FROM pytorch/pytorch:2.2.1-cuda12.1-cudnn8-runtime

# Instalujemy vLLM jednym poleceniem.
RUN pip install vllm

# Uruchamiamy serwer vLLM z Pana modelem.
CMD ["vllm", "serve", "adowu/falcon-h1-3b-instruct-pl-bfloat16", "--host", "0.0.0.0", "--trust-remote-code"]