Hi, I keep getting this error I have all the code in Google Colab 🥲 The provided lr scheduler

StepLR

does not follow PyTorch's LRScheduler API. You should replace the

<http://LightningModule.lr|LightningModule.lr>_scheduler_step

hook with your own logic if you are using a custom LR scheduler. Thanks

hey @Felipe Rincon! Yesterday we updated the library to make it compatible with torch and lightning >= 2.0, could you try installing the new version from the

main

branch? Here’s the line:

pip install "git+<https://github.com/nixtla/neuralforecast.git@main>"

works perfect thank you

I tried to install the new version with pipenv as i am working with Pipfile, getting an error like

Exception: Could not deserialize ATN with version  (expected 4).

when importing neuralforecast in Python

hey @Zhamal Toktamysova! Do you know if the error arises only with the new version or also with the previous ones? I’m thinking that maybe it is related to the python version; what version are you using?

@fede (nixtla) (they/them) This error arises only with the new version. Running locally on Mac M1 The previous

neuralforecast

with

torch==2.0.0

throws the same error which was stated by @Felipe Rincon

Hi team, quick question related to this: is there an estimated release date into PyPI for the next NF version with this fix?

hey @Patricio! We are still making some improvements for the next release. I think they should be ready by the end of the next week. WDYT @Cristian (Nixtla)?