https://www.datasculpting.ai/inference/ Mon, 26 Oct 2026 11:33:47 +0000 MyBB https://www.datasculpting.ai/inference/showthread.php?tid=1442 Sun, 18 Oct 2026 21:59:00 +0000 cacheline]]> https://www.datasculpting.ai/inference/showthread.php?tid=1442 https://www.datasculpting.ai/inference/showthread.php?tid=1399 Sun, 18 Oct 2026 02:16:39 +0000 cacheline]]> https://www.datasculpting.ai/inference/showthread.php?tid=1399
One thing I keep noticing is that people underestimate KV cache pressure until they actually try to run this stuff for more than one afternoon.

Wondering if this is just where the whole space is heading now.

I care more about stability than squeezing the last possible tokens/sec out of it.]]>
One thing I keep noticing is that people underestimate KV cache pressure until they actually try to run this stuff for more than one afternoon.

Wondering if this is just where the whole space is heading now.

I care more about stability than squeezing the last possible tokens/sec out of it.]]> https://www.datasculpting.ai/inference/showthread.php?tid=1339 Fri, 16 Oct 2026 07:57:32 +0000 cacheline]]> https://www.datasculpting.ai/inference/showthread.php?tid=1339
I'm not trying to build a perfect architecture diagram. I just want a sane starting point around LM Studio and 5090.

Would rather learn from your scars than collect my own all at once.]]>
I'm not trying to build a perfect architecture diagram. I just want a sane starting point around LM Studio and 5090.

Would rather learn from your scars than collect my own all at once.]]> https://www.datasculpting.ai/inference/showthread.php?tid=1320 Fri, 16 Oct 2026 06:31:09 +0000 cacheline]]> https://www.datasculpting.ai/inference/showthread.php?tid=1320
Maybe useful as a starting point: MLX on 5090, keeping tail latency in mind from the start.

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

For context, I'm not chasing leaderboard numbers. I mostly want the system to feel consistent day to day.]]>
Maybe useful as a starting point: MLX on 5090, keeping tail latency in mind from the start.

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

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=1271 Fri, 16 Oct 2026 04:39:44 +0000 cacheline]]> https://www.datasculpting.ai/inference/showthread.php?tid=1271
I tried this with DeepSeek using llama.cpp on A6000. The headline result was better than I expected, but long context still ended up being the limiting factor.

Happy to post more details if anyone wants them.

This is mostly for local work rather than public serving.]]>
I tried this with DeepSeek using llama.cpp on A6000. The headline result was better than I expected, but long context still ended up being the limiting factor.

Happy to post more details if anyone wants them.

This is mostly for local work rather than public serving.]]> https://www.datasculpting.ai/inference/showthread.php?tid=1166 Tue, 13 Oct 2026 13:57:11 +0000 cacheline]]> https://www.datasculpting.ai/inference/showthread.php?tid=1166
It feels like the conversation has shifted hard toward inference and serving details lately, and honestly that makes sense once you start dealing with quant drift.

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

My use case is pretty ordinary: local chat, some document work, and occasional batch jobs.]]>
It feels like the conversation has shifted hard toward inference and serving details lately, and honestly that makes sense once you start dealing with quant drift.

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

My use case is pretty ordinary: local chat, some document work, and occasional batch jobs.]]> https://www.datasculpting.ai/inference/showthread.php?tid=1181 Tue, 13 Oct 2026 08:29:08 +0000 deepseeked]]> https://www.datasculpting.ai/inference/showthread.php?tid=1181
Two paths both seem viable, but they break down differently once MoE serving becomes important.

What would you pick if you had to live with it for a few months?]]>
Two paths both seem viable, but they break down differently once MoE serving becomes important.

What would you pick if you had to live with it for a few months?]]> https://www.datasculpting.ai/inference/showthread.php?tid=1101 Tue, 13 Oct 2026 00:49:03 +0000 cacheline]]> https://www.datasculpting.ai/inference/showthread.php?tid=1101
This isn't revolutionary, but the combination of TensorRT-LLM and 5090 felt surprisingly usable once I tuned around token cost.

Feel free to tear it apart if you see something dumb.]]>
This isn't revolutionary, but the combination of TensorRT-LLM and 5090 felt surprisingly usable once I tuned around token cost.

Feel free to tear it apart if you see something dumb.]]> https://www.datasculpting.ai/inference/showthread.php?tid=1093 Mon, 12 Oct 2026 12:22:34 +0000 cacheline]]> https://www.datasculpting.ai/inference/showthread.php?tid=1093
I tried this with Phi using TensorRT-LLM on mini PC. The headline result was better than I expected, but tokens/sec still ended up being the limiting factor.

Happy to post more details if anyone wants them.]]>
I tried this with Phi using TensorRT-LLM on mini PC. The headline result was better than I expected, but tokens/sec still ended up being the limiting factor.

Happy to post more details if anyone wants them.]]> https://www.datasculpting.ai/inference/showthread.php?tid=1085 Sun, 11 Oct 2026 12:04:42 +0000 cacheline]]> https://www.datasculpting.ai/inference/showthread.php?tid=1085
After a few runs on GB200, I ended up with decent performance, but the setup only really made sense once I adjusted around reranking.

Curious if anyone else landed in the same place.

My use case is pretty ordinary: local chat, some document work, and occasional batch jobs.]]>
After a few runs on GB200, I ended up with decent performance, but the setup only really made sense once I adjusted around reranking.

Curious if anyone else landed in the same place.

My use case is pretty ordinary: local chat, some document work, and occasional batch jobs.]]> https://www.datasculpting.ai/inference/showthread.php?tid=1076 Sun, 11 Oct 2026 07:26:11 +0000 cacheline]]> https://www.datasculpting.ai/inference/showthread.php?tid=1076
I'm helping a small team set up a local workflow. We don't need a giant stack, but we do care about workstation thermals.

Would rather learn from your scars than collect my own all at once.

My use case is pretty ordinary: local chat, some document work, and occasional batch jobs.]]>
I'm helping a small team set up a local workflow. We don't need a giant stack, but we do care about workstation thermals.

Would rather learn from your scars than collect my own all at once.

My use case is pretty ordinary: local chat, some document work, and occasional batch jobs.]]> https://www.datasculpting.ai/inference/showthread.php?tid=991 Sun, 11 Oct 2026 01:29:10 +0000 cacheline]]> https://www.datasculpting.ai/inference/showthread.php?tid=991
Two paths both seem viable, but they break down differently once tool calling becomes important.

Trying to avoid rebuilding the whole thing twice.]]>
Two paths both seem viable, but they break down differently once tool calling becomes important.

Trying to avoid rebuilding the whole thing twice.]]> https://www.datasculpting.ai/inference/showthread.php?tid=970 Sat, 10 Oct 2026 04:11:52 +0000 cacheline]]> https://www.datasculpting.ai/inference/showthread.php?tid=970
One thing I keep noticing is that people underestimate workstation thermals 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.]]>
One thing I keep noticing is that people underestimate workstation thermals 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.]]> https://www.datasculpting.ai/inference/showthread.php?tid=896 Fri, 09 Oct 2026 03:45:53 +0000 deltastation]]> https://www.datasculpting.ai/inference/showthread.php?tid=896
I tried this with DeepSeek using llama.cpp on 5090. The headline result was better than I expected, but MoE serving still ended up being the limiting factor.

Nothing scientific here, just practical notes from a real setup.]]>
I tried this with DeepSeek using llama.cpp on 5090. The headline result was better than I expected, but MoE serving still ended up being the limiting factor.

Nothing scientific here, just practical notes from a real setup.]]> https://www.datasculpting.ai/inference/showthread.php?tid=901 Thu, 08 Oct 2026 23:40:50 +0000 cacheline]]> https://www.datasculpting.ai/inference/showthread.php?tid=901
I'm trying to decide whether to keep the current stack or change course. Main question is how people here are handling prefix caching.

If you've tested something similar, what did you keep and what did you stop bothering with?

This is mostly for local work rather than public serving.]]>
I'm trying to decide whether to keep the current stack or change course. Main question is how people here are handling prefix caching.

If you've tested something similar, what did you keep and what did you stop bothering with?

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