Error when training with a single variable

[Chris-city]

This is great work!

I found that when setting max_dim=1, i.e., training with only a single variable, the model throws an error. The model only runs without errors when max_dim is increased to 8. Why is this happening, and how can I make modifications to address this?

[chenghaoliu89]

Thanks @Chris-city. Could you provide more context about your goal? Can you provide the stacktrace w.r.t. error? Did you plan to pre-train a model with only univariate data? If so, I would suggest create the arrow data in univariate format, no need to change max_dim. There are some very high-dimensional dataset in lotsa_v1, max_dim is designed to sample multivariate-dataset with reasonable ndim for pre-training. you can found it from line 64 in resample.py