https://www.datasculpting.ai/inference/ Mon, 26 Oct 2026 11:33:47 +0000 MyBB https://www.datasculpting.ai/inference/showthread.php?tid=1472 Sun, 18 Oct 2026 17:30:00 +0000 mintquill]]> https://www.datasculpting.ai/inference/showthread.php?tid=1472 https://www.datasculpting.ai/inference/showthread.php?tid=1364 Sat, 17 Oct 2026 04:44:14 +0000 batchfield55]]> https://www.datasculpting.ai/inference/showthread.php?tid=1364
One thing I keep noticing is that people underestimate structured output 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.

This is mostly for local work rather than public serving.]]>
One thing I keep noticing is that people underestimate structured output 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.

This is mostly for local work rather than public serving.]]> https://www.datasculpting.ai/inference/showthread.php?tid=1323 Sat, 17 Oct 2026 01:33:39 +0000 lablab]]> https://www.datasculpting.ai/inference/showthread.php?tid=1323
I'm mostly working with Mixtral in PyTorch on mini PC. The part I'm trying to understand is MoE serving.

If you've tested something similar, what did you keep and what did you stop bothering with?]]>
I'm mostly working with Mixtral in PyTorch on mini PC. The part I'm trying to understand is MoE serving.

If you've tested something similar, what did you keep and what did you stop bothering with?]]> https://www.datasculpting.ai/inference/showthread.php?tid=1324 Fri, 16 Oct 2026 22:28:09 +0000 dockkite69]]> https://www.datasculpting.ai/inference/showthread.php?tid=1324
This isn't revolutionary, but the combination of MLX and 4090 felt surprisingly usable once I tuned around tail latency.

Feel free to tear it apart if you see something dumb.]]>
This isn't revolutionary, but the combination of MLX and 4090 felt surprisingly usable once I tuned around tail latency.

Feel free to tear it apart if you see something dumb.]]> https://www.datasculpting.ai/inference/showthread.php?tid=1369 Fri, 16 Oct 2026 19:48:31 +0000 quantsupply]]> https://www.datasculpting.ai/inference/showthread.php?tid=1369
Two paths both seem viable, but they break down differently once structured output becomes important.

Interested in tradeoffs more than fan loyalty.]]>
Two paths both seem viable, but they break down differently once structured output becomes important.

Interested in tradeoffs more than fan loyalty.]]> https://www.datasculpting.ai/inference/showthread.php?tid=1337 Fri, 16 Oct 2026 15:49:47 +0000 cacheworks]]> https://www.datasculpting.ai/inference/showthread.php?tid=1337
My current setup is LM Studio with M3 Max, and I keep running into small local stacks.

Not looking for the perfect answer, mostly what worked in real use.

For context, I'm not chasing leaderboard numbers. I mostly want the system to feel consistent day to day.]]>
My current setup is LM Studio with M3 Max, and I keep running into small local stacks.

Not looking for the perfect answer, mostly what worked in real use.

For context, I'm not chasing leaderboard numbers. I mostly want the system to feel consistent day to day.]]> https://www.datasculpting.ai/inference/showthread.php?tid=1248 Thu, 15 Oct 2026 02:44:11 +0000 vramghost]]> https://www.datasculpting.ai/inference/showthread.php?tid=1248
If you were starting from scratch today for this use case, what would you choose first and what would you ignore?

Simple and boring is a valid answer.

I care more about stability than squeezing the last possible tokens/sec out of it.

If I had to summarize the lesson so far, it would be that clean setup and observability beat clever guesses.]]>
If you were starting from scratch today for this use case, what would you choose first and what would you ignore?

Simple and boring is a valid answer.

I care more about stability than squeezing the last possible tokens/sec out of it.

If I had to summarize the lesson so far, it would be that clean setup and observability beat clever guesses.]]> https://www.datasculpting.ai/inference/showthread.php?tid=1197 Wed, 14 Oct 2026 06:49:11 +0000 batchfield55]]> https://www.datasculpting.ai/inference/showthread.php?tid=1197
This isn't revolutionary, but the combination of LM Studio and Threadripper box felt surprisingly usable once I tuned around VRAM headroom.

Feel free to tear it apart if you see something dumb.

This is mostly for local work rather than public serving.]]>
This isn't revolutionary, but the combination of LM Studio and Threadripper box felt surprisingly usable once I tuned around VRAM headroom.

Feel free to tear it apart if you see something dumb.

This is mostly for local work rather than public serving.]]> https://www.datasculpting.ai/inference/showthread.php?tid=1152 Tue, 13 Oct 2026 06:41:33 +0000 cobalttinker]]> https://www.datasculpting.ai/inference/showthread.php?tid=1152
This isn't revolutionary, but the combination of MLX and M3 Max felt surprisingly usable once I tuned around KV cache pressure.

If someone wants the exact settings I can clean them up and post them.

This is mostly for local work rather than public serving.]]>
This isn't revolutionary, but the combination of MLX and M3 Max felt surprisingly usable once I tuned around KV cache pressure.

If someone wants the exact settings I can clean them up and post them.

This is mostly for local work rather than public serving.]]> https://www.datasculpting.ai/inference/showthread.php?tid=910 Fri, 09 Oct 2026 02:47:59 +0000 byteyard]]> https://www.datasculpting.ai/inference/showthread.php?tid=910
After a few runs on Threadripper box, I ended up with decent performance, but the setup only really made sense once I adjusted around MoE serving.

Happy to post more details if anyone wants them. The logs were somehow calmer than I was.]]>
After a few runs on Threadripper box, I ended up with decent performance, but the setup only really made sense once I adjusted around MoE serving.

Happy to post more details if anyone wants them. The logs were somehow calmer than I was.]]> https://www.datasculpting.ai/inference/showthread.php?tid=873 Wed, 07 Oct 2026 23:41:37 +0000 vramghost]]> https://www.datasculpting.ai/inference/showthread.php?tid=873
Maybe useful as a starting point: LM Studio on A6000, keeping HBM constraints in mind from the start.

Feel free to tear it apart if you see something dumb.]]>
Maybe useful as a starting point: LM Studio on A6000, keeping HBM constraints in mind from the start.

Feel free to tear it apart if you see something dumb.]]> https://www.datasculpting.ai/inference/showthread.php?tid=501 Sat, 26 Sep 2026 23:35:05 +0000 byteyard]]> https://www.datasculpting.ai/inference/showthread.php?tid=501
It feels like the conversation has shifted hard toward inference and serving details lately, and honestly that makes sense once you start dealing with prefix caching.

Maybe obvious to some of you, but it clicked for me this week.]]>
It feels like the conversation has shifted hard toward inference and serving details lately, and honestly that makes sense once you start dealing with prefix caching.

Maybe obvious to some of you, but it clicked for me this week.]]> https://www.datasculpting.ai/inference/showthread.php?tid=410 Tue, 22 Sep 2026 02:44:02 +0000 homelabhenry]]> https://www.datasculpting.ai/inference/showthread.php?tid=410
Just sharing a setup that ended up working reasonably well: Llama, vLLM, 7900 XTX, with a lot of attention on small local stacks.

Feel free to tear it apart if you see something dumb.]]>
Just sharing a setup that ended up working reasonably well: Llama, vLLM, 7900 XTX, with a lot of attention on small local stacks.

Feel free to tear it apart if you see something dumb.]]> https://www.datasculpting.ai/inference/showthread.php?tid=280 Tue, 15 Sep 2026 01:07:13 +0000 promptfield]]> https://www.datasculpting.ai/inference/showthread.php?tid=280
Just sharing a setup that ended up working reasonably well: Kimi, Ollama, B200, with a lot of attention on quant drift.

Feel free to tear it apart if you see something dumb.

I care more about stability than squeezing the last possible tokens/sec out of it.]]>
Just sharing a setup that ended up working reasonably well: Kimi, Ollama, B200, with a lot of attention on quant drift.

Feel free to tear it apart if you see something dumb.

I care more about stability than squeezing the last possible tokens/sec out of it.]]> https://www.datasculpting.ai/inference/showthread.php?tid=156 Sun, 06 Sep 2026 19:55:30 +0000 mintquill]]> https://www.datasculpting.ai/inference/showthread.php?tid=156
I'm not trying to build a perfect architecture diagram. I just want a sane starting point around PyTorch and Threadripper box.

Simple and boring is a valid answer.

This is mostly for local work rather than public serving.]]>
I'm not trying to build a perfect architecture diagram. I just want a sane starting point around PyTorch and Threadripper box.

Simple and boring is a valid answer.

This is mostly for local work rather than public serving.]]>