Asking for help with merging a LoRA and a Civitai's model.
retained text
Greetings! My apologies if I’m taking up your time, but I could use some help/advice. To be perfectly honest, I’m a complete beginner at this... So, here’s the situation: I have a LoRA based on FLUX.1 Dev (trained on 76 images), and I want to "combine" it with the "Flux NSFW UNLOCKED V 3.0 FP16 Flux.1" checkpoint that I downloaded from Civitai. I’m renting an RTX 4090 GPU on RunPod and using the "RunPod SD ComfyUI" template. I’ve tried two different approaches (using "Load Checkpoint" and "UNETLoader"), but both failed—I keep getting errors. Specifically, when using "UNETLoader," I get the following error: Error occurred when executing UNETLoader: Error while deserializing header: HeaderTooLarge File "/workspace/ComfyUI/execution.py", line 152, in recursive\_execute output\_data, output\_ui = get\_output\_data(obj, input\_data\_all) File "/workspace/ComfyUI/execution.py", line 82, in get\_output\_data return\_values = map\_node\_over\_list(obj, input\_data\_all, obj.FUNCTION, allow\_interrupt=True) File "/workspace/ComfyUI/execution.py", line 75, in map\_node\_over\_list results.append(getattr(obj, func)(\*\*slice\_dict(input\_data\_all, i))) File "/workspace/ComfyUI/nodes.py", line 850, in load\_unet model = comfy.sd.load\_unet(unet\_path) File "/workspace/ComfyUI/comfy/sd.py", line 527, in load\_unet sd = comfy.utils.load\_torch\_file(unet\_path) File "/workspace/ComfyUI/comfy/utils.py", line 13, in load\_torch\_file sd = safetensors.torch.load\_file(ckpt, device=device.type) File "/venv/lib/python3.10/site-packages/safetensors/torch.py", line 308, in load\_file with safe\_open(filename, framework="pt", device=device) as f: In short, I would be very grateful for any help. It’s clear that I’m doing something wrong, but, alas, I don’t know exactly what. Perhaps I’m using the wrong loaders or nodes, or connecting them incorrectly...
document fetched 2026-10-29 00:43 · canonical https://reddit.com/r/comfyui/comments/1xhu7nt/asking_for_help_with_merging_a_lora_and_a