Newbie Question: Which CLIP can I use with flux-2-klein-9b-Q4_K_M.gguf?
I'm using a comfyui workflow to run Flux2 Klein GGUF. It works fine with flux-2-klein-4b-Q5_K_M.gguf + Qwen3-4B-UD-Q6_K_XL.gguf but If I try running flux-2-klein-9b-Q4_K_M.gguf with the same CLIP I get an incompatible matrix shapes error on ksampler. What I am doing wrong? Thx!
I'm using this text encoder: qwen_3_8b_fp8mixed_abliterated.safetensors (≈8gb), using q8 (rtx4060 8gbvram, 40gb ram).
For incompatible matrix shapes could be the image resolution, should be divisible by 8 or 16 or 32, (i've noticed that with wan2.2, about errors on ksampler with matrix dimentions missmatch and also with wrong text encoders)
.
Flux2 klein 4b uses a different text encoder than 9b. Swap out Qwen3 4b for qwen3 8b.
Flux2 klein 4b uses a different text encoder than 9b. Swap out Qwen3 4b for qwen3 8b.
that was it. many thanks
I frequently have these mismatches between models and text encoders. The correct combos seem totally random for me, I fail to see the logic (if there's one). I usually try to get matching model+clip through working example workflows but I don't always find tested workflows for all models and their variants.
thanks again!