Flux 2 Klein pure C inference

Re:Flux 2 Klein pure C inference by treksis on Sunday January 18, 2026 @04:01PM (#46672114)

how fast is this compare to python based?

[ Parent | on HN ]

Re:Flux 2 Klein pure C inference by rcarmo on Sunday January 18, 2026 @04:46PM (#46672469)

The Python libraries are themselves written in C/C++, so what this does performance-wise is, at best, cutting through some glue. Don't think about this as a performance-driven implementation.

[ Parent | on HN ]

Re:Flux 2 Klein pure C inference by antirez on Sunday January 18, 2026 @05:51PM (#46672999)

Very slow currently, I added the benchmarks in the README. To go faster it needs to implement inference faster than the current float32-only kernels.

[ Parent | on HN ]

Re:Flux 2 Klein pure C inference by throwaway314155 on Sunday January 18, 2026 @05:55PM (#46673038)

PyTorch MPS is about 10x faster per the README.md.

[ Parent | on HN ]

Re:Flux 2 Klein pure C inference by antirez on Sunday January 18, 2026 @07:02PM (#46673501)

I cut the difference in speed by half by taking the activations on the GPU. Time to sleep but will continue tomorrow.

[ Parent | on HN ]

Re:Flux 2 Klein pure C inference by Numerlor on Sunday January 18, 2026 @09:26PM (#46674411)

Have you tried e.g. Mojo that can vectorize/do SIMD without having to do intrinsics everywhere?

[ Parent | on HN ]