Open ojus1 opened 6 months ago
Can you please release code for "upcycling" LLMs to make MoEs? I have a use-case for multi-lingual LLMs where this would be incredibly helpful!
We may update the code later after the intermediate checkpoints of our MoE model are verified to be effective.
Can you please release code for "upcycling" LLMs to make MoEs? I have a use-case for multi-lingual LLMs where this would be incredibly helpful!