Model deployment instructions seem outdated and inconsistent

[meng-pinetree]

Hi, thanks for open-sourcing this project.

I am trying to reproduce ComputerRL on Google Cloud, but I ran into an issue with the model deployment instructions in the README. The documented model entry points seem inconsistent or no longer publicly accessible.

According to the README:

• open-source text-only model: ModelScope/ComputerRL
• open-source multimodal model: ModelScope/ComputerRL-V

But the sglang.launch_server example uses:

python -m sglang.launch_server \
  --model zai-org/autoglm-os-9b \
  --host 0.0.0.0 --port 30000 --served-model-name autoglm-os

However, I observed the following:

1. zai-org/autoglm-os-9b does not work as a public Hugging Face model ID for me.
   huggingface_hub returns a 401 / repository not found style error, and sglang.launch_server fails with:

OSError: zai-org/autoglm-os-9b is not a local folder and is not a valid model identifier listed on 'https://huggingface.co/models'

2. ModelScope/ComputerRL also does not seem to exist.
   Trying to download it with modelscope.snapshot_download('ModelScope/ComputerRL', ...) returns 404.

So currently:

• zai-org/autoglm-os-9b is not accessible to me as a public model
• ModelScope/ComputerRL returns 404
• the README contains both names, so it is unclear which model path is the correct one today

For reference, my environment is:

• Google Cloud
• A100 80GB
• CUDA 12.4
• Python 3.10

The base GPU environment is working, so the blocker seems specifically to be the model identifier / access path.

Could you clarify:

1. What is the correct current model path to use with sglang.launch_server?
2. Is zai-org/autoglm-os-9b private / gated / renamed?
3. Are ModelScope/ComputerRL and ModelScope/ComputerRL-V still valid, or have they moved?

Thanks!