v2.3.0

• RKNN-Toolkit2 support ARM64 architecture

• RKNN-Toolkit-Lite2 support installation via pip

• Add support for W4A16 symmetric quantization (RK3576)

• Operator optimization, such as LayerNorm, LSTM, Transpose, MatMul, etc.