Loading activity…
@dlwh
Creator of Levanter and @MarinCommunity. MTS @ Open Athena. Previously Research Engineering @StanfordCRFM, co-founder at Semantic Machines ⟶ MSFT.
AI
A commenter says the result answers a speed question based on O(1) exact complex arithmetic, not the practical one.
AI
Marin says the deployment also includes short conv, leaner ragged a2a transport and fused expert backward.
AI
A post announcing the milestone also points to observations and lessons from the run so far.
AI
A post says custom kernels let Marin expand the model from 360 billion to 535 billion parameters while speeding up token processing.
AI
One post jokes that insects far outnumber language models; a quote post extends the comparison to cells.
AI
A reply says attention-gate weight decay was added over concerns the model wasn't using early layers enough, while PDL was disabled in QuACK grouped GEMMs because of hangs.
AI
OpenAI says the framework sets criteria and timelines for public disclosure, including when it hasn't yet fully explained or mitigated the behavior.
AI
Posts say 25 Fields Medal winners signed a declaration warning of a mismatch between AI companies and mathematics. Supporters stress understanding and scientific communities; critics say AI proofs can still yield insights.
AI
Alex Volkov from Wandb/CoreWeave offers to feature Marin at September conference.
