Request: I think the name should be "34b" rather than "c34b".
As mentioned in the title, I have an opinion. I believe that people don't see your 34b models during searches, because they have to attach an "c" to the parameter count in order for your CodeLlama variants to show up.
I was still trying to be optimistic and holding out for a base llama-34b. But, alas, it may never happen.
Personally, I think that CodeLlama is basically Llama 2.1. Don't think I heard anyone say that 34b had the issues that were reported for other sizes.
I would assume "CodeLlama" has been trained on a majority code tasks and logic, and thus is less optimised for conversations and story prompts, hence adding the "c". It'd be misleading to name this in a way that doesn't denote the fact that it's not a "standard" llama2 34b model.
However you could name it "spicyboros-cl2-34b-2.2" to avoid any mistaken interpretations of the "34b" value. I can imagine some programs/scripts could attempt parsing it, for example.
Also, the model card doesn't mention the context size, but the gguf suggests that its 16k, is that correct? What impact does using it at non-16k context sized have on its results?