chatllm.cpp adds support of this model
#16
by
J22
- opened
chatllm.cpp now supports this model.
Initial tests show that this model is strong on solving some tricky math problems.
server ---chat :step3-vl -ngl all --max-length 10000 +detect-thoughts
As a bonus, since this model can support native resolution mathematically using only the global view, I have added an option to test this --set native-resolution 1.
can you make a full tutorial of how to download and use this model in chatllm.cpp? it looks like gguf is not working on your fork.
@rosspanda0 This command will download quantized model automatically.
server ---chat :step3-vl .............
better doc is very necessary for good stuff.
server is built from server.nim:
https://github.com/foldl/chatllm.cpp/blob/master/docs/binding.md#web-demo