update to ollama 0.4.0

[Matthww]

Hi,

Is it possible that this project could be updated to support ollama 0.4.0. I want to try the new LLama Vision model but to run those models you need atleast version 0.4.0.

Thanks!

[user7z]

i wonder why cant it just be integrated with upstream ollama ? just like nvidia & amd do

[tristan-k]

+1

[rnwang04]

Hi @Matthww , could you please let us know which exactly model do you want to run ?

[user7z]

@Matthww this one

[Matthww]

Hi @Matthww , could you please let us know which exactly model do you want to run ?

Hi @rnwang04 like @user7z mentioned I'm talking indeed about the Llama 3.2-Vision collection that can be found on ollama's model page: https://ollama.com/library/llama3.2-vision .

Ollama was updated accordingly to support these new vision models with version 0.4.0 that's available on their github repository in the releases tab.

[user7z]

@Matthww but were toaking here about the one provided with ipex-llm gpu-acceleration wich provides just 0.3.6 ollama , i think you know that ipex-llm is not integrated with the official ollama , so the context here is the accerated ollama provided with ipex-llm in this github repo

[rnwang04]

We are planning for a new rebase, if there is any progress, we will update here to let you know.

[user7z]

@wbste @rnwang04 iam not aware about the techncall details , but i think the propriatary way that ipex-llm works wouldnt make it with ollama , i mean Rocm is open , cuda mostlly open , so thats one high wall between ipex-llm & ollama , & even within intel ecosystem ipex-llm relies on oneAPI basekit , & it take lot of time for devs to be with the latest , i think for this project to continue it needs just two things :
1)_ rebase the designe
2)_ make it open as possible
Without these 2 things we cant toak about integrating this into the upstream ollama , its sad that the devs here do double work , developping ipex-llm & tinkring with ollama & every single AI utility that shows up , the devs needs to focus more on the core thing & eleminating the double work by integrating those stuff with community workflow , that would be the needed boost !

[jasonwang178]

We are planning for a new rebase, if there is any progress, we will update here to let you know.

I strongly recommend making it open source and merging the changes into the official Ollama project. This will attract more contributors and potential donations.

[qiuxin2012]

Any updates? Or is it possible for the end user to manually just grab the latest zip (i.e. ollama-windows-amd64.zip for Windows) from ollama and replacing the existing one? https://github.com/ollama/ollama/releases

No, you can't. We are still working on it.

[yizhangliu]

So slow update, sad.
Intel , are you OK?

[user7z]

@yizhangliu if your are serious about AI you should go for nvidia cuda, or amd rocm

[jason-dai]

So slow update, sad. Intel , are you OK?

Our current version is consistent with v0.4.6 of ollama. See https://github.com/intel-analytics/ipex-llm/blob/main/docs/mddocs/Quickstart/ollama_quickstart.md

[hkvision]

Do the docs not auto-build? Seems the link above says something different than the official docs...maybe when the project name changed that was left running? Or do I have the wrong URL?

Hi, the link you are referring to is not the official docs actually, it is just a previous development version and is outdated. The official one should be the link given in the previous comment below:

Our current version is consistent with v0.4.6 of ollama. See https://github.com/intel-analytics/ipex-llm/blob/main/docs/mddocs/Quickstart/ollama_quickstart.md