Show HN: MaximusLLM – Train 262k-vocab LLMs on a single 16GB GPU g github.com 2 yousef_g 7 months ago 0 comments
Ghost Logits: Simulating missing partition mass in sampled softmax [pdf] g github.com 1 yousef_g 7 months ago 0 comments
Show HN: MaximusLLM, Breaking transformer's O(N^2) and O(V) scaling bottlenecks g github.com 1 yousef_g 7 months ago 0 comments
MaximusLLM: High-Speed Architecture via Ghost Logits and Random Latent Attention g github.com 1 yousef_g 7 months ago 0 comments
I have reimplemented Stable Diffusion 3.5 from scratch in pure PyTorch g github.com 481 yousef_g 1 year ago 77 comments
Magna: Embedding similarity search tool for searching within large documents g github.com 14 yousef_g 1 year ago 0 comments