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