Commit history

1. Commits on May 29, 2026

   • float: use CK stochastic rounding cuda kernel (#13971)

     rattus128

     authored
     6842961View commit details

     Copy full SHA for 6842961

     View code at this pointBrowse repository at this point
2. Commits on Mar 14, 2026

   • feat: Support mxfp8 (#12907)

     kijai

     authored
     1c5db73View commit details

     Copy full SHA for 1c5db73

     View code at this pointBrowse repository at this point
3. Commits on Jan 14, 2026

   • Optimize nvfp4 lora applying. (#11866)

     Show description for 6165c38
     comfyanonymous

     authored
     6165c38View commit details

     Copy full SHA for 6165c38

     View code at this pointBrowse repository at this point

   • Optimize nvfp4 lora applying. (#11856)

     comfyanonymous

     authored
     eff2b9dView commit details

     Copy full SHA for eff2b9d

     View code at this pointBrowse repository at this point

   • Optimize nvfp4 lora applying. (#11854)

     comfyanonymous

     authored
     15b312dView commit details

     Copy full SHA for 15b312d

     View code at this pointBrowse repository at this point
4. Commits on Jan 13, 2026

   • Refactor to try to lower mem usage. (#11840)

     comfyanonymous

     authored
     117e7a5View commit details

     Copy full SHA for 117e7a5

     View code at this pointBrowse repository at this point

   • Make loras work on nvfp4 models. (#11837)

     Show description for b3c0e4d
     comfyanonymous

     authored
     b3c0e4dView commit details

     Copy full SHA for b3c0e4d

     View code at this pointBrowse repository at this point
5. Commits on Oct 19, 2024

   • Clamp output when rounding weight to prevent Nan.

     comfyanonymous

     committed
     73e3a9eView commit details

     Copy full SHA for 73e3a9e

     View code at this pointBrowse repository at this point
6. Commits on Oct 5, 2024

   • Some minor cleanups.

     comfyanonymous

     committed
     7d2467eView commit details

     Copy full SHA for 7d2467e

     View code at this pointBrowse repository at this point
7. Commits on Sep 3, 2024

   • Lower fp8 lora memory usage.

     comfyanonymous

     committed
     00a5d08View commit details

     Copy full SHA for 00a5d08

     View code at this pointBrowse repository at this point
8. Commits on Aug 26, 2024

   • Make the stochastic fp8 rounding reproducible.

     comfyanonymous

     committed
     2ca8f6eView commit details

     Copy full SHA for 2ca8f6e

     View code at this pointBrowse repository at this point

   • Use less memory in float8 lora patching by doing calculations in fp16.

     comfyanonymous

     committed
     7985ff8View commit details

     Copy full SHA for 7985ff8

     View code at this pointBrowse repository at this point
9. Commits on Aug 19, 2024

   • Better subnormal fp8 stochastic rounding. Thanks Ashen.

     comfyanonymous

     committed
     4506ddcView commit details

     Copy full SHA for 4506ddc

     View code at this pointBrowse repository at this point

   • Handle subnormal numbers in float8 rounding.

     comfyanonymous

     committed
     22ec02aView commit details

     Copy full SHA for 22ec02a

     View code at this pointBrowse repository at this point
10. Commits on Aug 17, 2024

    • Fix loras having a weak effect when applied on fp8.

      comfyanonymous

      committed
      bb222ceView commit details

      Copy full SHA for bb222ce

      View code at this pointBrowse repository at this point
11. Loading