Hope to continue updating for Ollama.

[yizhangliu]

https://github.com/ipex-llm/ipex-llm/releases/tag/v2.3.0-nightly Hope to continue updating.

[kflnig]

hope so.

[danielmayost]

I see that Ollama is now working on a Vulkan backend (ollama/ollama#11835). Also, the latest Intel AI-Playground update switched from SYCL to Vulkan: https://github.com/intel/AI-Playground/releases.

I’m curious about the performance differences between these and IPEX-LLM. Initially, it was quite slow, but I think the recent updates have improved it.

[savvadesogle]

I’m curious about the performance differences between these and IPEX-LLM

Here is some tests Vulkan and SYCL from llama.cpp (A770) on Windows

https://github.com/ggml-org/llama.cpp/discussions/10879#discussioncomment-14467566

And I do not know how to assemble it correctly llama.cpp and ollama to get good performance as in ipex-llm. https://github.com/intel/ipex-llm/issues/13309

[danielmayost]

https://github.com/ollama/ollama/pull/11835#issuecomment-3283241149

[Ellie-Williams-007]

https://github.com/intel/ipex-llm/issues/13308#issuecomment-3319233504