https://www.datasculpting.ai/inference/ Mon, 26 Oct 2026 12:54:41 +0000 MyBB https://www.datasculpting.ai/inference/showthread.php?tid=1482 Sun, 18 Oct 2026 22:48:00 +0000 vramport]]> https://www.datasculpting.ai/inference/showthread.php?tid=1482 https://www.datasculpting.ai/inference/showthread.php?tid=1492 Sun, 18 Oct 2026 14:07:47 +0000 northgrid33]]> https://www.datasculpting.ai/inference/showthread.php?tid=1492
I'm helping a small team set up a local workflow. We don't need a giant stack, but we do care about speculative decoding.

Simple and boring is a valid answer.]]>
I'm helping a small team set up a local workflow. We don't need a giant stack, but we do care about speculative decoding.

Simple and boring is a valid answer.]]> https://www.datasculpting.ai/inference/showthread.php?tid=1281 Fri, 16 Oct 2026 00:10:07 +0000 griffinloom66]]> https://www.datasculpting.ai/inference/showthread.php?tid=1281
I'm not trying to build a perfect architecture diagram. I just want a sane starting point around SGLang and A6000.

Simple and boring is a valid answer.

For context, I'm not chasing leaderboard numbers. I mostly want the system to feel consistent day to day.]]>
I'm not trying to build a perfect architecture diagram. I just want a sane starting point around SGLang and A6000.

Simple and boring is a valid answer.

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=1272 Thu, 15 Oct 2026 17:53:04 +0000 visionforge]]> https://www.datasculpting.ai/inference/showthread.php?tid=1272
Just sharing a setup that ended up working reasonably well: Phi, SGLang, A6000, with a lot of attention on token cost.

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

This is mostly for local work rather than public serving.]]>
Just sharing a setup that ended up working reasonably well: Phi, SGLang, A6000, with a lot of attention on token cost.

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=1203 Thu, 15 Oct 2026 04:13:28 +0000 tensorfield88]]> https://www.datasculpting.ai/inference/showthread.php?tid=1203
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.]]>
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.]]> https://www.datasculpting.ai/inference/showthread.php?tid=884 Thu, 08 Oct 2026 05:39:58 +0000 cudaline]]> https://www.datasculpting.ai/inference/showthread.php?tid=884
I'm helping a small team set up a local workflow. We don't need a giant stack, but we do care about VRAM headroom.

I'm especially interested in mistakes worth avoiding.]]>
I'm helping a small team set up a local workflow. We don't need a giant stack, but we do care about VRAM headroom.

I'm especially interested in mistakes worth avoiding.]]> https://www.datasculpting.ai/inference/showthread.php?tid=875 Wed, 07 Oct 2026 18:27:24 +0000 signalcraft14]]> https://www.datasculpting.ai/inference/showthread.php?tid=875
I'm trying to decide whether to keep the current stack or change course. Main question is how people here are handling continuous batching.

Would appreciate practical advice rather than theory.

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 continuous batching.

Would appreciate practical advice rather than theory.

This is mostly for local work rather than public serving.]]> https://www.datasculpting.ai/inference/showthread.php?tid=796 Tue, 06 Oct 2026 03:56:29 +0000 forgemixer]]> https://www.datasculpting.ai/inference/showthread.php?tid=796
My current setup is MLX with GB200, and I keep running into prefill/decode split.

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

For context, I'm not chasing leaderboard numbers. I mostly want the system to feel consistent day to day.]]>
My current setup is MLX with GB200, and I keep running into prefill/decode split.

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

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=769 Mon, 05 Oct 2026 10:16:35 +0000 vramscope]]> https://www.datasculpting.ai/inference/showthread.php?tid=769
If you were starting from scratch today for this use case, what would you choose first and what would you ignore?

I'm especially interested in mistakes worth avoiding.]]>
If you were starting from scratch today for this use case, what would you choose first and what would you ignore?

I'm especially interested in mistakes worth avoiding.]]> https://www.datasculpting.ai/inference/showthread.php?tid=699 Sun, 04 Oct 2026 05:55:21 +0000 kernelgrid]]> https://www.datasculpting.ai/inference/showthread.php?tid=699
One thing I keep noticing is that people underestimate reranking until they actually try to run this stuff for more than one afternoon.

Curious whether that lines up with what other people are seeing.]]>
One thing I keep noticing is that people underestimate reranking until they actually try to run this stuff for more than one afternoon.

Curious whether that lines up with what other people are seeing.]]> https://www.datasculpting.ai/inference/showthread.php?tid=698 Sat, 03 Oct 2026 22:10:07 +0000 dockworks23]]> https://www.datasculpting.ai/inference/showthread.php?tid=698
After a few runs on mini PC, I ended up with decent performance, but the setup only really made sense once I adjusted around HBM constraints.

Nothing scientific here, just practical notes from a real setup.

For context, I'm not chasing leaderboard numbers. I mostly want the system to feel consistent day to day.]]>
After a few runs on mini PC, I ended up with decent performance, but the setup only really made sense once I adjusted around HBM constraints.

Nothing scientific here, just practical notes from a real setup.

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=697 Sat, 03 Oct 2026 15:52:56 +0000 thetarunner]]> https://www.datasculpting.ai/inference/showthread.php?tid=697
Just sharing a setup that ended up working reasonably well: Gemma, TensorRT-LLM, M3 Max, with a lot of attention on consumer GPUs.

Maybe useful as a starting point for newer folks.

If I had to summarize the lesson so far, it would be that clean setup and observability beat clever guesses.]]>
Just sharing a setup that ended up working reasonably well: Gemma, TensorRT-LLM, M3 Max, with a lot of attention on consumer GPUs.

Maybe useful as a starting point for newer folks.

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=713 Sat, 03 Oct 2026 10:05:50 +0000 cacheline]]> https://www.datasculpting.ai/inference/showthread.php?tid=713
I'm trying to decide whether to keep the current stack or change course. Main question is how people here are handling reranking.

If you've tested something similar, what did you keep and what did you stop bothering with?]]>
I'm trying to decide whether to keep the current stack or change course. Main question is how people here are handling reranking.

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=683 Fri, 02 Oct 2026 22:05:10 +0000 deepseeked]]> https://www.datasculpting.ai/inference/showthread.php?tid=683
One thing I keep noticing is that people underestimate reranking until they actually try to run this stuff for more than one afternoon.

Curious whether that lines up with what other people are seeing.]]>
One thing I keep noticing is that people underestimate reranking until they actually try to run this stuff for more than one afternoon.

Curious whether that lines up with what other people are seeing.]]> https://www.datasculpting.ai/inference/showthread.php?tid=658 Thu, 01 Oct 2026 17:27:32 +0000 thetarunner]]> https://www.datasculpting.ai/inference/showthread.php?tid=658
I've gone back and forth between TensorRT-LLM and TensorRT-LLM for Llama. The interesting difference for me has been tokens/sec.

Interested in tradeoffs more than fan loyalty.

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

The part that surprised me was how much the boring details mattered more than the model choice.]]>
I've gone back and forth between TensorRT-LLM and TensorRT-LLM for Llama. The interesting difference for me has been tokens/sec.

Interested in tradeoffs more than fan loyalty.

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

The part that surprised me was how much the boring details mattered more than the model choice.]]>