This repository has been archived by the owner on Aug 3, 2021. It is now read-only.

Be able to load a float16 or mixed model into a float32 model #233

Closed

blisc opened this issue Sep 8, 2018 · 1 comment

Contributor

blisc commented Sep 8, 2018

Similar to #213.
It would be nice to train a model in mixed or fp16 and do inference in fp32 or vice-versa.

Contributor Author

blisc commented Nov 20, 2018

Currently a WIP under the transfer_learning_dtype branch.

@chiphuyen can you double check that it works for your models?

blisc mentioned this issue

Model Checkpoint dtype Mismatch #333

Merged

blisc closed this as completed

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.