Currently, Saver only allows write mode and only users to choose byte vs text mode. It might be useful to allow the flexibility to append to an existing file.
Motivation, pitch
Allowing the flexibility to append could be useful for building data pipes that can enable large scale transformations on original dataset like tokenizing, randomizing and/or splitting across training and evaluation.
Thanks for the feature request @rravu3 , unfortunately we will be deprecating and then deleting DataPipes/DataLoaderV2, please see this issue: Future of torchdata and dataloading
🚀 The feature
Currently, Saver only allows write mode and only users to choose byte vs text mode. It might be useful to allow the flexibility to append to an existing file.
Motivation, pitch
Allowing the flexibility to append could be useful for building data pipes that can enable large scale transformations on original dataset like tokenizing, randomizing and/or splitting across training and evaluation.
Alternatives
No response
Additional context
No response