I quantized a MobileNetV2 with linear output. Conv+BN+Relu were fused to ConvBnReLU2d and the Linear+Relu was fused to a LinearReLU layer. When executing it, I get the error message:
Didn't find kernel to dispatch to for operator 'quantized::linear_relu'.
All the conv layers seem to work correctly. According to the documentation LinearReLU is supported: https://pytorch.org/docs/stable/quantization.html#torch.nn.intrinsic.LinearReLU