Patrick Devine

OK, I just tested this on on an a100 80 and it's working (albeit slowly!). ``` $ ollama run mixtral:8x22b-instruct-v0.1-q5_K_M pulling manifest pulling a83a0ad30b31... 100% ▕██████████████████████████████████████████████████████████████████████████████████████████████████████████▏ 99 GB pulling 43070e2d4e53......

Sorry guys, can you try again w/ `0.1.22` and make sure you the model you're trying to use.

Can you post the [server logs](https://github.com/ollama/ollama/blob/main/docs/troubleshooting.md#how-to-troubleshoot-issues)?

That's a very small GPU w/ 2 GB of memory. I would imagine only the very smallest models would work on it. What model were you trying to load? That...

There's actually an issue filed with llama.cpp here: https://github.com/ggerganov/llama.cpp/issues/6422 I'm going to go ahead and close this as it's being tracked in llama.cpp's repo.

Hey @cedricvidal , the `ollama pull` and `ollama run` commands talk directly to the ollama server using the REST API and do not look for models on disk at all....

@woshilaowanga I think check with the webui project? Maybe they are doing things incorrectly.

@UmutAlihan we've actually been building out a test farm to better catch these issues before we release, but there are a lot of different permutations to test. Stability is incredibly...

are you behind a corporate proxy/firewall? Also, are you using wsl2 or is it just plain Windows?

I don't _think_ it's a problem with Ollama. I think it's almost certainly a Windows issue w/ DNS, but I'm not sure what's causing it.