Add docker instructions to README.md

[tunglinwood]

Hi Moonshot team,

Thank you for providing this amazing model.
I have tried starting kimi-audio-7b with docker, and here is my walkthrough. Please kindly review and feel free to merge. Thank you so much 😄

[MoyanZitto]

@tunglinwood Thank you for your interest in our project. This PR adds more instructions that should be noticed by users, but I think it contains too many details. Here are my suggestions:

• Skip the explanations and verifications, and merge the key actions for preparing the environment into the "Quick Start" section (e.g., git clone --recurse-submodules, cd ..).
• For Docker, there are two things users might miss, so I think it would be helpful to include a section before "Quick Start":
  • We have a pre-built Docker image available: docker pull moonshotai/kimi-audio:v0.1. （which is documented in huggingface repo but not here )
  • You may need to mount your local directory to access the cloned repository, and you need to use --gpus to enable GPU support (i.e., in your Docker command).
• You don't need to manually download the Hugging Face model and pass the local path. Just use the model_id, pls verify this.

[tunglinwood]

Hi, @MoyanZitto I have updated the pull, please kindly review.

[tunglinwood]

@MoyanZitto Hello could you kindly help to review the pull?

[MoyanZitto]

Suggested change

	### 1. Inference with Docker
	### 1. Set environment

[MoyanZitto]

Suggested change

	- [1. Inference with Docker](#1-inference-with-docker)
	- [1. Set environment](#1-set-environment)

[MoyanZitto]

I believe we can reorganize readme:

• get code ( note you need submodule update --init）
• set envrionment
  • use pre-built image ( we should prefer this because it's simpler)
  • build your own image
• run demo
  • i.e inference with python