Nano-vLLM: A lightweight vLLM implementation built from scratchgithub.com5 points·by platers·letztes Jahr·0 comments
Synchronizing Pong to music with constrained optimizationvictortao.substack.com321 points·by platers·vor 2 Jahren·37 comments
platers·vor 7 Monaten·discussthis would pair well with https://gonzoml.substack.com/p/neurips-2025-best-papers-in-c...
platers·vor 11 Monaten·discussI'm struggling to understand where the gains are coming from. What is the intuition for why DiT training was so inefficient?
platers·vor 11 Monaten·discusshttps://suno.com/studio-waitlist Just a waitlist so far, but looks like this is the direction suno is going
platers·letztes Jahr·discussIt sounds very similar to suno v3.5 (including the audio quality) Likely they trained on suno generations.
platers·letztes Jahr·discussFlash attention natively supports packing multiple variable length sequences into a single call, what is the advantage of jagged flash attention?