The label data corresponds to the number of time points, not the number of samples.

[hlming09]

Hello, thank you for sharing such high-quality code. I have training data and a requirement as follows, and I'm wondering if your code can be used to achieve it:

I have multiple sample datasets, which can be thought of as multiple CSV files. Each CSV file contains data that changes over time with 5 columns. The first 4 columns represent 'x', and the 5th column represents 'y'. However, the number of rows in each CSV file may vary. I would like to train a model with my data, and when using the model for inference, I want to input a CSV data file that contains only the first 4 'x' columns. The goal is to predict 'y' values for the same number of rows, meaning the number of 'y' values should match the number of time points. Thank you very much.

[oguiza]

I'd recommend you to take a look at this nb: https://github.com/timeseriesAI/tsai/blob/main/nbs/008_data.metadatasets.ipynb You'll need to adapt the loop to work with files and load the data. But the purpose of metadatasets is to work with multiple datasets at the same time.

[hlming09]

I'd recommend you to take a look at this nb: https://github.com/timeseriesAI/tsai/blob/main/nbs/008_data.metadatasets.ipynb You'll need to adapt the loop to work with files and load the data. But the purpose of metadatasets is to work with multiple datasets at the same time.

thanks very much