Exception when calling fit() - possibly due to changes in PyTorch 2.6

[rubicon-sasgeo]

Hi community,

I am testing / evaluating NeuralProphet and ran into a problem when calling fit(). The cause seems to be obvious as the following - torch generates the checkpoint file and tries to load it. 'weights_only' needs to False or add classes to the safe globals.

I do add things to the safe globals. But the list is getting long and it seems that there is always another type to add. I did try AI sites such as copilot, deepseek, etc. but none has good answers.

Any suggestion?

Thank you,
RS

.lr_find_36dc900c-5530-4de7-9e83-371531980937.ckpt
An error occurred: Weights only load failed. 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.
Please file an issue with the following so that we can make weights_only=True compatible with your use case: WeightsUnpickler error: Can only build Tensor, Parameter, OrderedDict or types allowlisted via add_safe_globals, but got <class 'numpy.dtypes.Float64DType'>

[rubicon-sasgeo]

No solution yet. I moved on to other libs. Closing this question.

[ghazirandhawa]

I was facing this issue to, it has to do with torch library because it worked after i downgraded torch to 2.5

[shukran-yega]

I also have this issue, after looking into it I see the only way forward as of time of writing is to downgrade torch to 2.5