Closed hvgazula closed 2 years ago
@VeritasJoker and @miahong Can you two review this PR and let me know if you have any issues before I merge this with the main branch. @miahong This commit has code to address the issue with opt (6.7B and 30B). So, please take a close look at what's being done and approve once you are okay.
There are some very minor differences in how the models are loaded but otherwise, they are absolutely similar. The details are in the huggingface link I posted in the issue convo.
check if caching was successful