Fix Python usage example: use llm.chat() with the chat template

Add FP8 quantization of ornith-ai/Ornith-1.5-9B

initial commit