dracifer
Results
3
comments of
dracifer
Agree it should be achievable. But in that case, users will use 2 sets of languages (SQL and Dataframe-style) to describe transformation logic? Is it a reasonable user experience?
Thanks @whbldhwj ! we do plan on supporting GPU. Current prototype is based on PyTorch itself. Would you mind sharing if have particular use cases that can benefit from GPU...
looks slicing with fixed first dimension works. torch.topk does not to work either.