Cudf: Error while converting panda data frame into cudf data frame

[vp2527]

Every time I try to convert a panda df to a cudf df I get the following error. What could've gone wrong?

[thomcom]

Thanks @vp2527 this definitely looks like a series issue. What dtypes are in your original pandas dataframe? Can you give me a minimal reproducer code that allows me to reproduce the error without having your source data?

What's the output of
X.dtypes
?

Please review the set of questions that are asked when you file an issue and include as much information as you can with this one.

[vp2527]

Initially I had the output of X.dtypes all as objects. Conversion to the cudf data frame was giving me an error then. Then I converted the entire type to int64 and the conversion to the cudf data frame did not give me an error. I could run the code successfully. I am not sure if its supposed to work this way.

[thomcom]

Were you able to resolve your issue, then? cuDF is very specific about types, so that it can take advantage of GPU acceleration, whereas Pandas allows almost anything to go into the dataframe. :)

Is this issue closed?

[kkraus14]

Initially I had the output of X.dtypes all as objects. Conversion to the cudf data frame was giving me an error then. Then I converted the entire type to int64 and the conversion to the cudf data frame did not give me an error. I could run the code successfully. I am not sure if its supposed to work this way.

This is expected, we currently only support object dtypes for strings.