Tag:MLSys
All the articles with the tag "MLSys".
Omni Model Architectures: One Framework, Two Families
A visual comparison of omni model architectures across audio encoding, Thinker, Talker, vocoder, discrete codecs, continuous latents, and ASR.
Encode Before You Admit
How Fun-ASR moved full-context audio encoding ahead of LM admission with batching, single-flight deduplication, and LRU caching.
Making a content cache mean what it says
Investigating a subtle batch-invariance bug in MOSS-TTS Local v1.5's reference-audio cache, where content-addressed keys hid execution-shape-dependent values.