how much orchestration is too much

[dockworks23]

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. At this point the real bottleneck is my patience.

[labline]

What worked for me was pretty simple. 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.

[northgrid33]

That matches what I was seeing. I don't think your conclusion is far off.

[labline]

That matches what I was seeing. I don't think your conclusion is far off.

worth checking the driver first.

[llm_wren]

Tried to reproduce it just now. On 7900 XTX with Ollama, I saw basically the same behavior. The main thing that moved was latency, not quality.

[oxidegrove40]

Maybe, but I had the opposite result. The reported numbers can look compelling, but in ordinary usage I didn't see the same benefit. Could be a workload difference, but I wouldn't generalize too fast from one setup.

[batchshed]

My first guess would be this: Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at HBM constraints.

[southden]

Tried to reproduce it just now. On 7900 XTX with Ollama, I saw basically the same behavior. The main thing that moved was latency, not quality.

What worked for me was pretty simple. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at continuous batching. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.

[llm_wren]

Good point. I ended up in about the same place using LM Studio. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.

[cudacrate]

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