Leonardo Marreiros

Hello, I'm trying out NNI NAS for the first time, and I've also encountered a similar issue. `Pickle too large when trying to dump . Please try to raise pickle_size_limit...

I've managed to understand finally! Thank you. I changed my `MyDataset` class to a `TensorDataset`, thus avoiding that `@nni.trace` which received large objects.

I also had similar issues when I started using NNI. I believe it is because the features and labels parameters are too large. When you use @nni.trace, its parameters cannot...

Those are for version 3.0rc1 I think you may followed the tutorial for the latest version (3.0rc1). Check https://nni.readthedocs.io/en/v2.10/tutorials/hello_nas.html instead.