This model is a 4-bit quantized version of the MNN model exported from Qwen1.5-0.5B-Chat using llm-export.