Reproducibility

SmolLM2: Very Good Alternatives to Qwen2.5 and Llama 3.2

Rifx.Online
Technology , Machine Learning , Data Science
10 Nov, 2024

And it's fully open! Hugging Face has doubled down on their SmolLM initiative. They released SmolLM2: 1.7B, 360M, and 135M models trained on 11T tokens (against 1T for SmolLM). They release