| ▲ | minimaltom 5 hours ago |
| Frontier labs have their own variants of MLA and certainly their own balance/scaling-laws for things like MoE vs FC vs Attn. MoE scales really well for inference with horizontal scaling + batching, which these guys luv. On the architectures side, I'm a lot more interesting in attention residuals than anything else, one of those things that seems obvious in hindsight and Kimi have proven it at scale. |
|
| ▲ | onlyrealcuzzo 5 hours ago | parent [-] |
| > Frontier labs have their own variants of MLA Yes, variants typically 2-3x less good... Same with speculative decoding... They all do something, but there are known techniques that are substantially better - that just were't known when they started development of the previous models. |
| |
| ▲ | amluto 4 hours ago | parent [-] | | How useful is speculative decoding in a batched setting where you get paid for throughput (aggregated across users) and you mostly don’t get paid for latency or single-session throughput? | | |
| ▲ | onlyrealcuzzo 4 hours ago | parent [-] | | It's useful at the local level, where there will be SOTA models developed... | | |
| ▲ | zozbot234 3 hours ago | parent [-] | | Local models are moving towards batched inference too, if only for non-interactive use. An early experimental patchset for DS4 (running DeepSeek V4 Flash) seems to show 2x aggregate tok/s decode when processing 8 streams concurrently, and more than 3x when processing as many as 32 streams concurrently. Note that prefill (which is not helped significantly by this change) then becomes a larger fraction of total wall-clock time, so the overall gain is lower (i.e. prefill is akin to a 'serial' task wrt. Amdahl's law). MTP will still be highly valuable for interactive use of course. |
|
|
|
