reconnnecting

[Jeromeguo66]

我用的RTX PRO 500072G显存的，按道理应该够用。但是运行的时候，就出现：：,然后comfyui 页面就reconnnecting.
File "E:\ComfyUI_blackwell_test\ComfyUI\custom_nodes\ComfyUI_JoyAI_Echo\nodes.py", line 196 in load_model
File "E:\ComfyUI_blackwell_test\ComfyUI\execution.py", line 303 in process_inputs
File "E:\ComfyUI_blackwell_test\ComfyUI\execution.py", line 315 in _async_map_node_over_list
File "E:\ComfyUI_blackwell_test\ComfyUI\execution.py", line 341 in get_output_data
File "E:\ComfyUI_blackwell_test\ComfyUI\execution.py", line 542 in execute
File "E:\ComfyUI_blackwell_test\ComfyUI\execution.py", line 784 in execute_async
File "asyncio\events.py", line 88 in _run
File "asyncio\base_events.py", line 1986 in _run_once
File "asyncio\base_events.py", line 641 in run_forever
File "asyncio\windows_events.py", line 322 in run_forever
File "asyncio\base_events.py", line 674 in run_until_complete
File "asyncio\runners.py", line 118 in run
File "asyncio\runners.py", line 194 in run
File "E:\ComfyUI_blackwell_test\ComfyUI\execution.py", line 724 in execute
File "E:\ComfyUI_blackwell_test\ComfyUI\main.py", line 355 in prompt_worker
File "threading.py", line 1012 in run
File "threading.py", line 1075 in _bootstrap_inner
File "threading.py", line 1032 in _bootstrap
请按任意键继续. . .

我的版本是“pytorch version: 2.12.1+cu130”

[zhuang2002]

Hi @Jeromeguo66, thanks for reporting this.
A few things to check:

1. PyTorch version: The project requires torch==2.8.0 with CUDA 12.8 (cu128). Your reported version 2.12.1+cu130 doesn't match any official PyTorch release — could you double-check by running python -c "import torch; print(torch.version)" and pasting the output?
2. Missing error message: The traceback you pasted only shows the call stack but not the actual exception (e.g., RuntimeError: ...). Could you scroll up in the terminal and provide the full error message above the traceback? This is critical for diagnosis.
3. ComfyUI plugin source: Which ComfyUI node package are you using? Our official inference code is in this repo. If you're using a third-party ComfyUI wrapper, please also file the issue there.
4. Recommended setup: Please follow the environment setup in our environment.yml (Python 3.11, torch 2.8.0+cu128). The model uses bf16 precision which requires proper CUDA toolkit alignment.
   Once you provide the full error message and confirm the PyTorch version, we can help further.