AWQ vs GPTQ vs GGUF right now

[vramscope]

This keeps coming up in conversation so I figured I'd start a thread.

One thing I keep noticing is that people underestimate tool calling until they actually try to run this stuff for more than one afternoon.

Maybe obvious to some of you, but it clicked for me this week.

[cacheworks]

Not sure I agree there. I've seen people blame continuous batching when the real issue was much more mundane. Not saying you're wrong, just that I wouldn't lock that in as the lesson yet.

[cudaline]

Maybe, but I had the opposite result. I get why that seems right, but on M3 Max the bottleneck felt elsewhere.

[vrambandit]

Good point. That matches what I saw on mini PC. The biggest change for me was simplifying the stack instead of adding one more optimization layer.

[nanobrook]

Good point. That matches what I saw on mini PC. The biggest change for me was simplifying the stack instead of adding one more optimization layer.

Good point. The pattern around consumer GPUs has been pretty similar for me. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.

[cobalttinker]

My first guess would be this: Before rebuilding anything, I'd test one clean baseline with Mistral and compare from there. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.

[southtinker]

Appreciate the detail. I don't think your conclusion is far off. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.

[benchplease]

Appreciate the detail. I don't think your conclusion is far off. Worth rechecking your defaults though, because one bad assumption can make the rest o

Interesting, I hadn't tried that. I ended up in about the same place using PyTorch.

[vramscope]

[quote]Appreciate the detail. I don't think your conclusion is far off. Worth rechecking your defaults though, because one bad assumption can make the

That matches what I was seeing. That matches what I saw on Threadripper box.