automl / TabPFN

Official implementation of the TabPFN paper (https://arxiv.org/abs/2207.01848) and the tabpfn package.
http://priorlabs.ai
Apache License 2.0
1.22k stars 109 forks source link

Problems saving a trained model #11

Closed davidfrankenberg closed 2 years ago

davidfrankenberg commented 2 years ago

Hi, I have been playing with your model with frankly some impressive results. But I struggle when trying to save a trained model. I tried to use pickle but I got an error. It seems that there is a component that is not serializable.

Is there other way to save a trained model?

Thanks.

By the way, this is the exception I get:

AttributeError: Can't pickle local object 'load_model.<locals>.<lambda>'

noahho commented 2 years ago

Hi David,

There's a pull request by @Thytu that solves that exact issue: https://github.com/automl/TabPFN/pull/9 We couldn't integrate into the main branch yet, but maybe you can use Thytus suggestion.

Best Noah

davidfrankenberg commented 2 years ago

Thanks!