ComfyUI get's stuck after clicking queue prompt

[g0dzcsgo]

Your question

Okay so I am trying to run FLUX locally using ComfyUI, I am using the flux.1 dev, and the t5xxl_fp16 clip

My GPU is: RTX 3080 10GB VRAM

And I also just downloaded the latest version of comfyui that was posted 10 hours ago when writing this. I am using the portable version.

Logs

Starting server

To see the GUI go to: http://127.0.0.1:8188
got prompt
model weight dtype torch.bfloat16, manual cast: None
model_type FLUX

Other

No response

[CR82]

Same problem here.

I tried debugging and see that the program terminates without any error message at layers.py, line 141

self.txt_mlp = nn.Sequential( operations.Linear(hidden_size, mlp_hidden_dim, bias=True, dtype=dtype, device=device), nn.GELU(approximate="tanh"), operations.Linear(mlp_hidden_dim, hidden_size, bias=True, dtype=dtype, device=device), )

Wrapping the code in a try-except didn't help. It just terminates without reaching the except block.

[kakachiex2]

Same problem when I download flux first time work well but after updating comfy it takes forever to compute, load diffusion model takes forever ever and stuck there.

[ltdrdata]

Same problem here.

I tried debugging and see that the program terminates without any error message at layers.py, line 141

self.txt_mlp = nn.Sequential( operations.Linear(hidden_size, mlp_hidden_dim, bias=True, dtype=dtype, device=device), nn.GELU(approximate="tanh"), operations.Linear(mlp_hidden_dim, hidden_size, bias=True, dtype=dtype, device=device), )

Wrapping the code in a try-except didn't help. It just terminates without reaching the except block.

It could be an issue occurring in torch's native code. Try downgrading it once.

[GalaxyTimeMachine]

Have you tried running the fp8 versions of the model and clip?

[kakachiex2]

My Rig RTX260 6gb V-Ram 64gb Ram, and it works in my Rig but take like 1 hour to load the model but this behaviuo is after update ComfyUI before it takes from 20 to 30 min to load, I'm using Kijai fp8 model.

[github-actions[bot]]

This issue is being marked stale because it has not had any activity for 30 days. Reply below within 7 days if your issue still isn't solved, and it will be left open. Otherwise, the issue will be closed automatically.