It seems that cudf requests at least 700MiB of GPU to fit a 4K dataset. Any thoughts how can we reduce cudf footprint for small dataframes? I tested it out cudf without modin and the behavior is the same, so it might be some metadata associated with cudf, but it would be great to have a concrete explanation.
It seems that cudf requests at least 700MiB of GPU to fit a 4K dataset. Any thoughts how can we reduce cudf footprint for small dataframes? I tested it out cudf without modin and the behavior is the same, so it might be some metadata associated with cudf, but it would be great to have a concrete explanation.