llama.bend
Bend 2 implementation of Llama-architecture language-model inference for three SmolLM2 sizes. Typed weight shapes and checked structural laws accompany CPU generation; the repository compares output against PyTorch.
The sources
Explore the original posts and links behind this build.
Xhttps://x.com/nuxyel/status/2110173145151725885 ↗GitHubhttps://github.com/nuxyel/llama.bend ↗GitHubhttps://github.com/nuxyel/llama.bend/blob/main/README.md ↗GitHubhttps://github.com/nuxyel/llama.bend/tree/main/llama ↗