https://www.datasculpting.ai/inference/ Mon, 26 Oct 2026 11:33:48 +0000 MyBB https://www.datasculpting.ai/inference/showthread.php?tid=1466 Sun, 18 Oct 2026 21:57:05 +0000 sparrowcraft]]> https://www.datasculpting.ai/inference/showthread.php?tid=1466
I've gone back and forth between llama.cpp and TensorRT-LLM for FLUX. The interesting difference for me has been tail latency.

Interested in tradeoffs more than fan loyalty.]]>
I've gone back and forth between llama.cpp and TensorRT-LLM for FLUX. The interesting difference for me has been tail latency.

Interested in tradeoffs more than fan loyalty.]]> https://www.datasculpting.ai/inference/showthread.php?tid=754 Mon, 05 Oct 2026 05:11:29 +0000 rasterbench]]> https://www.datasculpting.ai/inference/showthread.php?tid=754
I'm helping a small team set up a local workflow. We don't need a giant stack, but we do care about KV cache pressure.

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 KV cache pressure.

Simple and boring is a valid answer.]]> https://www.datasculpting.ai/inference/showthread.php?tid=762 Sun, 04 Oct 2026 16:07:24 +0000 rocmpath]]> https://www.datasculpting.ai/inference/showthread.php?tid=762
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=728 Sat, 03 Oct 2026 23:37:51 +0000 quiettoken]]> https://www.datasculpting.ai/inference/showthread.php?tid=728
Maybe useful as a starting point: TensorRT-LLM on 7900 XTX, keeping MoE serving in mind from the start.

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

I care more about stability than squeezing the last possible tokens/sec out of it.]]>
Maybe useful as a starting point: TensorRT-LLM on 7900 XTX, keeping MoE serving in mind from the start.

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

I care more about stability than squeezing the last possible tokens/sec out of it.]]> https://www.datasculpting.ai/inference/showthread.php?tid=684 Sat, 03 Oct 2026 03:13:20 +0000 localcraft17]]> https://www.datasculpting.ai/inference/showthread.php?tid=684
My current setup is MLX with A6000, and I keep running into consumer GPUs.

Not looking for the perfect answer, mostly what worked in real use.]]>
My current setup is MLX with A6000, and I keep running into consumer GPUs.

Not looking for the perfect answer, mostly what worked in real use.]]> https://www.datasculpting.ai/inference/showthread.php?tid=591 Tue, 29 Sep 2026 11:32:43 +0000 llmarch51]]> https://www.datasculpting.ai/inference/showthread.php?tid=591
I tried this with Gemma using TensorRT-LLM on 7900 XTX. The headline result was better than I expected, but HBM constraints still ended up being the limiting factor.

Curious if anyone else landed in the same place.]]>
I tried this with Gemma using TensorRT-LLM on 7900 XTX. The headline result was better than I expected, but HBM constraints still ended up being the limiting factor.

Curious if anyone else landed in the same place.]]> https://www.datasculpting.ai/inference/showthread.php?tid=418 Tue, 22 Sep 2026 21:14:01 +0000 audiopoint18]]> https://www.datasculpting.ai/inference/showthread.php?tid=418
Two paths both seem viable, but they break down differently once workstation thermals 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 workstation thermals 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=347 Fri, 18 Sep 2026 18:43:25 +0000 bytelab]]> https://www.datasculpting.ai/inference/showthread.php?tid=347
I tried this with DeepSeek using SGLang on B200. The headline result was better than I expected, but batch size still ended up being the limiting factor.

Nothing scientific here, just practical notes from a real setup.]]>
I tried this with DeepSeek using SGLang on B200. The headline result was better than I expected, but batch size 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=281 Mon, 14 Sep 2026 15:54:37 +0000 torchdesk]]> https://www.datasculpting.ai/inference/showthread.php?tid=281
I've gone back and forth between LM Studio and llama.cpp for Llama. The interesting difference for me has been reranking.

Interested in tradeoffs more than fan loyalty.

My use case is pretty ordinary: local chat, some document work, and occasional batch jobs.]]>
I've gone back and forth between LM Studio and llama.cpp for Llama. The interesting difference for me has been reranking.

Interested in tradeoffs more than fan loyalty.

My use case is pretty ordinary: local chat, some document work, and occasional batch jobs.]]> https://www.datasculpting.ai/inference/showthread.php?tid=243 Sat, 12 Sep 2026 11:23:35 +0000 mangowire]]> https://www.datasculpting.ai/inference/showthread.php?tid=243
I tried this with Phi using llama.cpp on A6000. The headline result was better than I expected, but VRAM headroom still ended up being the limiting factor.

Happy to post more details if anyone wants them.]]>
I tried this with Phi using llama.cpp on A6000. The headline result was better than I expected, but VRAM headroom still ended up being the limiting factor.

Happy to post more details if anyone wants them.]]> https://www.datasculpting.ai/inference/showthread.php?tid=251 Sat, 12 Sep 2026 08:13:58 +0000 relayforge]]> https://www.datasculpting.ai/inference/showthread.php?tid=251
My current setup is PyTorch with A6000, and I keep running into prefill/decode split.

Would appreciate practical advice rather than theory.]]>
My current setup is PyTorch with A6000, and I keep running into prefill/decode split.

Would appreciate practical advice rather than theory.]]> https://www.datasculpting.ai/inference/showthread.php?tid=171 Tue, 08 Sep 2026 02:50:01 +0000 visionforge]]> https://www.datasculpting.ai/inference/showthread.php?tid=171
I'm trying to decide whether to keep the current stack or change course. Main question is how people here are handling embeddings.

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.]]>
I'm trying to decide whether to keep the current stack or change course. Main question is how people here are handling embeddings.

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=57 Fri, 28 Aug 2026 22:40:50 +0000 fastfield4]]> https://www.datasculpting.ai/inference/showthread.php?tid=57
Short version: llama.cpp behaved well for me, but only after I stopped ignoring speculative decoding.

Curious if anyone else landed in the same place.

I care more about stability than squeezing the last possible tokens/sec out of it.]]>
Short version: llama.cpp behaved well for me, but only after I stopped ignoring speculative decoding.

Curious if anyone else landed in the same place.

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