LLM inference optimization: Architecture, KV cache and Flash attention

Update Required To play the media you will need to either update your browser to a recent version or update your Flash plugin.