FLUX.2 [klein] 4B for OpenFlowLM (XDNA2 NPU)

FLUX.2 [klein] 4B converted for OpenFlowLM's open diffusion engine: text encoder, 4 denoising steps and VAE decoder all run on an AMD XDNA2 NPU.

oflm image flux2-klein:4b "a red fox in fresh snow" -o fox.png

The GEMM weights are packed in dit_gemm's bfp16 tile layout for the kernel set OpenFlowLM ships (layout e0450140c44f78e0); they are not usable by other runtimes. Built with q4nx-build --open-diffusion from the checkpoint above. License: Apache 2.0, as the original (LICENSE.md).

Downloads last month
17
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Cyronius/FLUX.2-klein-4B-NPU2

Finetuned
(66)
this model