Data Free Quantization

Data-Free Quantization is a technique aimed at achieving highly accurate quantized models without accessing any training data. This method effectively addresses the issue of reliance on original data in traditional quantization processes by synthesizing boundary support samples, among other means, thereby enhancing the flexibility and security of model deployment, especially in applications where data privacy is sensitive.