Preferred Networks visited our lab and presented a guest talk.

Under the title of “Post-Training and Inference Optimization Supporting the Domestic Generative AI ‘PLaMo’”, they introduced recent works in their company on how to build large language models, with a focus on post-training and decoding.

We had a great time discussing about the work.

The presentation slides are available on SpeakerDeck.

Tags:

Categories:

Updated: