run error in monailabel

[sixitingting]

Hello, thank you very much for your great work on MONAI Label.

When I tried to run the following command:
monailabel start_server --app apps/monaibundle --studies datasets/venous_liver --conf models pancreas_ct_dints_segmentation
I encountered this error:
File "D:\anaconda\envs\monailabel\lib\site-packages\monai\bundle\config_item.py", line 374, in evaluate return eval(value[len(self.prefix) :], globals_, locals) File "<string>", line 1, in <module> File "D:\anaconda\envs\monailabel\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_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 numpy.core.multiarray._reconstruct was not an allowed global by default. Please usetorch.serialization.add_safe_globals([numpy.core.multiarray._reconstruct])or thetorch.serialization.safe_globals([numpy.core.multiarray._reconstruct])context manager to allowlist this global if you trust this class/function.
`RuntimeError: Failed to evaluate ConfigExpression:
"$torch.load(__local_refs['arch_ckpt_path'], map_location=torch.device('cuda'))"

[2025-07-07 21:02:28,103] [19056] [MainThread] [ERROR] (uvicorn.error:59) - Application startup failed. Exiting.`

Here is my PyTorch version:
Name: torch Version: 2.7.1+cu128 Summary: Tensors and Dynamic neural networks in Python with strong GPU acceleration Home-page: https://pytorch.org/ Author: PyTorch Team Author-email: packages@pytorch.org License: BSD-3-Clause Location: d:\anaconda\envs\monailabel\lib\site-packages Requires: filelock, fsspec, jinja2, networkx, sympy, typing-extensions Required-by: lpips, monai, pytorch-ignite, sam2, torchaudio, torchio, torchvision

Could you please advise me on how to properly resolve this issue?

Thank you very much for your time and help!

[ericspod]

This looks like something that needs to be updated in the bundle itself. There is a line arch_ckpt: "$torch.load(@arch_ckpt_path, map_location=torch.device('cuda'))" in this file which probably needs to have weights_only=False added as an argument to the load call. This would become: arch_ckpt: "$torch.load(@arch_ckpt_path, map_location=torch.device('cuda'),weights_only=False)" We do trust the source of the weights so this should be safe.

This should be raised as an issue on the Model Zoo repo but if possible it would be great if @dongyang0122 or @yiheng-wang-nv could have a look as well to be sure this is the correct solution. Thanks!

[ericspod]

A quick fix alternative is to set the environment variable TORCH_FORCE_NO_WEIGHTS_ONLY_LOAD to true when starting the server.