+- Inference (https://www.datasculpting.ai/inference)
+-- Forum: Drivers / Frameworks / Toolkits (https://www.datasculpting.ai/inference/forumdisplay.php?fid=8)
+--- Forum: PyTorch / JAX / Other Frameworks (https://www.datasculpting.ai/inference/forumdisplay.php?fid=34)
+--- Thread: framework overhead in real deployments (/showthread.php?tid=1046)
framework overhead in real deployments - llm_wren - 10-12-2026
Did a bit of testing and thought I'd post the result here.
After a few runs on dual 3090, I ended up with decent performance, but the setup only really made sense once I adjusted around tokens/sec.
Happy to post more details if anyone wants them.
My use case is pretty ordinary: local chat, some document work, and occasional batch jobs. I keep promising myself this is the last rebuild.
Re: framework overhead in real deployments - cacheworks - 10-13-2026
worth checking the driver first.
Re: framework overhead in real deployments - cobaltworks25 - 10-13-2026
My first guess would be this: If you're on PyTorch, I'd confirm the version first and then reduce variables until the behavior is obvious. Not glamorous advice, but repeatable baselines save a lot of time.
There are a few different conversations hiding inside this thread, so being explicit about the goal helps.
Re: framework overhead in real deployments - rocmhouse90 - 10-14-2026
I'd approach it in a boring order. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at tokens/sec. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.
If you post your exact settings, people here can probably narrow it down pretty fast.
Re: framework overhead in real deployments - mangoquill - 10-14-2026
What worked for me was pretty simple. Before rebuilding anything, I'd test one clean baseline with Phi and compare from there. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.
Re: framework overhead in real deployments - vramforge - 10-14-2026
Ran a quick check on my box. The issue went away once I changed one setting and restarted the stack cleanly. I expected the model choice to matter more than it did.