fix typo in default model path

[ott2]

Elsewhere in the code the default model path is models/llama-7B/ggml-model.bin; that string appears in

examples/embedding/embedding.cpp
examples/save-load-state/save-load-state.cpp
examples/perplexity/perplexity.cpp
examples/main/main.cpp

This change makes ./main -h print the correct default path.

It might make sense to refactor the code to reference the default path only once. This change makes that refactoring easier to do.

[prusnak]

Readme mentions ./models/7B/ggml-model-q4_0.bin and I think we should use this everywhere for unification.

Suggested change

	std::string model = "models/llama-7B/ggml-model.bin"; // model path
	std::string model = "models/7B/ggml-model-q4_0.bin"; // model path

Can you also update this PR to remove params.model = "models/llama-7B/ggml-model.bin"; from files you mention? It doesn't make sense to override the value if it is the same as the default, no?

[ott2]

Should we use models/7B or models/llama-7B as the canonical path? I used to have just 7B but with all the new models supported by llama.cpp I'm now tending to prefer llama-7B for disambiguation.

[prusnak]

I prefer to keep 7B because that's how the directory is named when you download the model.

Also the instructiond should work no matter what model is located in the 7B directory (Alpaca, Vicuna, etc.)

[ott2]

I've implemented your suggested changes. The default model path is now defined in examples/common.h and this is used elsewhere implicitly, instead of repeating the string in several places.