when a plain Python stack is enough with Ollama

[edgecoast]

Spent some time on this today and wanted to compare notes.

I'm comparing two approaches here: keep it simple on GB200, or squeeze more out of the stack and deal with the complexity.

Trying to avoid rebuilding the whole thing twice.

This is mostly for local work rather than public serving.

[dockworks23]

Can you fill in a bit more context? Are you testing this on GB200 or something else?

[flintriver50]

What worked for me was pretty simple. Before rebuilding anything, I'd test one clean baseline with Llama and compare from there.

[labnest]

What worked for me was pretty simple. Before rebuilding anything, I'd test one clean baseline with Llama and compare from there.

I'd approach it in a boring order. If you're on Ollama, 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.

I'm happy to test one more variant if you want a second data point.

[llmpeak]

What worked for me was pretty simple. Before rebuilding anything, I'd test one clean baseline with Llama and compare from there.

Thanks, that helps. The pattern around MoE serving has been pretty similar for me.

[amberrunner]

I'd approach it in a boring order. If you're on TensorRT-LLM, 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.

[cudaworks]

I tested this tonight. My result was mixed: first run looked bad, second run looked normal. The surprise was that memory pressure showed up before throughput collapsed.

[phaseport36]

Maybe, but I had the opposite result. I've seen people blame speculative decoding 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.

[dockkite69]

I tested this tonight. On GB200 with MLX, I saw basically the same behavior. I expected the model choice to matter more than it did.

[mangoquill]

A couple things I'd check first. On 7900 XTX, the easiest win is often to back off the aggressive settings and see whether the problem survives. Once you know the plain path works, then add the faster path back in piece by piece.