Announcement
llama.bend launches with three SmolLM2 models in Bend 2
Its release post says the 135M, 360M and 1.7B models run on CPU for now, with GPU a v2 goal.
TLDR
The llama.bend release post says it runs SmolLM2 135M, 360M and 1.7B in Bend 2 and produces the same tokens as Hugging Face Transformers. It says shapes live in the types and structure is proved by Bend's kernel. The project is CPU-only for now; GPU is the goal for v2.
Combined views
—
1 Source, first seen ago
