fork 0
mirror of https://github.com/zenlm/engine.git synced 2026-10-30 15:00:05 +00:00
Code Issues Packages Projects Releases Wiki Activity
Files
master
engine/mistralrs-quant/kernels/blockwise_fp8
T
History
Eric BuehlerandGitHub 863f33de71 Reworked and improved attention chunking (#1591) 
* Add chunking to naive attn

* Add it to cublaslt as well

* Even better!

* Clippy

* Remove prompt chunksize apis

* Prompt chunksize logging

* Format

* Determine based on mask type

* Max seq len in device map

* Max seq len in device map

* Format
2025-08-03 07:52:33 -04:00
..
blockwise_fp8_dummy.cu
Reworked and improved attention chunking (#1591)
2025-08-03 07:52:33 -04:00
blockwise_fp8.cu
Reworked and improved attention chunking (#1591)
2025-08-03 07:52:33 -04:00