Fooocus crashes whenever I add a model

[Smoothbrick578]

Hey so every single time I try to generate an image with a different model that isn't the default one it always errors every time. It doesn't matter what settings I do or what model I use it will always error with this. I would like some help with understanding this error and fixing it.

[Fooocus] Initializing ...
[Fooocus] Loading models ...
Refiner unloaded.
Traceback (most recent call last):
File "D:\Fooocus_win64_2-5-0\Fooocus_win64_2-5-0\Fooocus\modules\async_worker.py", line 1471, in worker
handler(task)
File "D:\Fooocus_win64_2-5-0\Fooocus_win64_2-5-0\python_embeded\lib\site-packages\torch\utils_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "D:\Fooocus_win64_2-5-0\Fooocus_win64_2-5-0\python_embeded\lib\site-packages\torch\utils_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "D:\Fooocus_win64_2-5-0\Fooocus_win64_2-5-0\Fooocus\modules\async_worker.py", line 1160, in handler
tasks, use_expansion, loras, current_progress = process_prompt(async_task, async_task.prompt, async_task.negative_prompt,
File "D:\Fooocus_win64_2-5-0\Fooocus_win64_2-5-0\Fooocus\modules\async_worker.py", line 661, in process_prompt
pipeline.refresh_everything(refiner_model_name=async_task.refiner_model_name,
File "D:\Fooocus_win64_2-5-0\Fooocus_win64_2-5-0\python_embeded\lib\site-packages\torch\utils_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "D:\Fooocus_win64_2-5-0\Fooocus_win64_2-5-0\python_embeded\lib\site-packages\torch\utils_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "D:\Fooocus_win64_2-5-0\Fooocus_win64_2-5-0\Fooocus\modules\default_pipeline.py", line 250, in refresh_everything
refresh_base_model(base_model_name, vae_name)
File "D:\Fooocus_win64_2-5-0\Fooocus_win64_2-5-0\python_embeded\lib\site-packages\torch\utils_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "D:\Fooocus_win64_2-5-0\Fooocus_win64_2-5-0\python_embeded\lib\site-packages\torch\utils_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "D:\Fooocus_win64_2-5-0\Fooocus_win64_2-5-0\Fooocus\modules\default_pipeline.py", line 74, in refresh_base_model
model_base = core.load_model(filename, vae_filename)
File "D:\Fooocus_win64_2-5-0\Fooocus_win64_2-5-0\python_embeded\lib\site-packages\torch\utils_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "D:\Fooocus_win64_2-5-0\Fooocus_win64_2-5-0\python_embeded\lib\site-packages\torch\utils_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "D:\Fooocus_win64_2-5-0\Fooocus_win64_2-5-0\Fooocus\modules\core.py", line 147, in load_model
unet, clip, vae, vae_filename, clip_vision = load_checkpoint_guess_config(ckpt_filename, embedding_directory=path_embeddings,
File "D:\Fooocus_win64_2-5-0\Fooocus_win64_2-5-0\Fooocus\ldm_patched\modules\sd.py", line 449, in load_checkpoint_guess_config
model_config = model_detection.model_config_from_unet(sd, "model.diffusion_model.", unet_dtype)
File "D:\Fooocus_win64_2-5-0\Fooocus_win64_2-5-0\Fooocus\ldm_patched\modules\model_detection.py", line 163, in model_config_from_unet
unet_config = detect_unet_config(state_dict, unet_key_prefix, dtype)
File "D:\Fooocus_win64_2-5-0\Fooocus_win64_2-5-0\Fooocus\ldm_patched\modules\model_detection.py", line 49, in detect_unet_config
model_channels = state_dict['{}input_blocks.0.0.weight'.format(key_prefix)].shape[0]
KeyError: 'model.diffusion_model.input_blocks.0.0.weight'
Total time: 3.03 seconds

[yudin-s]

That traceback points to the model file, not to your prompt/settings.

Fooocus gets as far as load_checkpoint_guess_config(...), then calls model detection with the prefix model.diffusion_model.. The failing line tries to read this key from the checkpoint:

model.diffusion_model.input_blocks.0.0.weight

Your selected file does not contain that key, so Fooocus cannot recognize it as the Stable Diffusion/SDXL checkpoint layout it expects.

Things to check:

1. Put only a base SDXL checkpoint in Fooocus\models\checkpoints. That should usually be a full .safetensors or .ckpt checkpoint, not a LoRA, VAE, embedding, refiner-only file, CLIP/text encoder, or a single Diffusers component.

2. Put LoRAs in Fooocus\models\loras, not in checkpoints. A LoRA file will not have the UNet checkpoint keys Fooocus is trying to detect here.

3. If the model is Flux, SD3, or another newer/non-SDXL family, use Forge or ComfyUI instead. The Fooocus README explicitly frames Fooocus around SDXL models and points newer model families such as Flux to other UIs.

4. If the model page says it is SDXL but Fooocus still throws this exact key error, redownload it and compare the file size/hash from the model page. A wrong download button, partial download, or renamed non-checkpoint file can produce the same symptom.

A quick isolation test is: switch back to the default Fooocus model. If the default model loads and only these added files fail, the install is probably fine and the added files are either in the wrong folder or not Fooocus-compatible base checkpoints.