fix embedding by adding fixes from llama.cpp upstream

[deadbeef84]

Embedding appears broken since v0.1.32

See #3777 #4207 for details.

This PR applies fixes based on ggerganov/llama.cpp@1b67731#diff-87355a1a297a9f0fdc86af5e2a59cae153290f58d68822cd10c30fee4f7f7076.

I've tested it and embedding vectors looks correct after applying this patch.

[fredrik-smedberg]

I can confirm this PR indeed fixes very obvious issues I had when doing embedding and queries with mxbai-embed-large and nomic-embed-text models.

[multiduplikator]

Desperately waiting for this fix to be integrated into next version. Having to stay on 0.1.31 is kind of a pain ...

[fredrik-smedberg]

@multiduplikator are you on a Mac? If so, it's very easy to download, compile and run this branch on your computer.

1. Install Brew if you haven't already (https://brew.sh)

2. Check out this branch,

git clone https://github.com/ollama/ollama.git
cd ollama
git fetch origin pull/4399/head:pr-4399
git checkout pr-4399

3. Follow the instructions here https://github.com/ollama/ollama?tab=readme-ov-file#building

4. Then just run it, like in the examples here, https://github.com/ollama/ollama?tab=readme-ov-file#running-local-builds

[multiduplikator]

@fredrik-smedberg Not on a Mac. I still have time to wait ... would like to avoid getting into custom building. But thanks for the input. Will come in handy when the time comes :)

[0ssamaak0]

GJ @deadbeef84 🚀🚀
I just found out this is not working as expected after creating a vector db for 50k examples 😅😅 gonna use HF untill this get merged

[0ssamaak0]

@jmorganca please merge this! Thanks 😁😁

[youkefan18]

@jmorganca Hi, I can confirm this is important fix for any application running in prod.
The change in embedding results after upgrading ollama ruins all RAG related feature.