AWQ quantization of DeepSeek-V2.5-1210
To run on 8xH100 80GB, you can use vLLM with:
vllm serve adamo1139/DeepSeek-V2.5-1210-AWQ --tensor-parallel 8 --trust-remote-code
