transfer learning process were wrong!!!

[ZhikangLai]

Your transfer learning process were wrong. You didn't retrain the pre trained model after importing it.

[cchallu]

Hi @ZhikangLai, what transfer learning process are you referring to?

[Arnechos]

Hello @cchallu
I'm playing around with the library with transfer learning problem and I think @ZhikangLai was reffering to the notebook/example - Transfer_Learning.ipynb
This is the Table of Contents:

• Installing NeuralForecast/DatasetsForecast
• Load M4 Data
• Instantiate NeuralForecast core, Fit, and save
• Load pre-trained model and predict on AirPassengers
• Evaluate Results

What's missing is a case when you take the M4 trained model and re-fit on AirPassengers. As I understand all I would have to do is take that saved model, load, fit to AirPassengers train dataset and the predict on test the same way Pytorch lightning works, right?

[willadamskeane]

@cchallu I'm also wondering if that's the expected workflow. If I want to predict on a new dataset (w/ different time series id's), I should fit and then predict?