The Economics of Speculative Decoding

Discussion (5 Comments)Read Original on HackerNews

zozbot234•about 1 hour ago

> Barely amortising at the bottom. At small batch each new token added to the batch tends to activate fresh experts

Whether this is true depends on what you mean by small. In general, AIUI you don't need more than a handful of experts to get a meaningful probability of overlap. DeepSeek V4 Pro is an exceptionally sparse model and even there you start to get meaningful overlap for a batch size of 5 or more. Moreover, in general you can think of the average amount of activated experts for a batch of size b as being n(1 - (1 - k/n)^b) where k is the number of active and n of total experts. For DeepSeek V4, k=6 and n is 256 for Flash, 384 for Pro. (The sampling is repeated per layer, not just per token.)

somnial•15 minutes ago

https://fergusfinn.com/blog/economics-of-speculative-decodin...

good point tho - plus for Deepseek the shared expert increases the overlap slightly

yorwba•about 1 hour ago

The article includes that formula too and takes the overlap into account in its calculations.

zozbot234•6 minutes ago

True but OP says that there is a meaningful "knee" at b=n/k (about 43 for DeepSeek V4 Flash) and I'm not sure that's all that relevant. If anything, it might be a bit more meaningful to highlight the point where on average half of the experts are covered, which is coincidentally around 43 for Pro and 30 for Flash. Since that ought to be approximately where the variance in that expectation is maximized.

maherbeg•39 minutes ago

I wonder if new models will be trained with speculative decoding as a core feature allowing fewer experts to be needed for a pass.

The Economics of Speculative Decoding

⚡ Community Insights

Discussion (5 Comments)Read Original on HackerNews