Julian Yap

Yeah I’m saying the flag parsing should be consistent behavior, not trial and error.

Add a test case or test against the API?

Added test case. Also confirming this was tested against the Twitter API and actually works.

Mm, yeah on further testing with larger images (but within the 700kb documented limit) this doesn't work too well. Errors: * 430 Request Header Fields Too Large * stream error:...

@dghubble OK it works now with the latest commit.

Not sure if you’re still testing but this isn’t merged to Main branch.

Confirmed this fixed my issue.

For anyone wondering, here's how you can manually disable AVX to build Ollama. ``` $ git diff diff --git a/llm/llama.cpp/generate_linux.go b/llm/llama.cpp/generate_linux.go index ce9e78a..77c9795 100644 --- a/llm/llama.cpp/generate_linux.go +++ b/llm/llama.cpp/generate_linux.go @@ -14,13...

> I can confirm that the fix also allows loading the new 2.0 inpainting model without any issues. Can additionally confirm this fixes loading 2.0 inpainting model.

Is this related to Stable Diffusion?