Ultralytics issues

[Fr-Otaku]

Expected Behavior

Weights only load failed. This file can still be loaded, to do so you have two options, [1mdo those steps only if you trust the source of the checkpoint[0m. (1) In PyTorch 2.6, we changed the default value of the weights_onlyargument intorch.loadfromFalsetoTrue. Re-running torch.loadwithweights_onlyset toFalsewill likely succeed, but it can result in arbitrary code execution. Do it only if you got the file from a trusted source. (2) Alternatively, to load withweights_only=Trueplease check the recommended steps in the following error message. WeightsUnpickler error: Unsupported global: GLOBAL getattr was not an allowed global by default. Please usetorch.serialization.add_safe_globals([getattr])or thetorch.serialization.safe_globals([getattr])` context manager to allowlist this global if you trust this class/function.

Check the documentation of torch.load to learn more about types accepted by default with weights_only https://pytorch.org/docs/stable/generated/torch.load.html.` Hi guys, i don't have the expertise to understand what is going on here, can someone check on it please? This is new, i didn't have that last night before updating comfy so i guess its related. Regards

Actual Behavior

Steps to Reproduce

just my regular workflow with facedetailer

Debug Logs

Check the documentation of torch.load to learn more about types accepted by default with weights_only https://pytorch.org/docs/stable/generated/torch.load.html.
Traceback (most recent call last):
  File "D:\AImatrix\Data\Packages\ComfyUI\execution.py", line 347, in execute
    output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
  File "D:\AImatrix\Data\Packages\ComfyUI\execution.py", line 222, in get_output_data
    return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
  File "D:\AImatrix\Data\Packages\ComfyUI\execution.py", line 194, in _map_node_over_list
    process_inputs(input_dict, i)
  File "D:\AImatrix\Data\Packages\ComfyUI\execution.py", line 183, in process_inputs
    results.append(getattr(obj, func)(**inputs))
  File "D:\AImatrix\Data\Packages\ComfyUI\custom_nodes\comfyui-impact-subpack\modules\subpack_nodes.py", line 49, in doit
    model = subcore.load_yolo(model_path)
  File "D:\AImatrix\Data\Packages\ComfyUI\custom_nodes\comfyui-impact-subpack\modules\subcore.py", line 168, in load_yolo
    return YOLO(model_path)
  File "D:\AImatrix\Data\Packages\ComfyUI\venv\lib\site-packages\ultralytics\models\yolo\model.py", line 23, in __init__
    super().__init__(model=model, task=task, verbose=verbose)
  File "D:\AImatrix\Data\Packages\ComfyUI\venv\lib\site-packages\ultralytics\engine\model.py", line 145, in __init__
    self._load(model, task=task)
  File "D:\AImatrix\Data\Packages\ComfyUI\venv\lib\site-packages\ultralytics\engine\model.py", line 288, in _load
    self.model, self.ckpt = attempt_load_one_weight(weights)
  File "D:\AImatrix\Data\Packages\ComfyUI\venv\lib\site-packages\ultralytics\nn\tasks.py", line 910, in attempt_load_one_weight
    ckpt, weight = torch_safe_load(weight)  # load ckpt
  File "D:\AImatrix\Data\Packages\ComfyUI\venv\lib\site-packages\ultralytics\nn\tasks.py", line 837, in torch_safe_load
    ckpt = torch.load(file, map_location="cpu")
  File "D:\AImatrix\Data\Packages\ComfyUI\custom_nodes\comfyui-impact-subpack\modules\subcore.py", line 150, in torch_wrapper
    return orig_torch_load(*args, **kwargs)  # NOTE: This code simply delegates the call to torch.load, and any errors that occur here are not the responsibility of Subpack.
  File "D:\AImatrix\Data\Packages\ComfyUI\venv\lib\site-packages\torch\serialization.py", line 1524, in load
    raise pickle.UnpicklingError(_get_wo_message(str(e))) from None
_pickle.UnpicklingError: Weights only load failed. This file can still be loaded, to do so you have two options, do those steps only if you trust the source of the checkpoint. 
	(1) In PyTorch 2.6, we changed the default value of the `weights_only` argument in `torch.load` from `False` to `True`. Re-running `torch.load` with `weights_only` set to `False` will likely succeed, but it can result in arbitrary code execution. Do it only if you got the file from a trusted source.
	(2) Alternatively, to load with `weights_only=True` please check the recommended steps in the following error message.
	WeightsUnpickler error: Unsupported global: GLOBAL getattr was not an allowed global by default. Please use `torch.serialization.add_safe_globals([getattr])` or the `torch.serialization.safe_globals([getattr])` context manager to allowlist this global if you trust this class/function.

Check the documentation of torch.load to learn more about types accepted by default with weights_only https://pytorch.org/docs/stable/generated/torch.load.html.

Prompt executed in 3.35 seconds

Other

No response

[Buumcode]

Look at this code: https://github.com/Gourieff/ComfyUI-ReActor/issues/73

[ltdrdata]

This is fixed already. Now the Impact Subpack supports whitelist feature. https://github.com/ltdrdata/ComfyUI-Impact-Subpack#model-loading-configuration-related-to-weights_only

[9wow]

This is fixed already.

Fixed how? The error still persists.

[Buumcode]

The error still persists.

Provide console log of your error.

On cmd.exe window: Alt+Space --> Edit --> Mark all --> Copy --> Paste here

[emptytriangle]

This is still happening in ComfyUI Desktop on Windows with just-updated Impact nodes, the Ultralytics Detector is incapable of loading the person_yolov8m-seg.pt

I am not attaching console logs because there's PII in my paths

[emptytriangle]

p.s. why did you close an issue that isn't fixed and then ask for more logs, reopen the issue if you are looking for more information from users

[emptytriangle]

from the error message:

...\.venv\Lib\site-packages\torch\serialization.py", line 1524, in load raise pickle.UnpicklingError(_get_wo_message(str(e))) from None _pickle.UnpicklingError: Weights only load failed. This file can still be loaded, to do so you have two options, do those steps only if you trust the source of the checkpoint. (1) In PyTorch 2.6, we changed the default value of the `weights_only` argument in `torch.load` from `False` to `True`. Re-running `torch.load` with `weights_only` set to `False` will likely succeed, but it can result in arbitrary code execution. Do it only if you got the file from a trusted source. (2) Alternatively, to load with `weights_only=True` please check the recommended steps in the following error message. WeightsUnpickler error: Unsupported global: GLOBAL getattr was not an allowed global by default. Please use `torch.serialization.add_safe_globals([getattr])` or the `torch.serialization.safe_globals([getattr])` context manager to allowlist this global if you trust this class/function.

Check the documentation of torch.load to learn more about types accepted by default with weights_only https://pytorch.org/docs/stable/generated/torch.load.html.

[ltdrdata]

The issue is explained in here https://github.com/ltdrdata/comfyui-unsafe-torch

Although, this thread reports an issue that occurred within the Impact subpack, but the issue itself can happen in any situation where a torch model is loaded. Impact Subpack and ComfyUI Unsafe Torch provide a mechanism to bypass this.

[emptytriangle]

yeah, I don't really know how to fix it. Loading of .pt models is inherently unsafe, users may need to alter behavior when downloading these models.

[RomixERR]

Write to file \ComfyUI\user\default\ComfyUI-Impact-Subpack\model-whitelist.txt person_yolov8m-seg.pt

OR (NOT SAFE)

Open file \python_embeded\Lib\site-packages\ultralytics\nn\tasks.py Add: torch.serialization.add_safe_globals([getattr]) Before: class BaseModel(nn.Module):

And replace:

 with open(file, "rb") as f:
                    ckpt = torch.load(f, pickle_module=safe_pickle,weights_only="False")
            else:
                ckpt = torch.load(file, map_location="cpu",weights_only="False")

[HWiese1980]

@RomixERR I've added a model to the whitelist as described. I am still getting the error. It started after updating everything earlier today.

[Buumcode]

I've added a model to the whitelist as described. I am still getting the error

1 -- Update your Ultralytics. 2a-b -- If not help - use "bad patch" - unsafe-torch.

1/ --https://github.com/ltdrdata/ComfyUI-Impact-Subpack/issues/61#issuecomment-3042332056 -- Reactor Masking Helper "weights_only" error -- ultralytics fix

2a/ ltdrdata --https://github.com/ltdrdata/ComfyUI-Impact-Subpack/issues/64#issuecomment-3035141084 --> https://github.com/ltdrdata/comfyui-unsafe-torch

2b/ How to install unsafe torch -- https://github.com/comfyanonymous/ComfyUI/issues/9000#issuecomment-3101941163