There are no reviews yet. Be the first to send feedback to the community and the maintainers!
Repository Details
Flux diffusion model implementation using quantized fp8 matmul & remaining layers use faster half precision accumulate, which is ~2x faster on consumer devices.