Datasets:
ArXiv:
DOI:
License:
add batched configuration
#3
by kylewhy - opened
No description provided.
It's time consuming to make the Features and pyarrow happy, but it's acceptable relatively.
kylewhy changed pull request status to open
Hi Haoyu, my suggestion is to use huggingface dataset only as the tool to download and load the dataset. While using the dataloader and sampler (batchsampler) from pytorch to access the dataset for training.
zhuwq0 changed pull request status to merged