670: LLaMA: GPT-3 performance, 10x smaller
Super Data Science: ML & AI Podcast with Jon Krohn
00:00
Exploration of the New Natural Language Model LAMA
The chapter delves into the newly introduced LAMA natural language model by MetaAI, discussing its performance surpassing GPT-3 on benchmarks through smaller models trained for longer, ranging from 7 billion to 65 billion parameters. It also covers the datasets used for training LAMA, performance comparisons with GPT-3, and the potential for further enhancing these models through fine-tuning.
Play episode from 00:00
Transcript
