For those encountering this problem, I experienced it too but managed to resolve it! It seems that llama models are not optimal for embedding. There...