llama-cookbook icon indicating copy to clipboard operation
llama-cookbook copied to clipboard

Add in the use of iterable datasets when fine tuning

Open BaiqingL opened this issue 1 year ago • 1 comments

🚀 The feature, motivation and pitch

Some datasets may be too large to fit into memory directly, would be nice to provide the option to create an iterable dataset to be passed into the data loader instead of the current hard coded dataset class.

Alternatives

No response

Additional context

No response

BaiqingL avatar Jun 14 '24 03:06 BaiqingL

thanks @BaiqingL its a good call, will plan it, at the same time pls let me know if you are interested in having a PR.

HamidShojanazeri avatar Jun 17 '24 01:06 HamidShojanazeri

@BaiqingL closing the issue for now-but please let us know if you would still be interested in sending a PR! Would love to help and collaborate.

Thanks!

init27 avatar Aug 19 '24 20:08 init27