Updated to use the latest ollama models and pinned versions

[Otterwerks]

Changes

• Pinned package versions to latest stable releases
• Specified Python 3.11.8 as the required version
• Updated model handling code to be compatible with latest ollama package

Notes

The project needed updates to work with the latest versions of its dependencies, particularly the ollama package which has had significant updates. These changes ensure reproducible environments and compatibility with current package versions.

[asciimo]

I tested this PR and it works great. Thank you!

Until this is merged, a lazy workaround is pinning ollama==0.1.8 in requirements.txt, to get benchmark.py running as originally intended.

[bigmessydata]

great patch! works great on llama 3.3

[adamj34]

I also tested this PR locally and everything worked properly. Tested on Alibaba's QWQ model.

[mykytahordia]

Works good on my machine. Thanks for update, I wish to see it in original repo, because the error was annoying.