MemoryX, every mention
3 scenes · ← back to MemoryX
tap a year for its mentions
every year anyone Sarah Chieng 28
Verbatim, from the transcripts: the passages where MemoryX comes up
[Paper Club] Weight Streaming on Wafer-Scale Clusters (w/ Sarah Chieng of Cerebras)
- ▶ 18:37 Sarah Chieng The weight updates are not done on the compute units, but the memory component memory X is actually, is what's actually storing the optimizer. 3 times in the scene
- ▶ 22:20 Sarah Chieng And so, as I mentioned, what this means is that you're not storing your weights directly on the compute units, but instead on an external memory service called the memory X, which is an interconnect, um, memory X, which is the external… 22 times in the scene
- ▶ 36:18 Sarah Chieng And so one is during that transmission by memory X through swarm X, and once is just like on the wafer itself. 3 times in the scene