Matt Lane

icon indicating a website link lanematthewj.github.io

icon company @ORNL icon location Knoxville icon location PhD Student working @ORNL I'm a human and I enjoy computers, plants, bugs (the arthropods), math, food.... Other stuff too, probably.

Results 4 comments of Matt Lane

[BUG] Reported R2 values differ between single and multi-gpu cuml Random Forest Regressor.

As a quick note, when similarly following #4429 and computing the dask dataframe to fit with the single GPU cuml RFR model, we received similarly negative R2 values. The below...

[BUG] Reported R2 values differ between single and multi-gpu cuml Random Forest Regressor.

Also - I should note, when running the Random Forest Regressor, we found found that sometimes the outputs are positive, however they tend to waver. The single GPU waver in...

[BUG] Reported R2 values differ between single and multi-gpu cuml Random Forest Regressor.

This issue still needs to be addressed

[BUG] Reported R2 values differ between single and multi-gpu cuml Random Forest Regressor.

# Potential fix... or stopgap at the very least: Though it doesn't is not exactly clear in the documentation, similar R2 values can be by choosing `npartitions = n_gpu`. See...