How to fix the error when deploy deepseek-V3 with awq on 8xH800?

[huanqi-hu]

I follow this document to deploy DeepSeek-V3 with AWQ. I downloaded the model weights from here, and ran the command below:

python3 -m sglang.launch_server
--model-path [DeepSeek-V3-AWQ-MODEL_PATH]
--tp-size 8
--trust-remote
--dtype half
But I encountered the following error message:

ValueError: Unknown quantization method: awq. Must be one of ['fp8', 'blockwise_int8', 'modelopt', 'w8a8_int8', 'w8a8_fp8', 'compressed-tensors']
It seems that the awq quantization method is not included in BASE_QUANTIZATION_METHODS. So I added "awq": AWQConfig to the dictionary, but then another error occurred:

AttributeError: type object 'DummyConfig' has no attribute 'override_quantization_method'
It looks like awq is not officially supported yet.
I would like to know: has anyone else faced the same problem, and how did you fix it?