Module codellama 

CodeLlama Model V2 - Clean implementation

CodeLlama architecture features:

• Based on LLaMA 2 architecture
• Extended context window (up to 100k tokens)
• Infilling capability (FIM tokens)
• RoPE with extended frequencies

Structs

CodeLlamaAttention

CodeLlamaConfig

CodeLlamaLayer

CodeLlamaMLP

CodeLlamaModelV2