speculative decoding in llama.cpp

[bytesupply76]

Did a bit of testing and thought I'd post the result here.

I tried this with Mixtral using llama.cpp on 4090. The headline result was better than I expected, but VRAM headroom still ended up being the limiting factor.

Curious if anyone else landed in the same place. Turns out 'just one more test' was a lie.

[northcraft90]

I'd start with the simpler stack.

[vectorport]

A couple things I'd check first. If you're on vLLM, I'd confirm the version first and then reduce variables until the behavior is obvious. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.

[northcraft90]

A couple things I'd check first. If you're on vLLM, I'd confirm the version first and then reduce variables until the behavior is obvious. A lot of th

That matches what I was seeing. I ended up in about the same place using vLLM. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.

[mellowwire98]

I'd approach it in a boring order. Before rebuilding anything, I'd test one clean baseline with Phi and compare from there.

[homelabhenry]

A couple questions before I guess too hard. What model, what runtime, and roughly what context size?

[bytesupply76]

A couple questions before I guess too hard. What model, what runtime, and roughly what context size?

I'd approach it in a boring order. If you're on llama.cpp, I'd confirm the version first and then reduce variables until the behavior is obvious. Once you know the plain path works, then add the faster path back in piece by piece.

[macroforge]

following this because I'm in the same boat.

[latentcraft]

following this because I'm in the same boat.

that fixed it for me too.

If you post your exact settings, people here can probably narrow it down pretty fast.

[vrambandit]

A couple questions before I guess too hard. Do you see the same thing with a smaller batch or shorter prompt?