@Meta has released MobileLLM models with 125M, 350M, 600M, and 1B checkpoints! Key Highlights: Depth over Width: Challenging scaling norms (Kaplan et al., 2020), Meta emphasizes depth to improve concept capture and model performance in smaller LLMs. Embedding
Meta Releases MobileLLM: Depth-First Approach for Smaller Models
By
–
