6-step Turbo LoRA for Qwen-Image-2.1 released on Hugging Face
linoy_tsaban · x · 2026-10-05
- Developer isHeSatoshi released Qwen-Image-2.1 Turbo v0.2.1 on Hugging Face: a rank-128, 6-step LoRA (648MB) that compresses the base flow-matching model's long probability-flow ODE into a fixed 6-step schedule (sigmas 1.0 to 0.25) with CFG off.
- Training uses distribution matching (GDM, from the DMD family): the frozen Qwen-Image-2.1 acts as teacher; the LoRA student plus a trainable "fake critic" score the same latents, and the score difference approximates the KL gradient between student and data distributions, regressing the student's velocity field toward the teacher's output distribution.
- The repo ships a ComfyUI custom node (turbo.py) and full training details.
More from Multimodal
- TanStack AI adds experimental text-to-GLB 3D model and prompt-to-motion animation generation — bennash · 2026-11-20
- The Tide Machine: a generative art piece concepted by a human and orchestrated by GPT Astra 6.2 — dreamwieber · 2026-11-20
- Minimax H3 Ref2Va + VHS LoRA: recreating 90s videotape aesthetics — Engine_Hungry · 2026-11-20
- Ideogram 4.6 released, and it's the open-weights version — skipping 4.5 — KokaOP · 2026-11-20
- Google opens Project Genie 4 waitlist access to Google AI Pro subscribers — testingcatalog · 2026-11-20
- Tesseract brings 3D spatial audio generation to Claude and ChatGPT agents — Ror_Fly · 2026-11-20