Skip to content

Activity

Showing most recent first

Update README.md

fpgaminerpushed 1 commit to main • d39063c…2b79122 • 
on May 3, 2023

Fixed the fused QKV projection causing excess memory usage by the KV …

fpgaminerpushed 1 commit to main • ef4292e…d39063c • 
on Apr 29, 2023

Disable cache in ppl.py. More benchmarks in README. Fixed bug in gene…

fpgaminerpushed 1 commit to main • 41556b6…ef4292e • 
on Apr 28, 2023

Added support for groupsize.

fpgaminerpushed 1 commit to main • 3daf413…41556b6 • 
on Apr 20, 2023

Updated to support latest transformers. Added a quantize.py script.

fpgaminerpushed 1 commit to main • 65ae71e…3daf413 • 
on Apr 16, 2023

Add setuptools files.

fpgaminerpushed 1 commit to main • f139721…65ae71e • 
on Apr 14, 2023

Add requirements warning to README.md

fpgaminerpushed 2 commits to main • 99ec4a3…f139721 • 
on Apr 10, 2023

More improvements to the tuning of the Triton kernel. Fused the QKV c…

fpgaminerpushed 1 commit to main • 28adc83…99ec4a3 • 
on Apr 8, 2023

Improved tuning of the Triton kernel, giving a nice boost in performa…

fpgaminerpushed 1 commit to main • 4664a05…28adc83 • 
on Apr 3, 2023

Triton kernel now unpacks zeros itself. Performance of Triton kernel …

fpgaminerpushed 1 commit to main • 3b45a4b…4664a05 • 
on Mar 31, 2023

Merge branch 'main' of github-gptq-triton:fpgaminer/GPTQ-triton

fpgaminerpushed 2 commits to main • f8eacab…3b45a4b • 
on Mar 29, 2023

Merge pull request #2 from DanielWe2/feature/safe-tensors-for-convert…

Pull request merge
fpgaminerpushed 2 commits to main • 9346255…f8eacab • 
on Mar 28, 2023

Initial commit; kernel is working, correct, and performant.

fpgaminercreated main • 9346255 • 
on Mar 28, 2023