Martin7-1

Sounds great for Chinese Developers. Maybe I will PR for it in 3-4 weeks

Hi @imoxu , I'm tring my best to contribute, maybe in 1-2 weeks I will PR for it.

I also want to know how `llama2` get embeddings...It confuses me few days. In `llama.cpp` it looks like get embeddings in decode phase (see [`llama_decode_internal()`](https://github.com/ggerganov/llama.cpp/blob/master/llama.cpp#L6203)), but I don't know exactly...

Any update on this issue? I observe the same thing.

BTW, I found that there are some token usage differences. I don't know what the reason is

> HI @Martin7-1 thank you, this is long due! > > A few thoughts: > > * while it is great to have a default implementation that will work for...

> We can add a mapping (model->dimensions) and a conditional logic in the `dimension()` method. If the model is known (is in the mapping), return mapped value. If not, do...

Sounds great. I think it's a good way to manage artifacts.