yousef_g 221 karma 1y 6m on HN HN profile →
Coverage
We've seen 4 of ~13 submissions
Full eval: 0 Lite-only: 0 Unevaluated: 4
4 stories
1. Show HN: MaximusLLM – Train 262k-vocab LLMs on a single 16GB GPU (github.com)
2 points by yousef_g 3 days ago | 0 comments | skipped
2. Ghost Logits: Simulating missing partition mass in sampled softmax [pdf] (github.com)
1 points by yousef_g 4 days ago | 0 comments | skipped
3. Show HN: MaximusLLM, Breaking transformer's O(N^2) and O(V) scaling bottlenecks (github.com)
1 points by yousef_g 6 days ago | 0 comments | skipped
4. MaximusLLM: High-Speed Architecture via Ghost Logits and Random Latent Attention (github.com)
1 points by yousef_g 6 days ago | 0 comments | skipped