Quantized models and legalization pass

Development
4

Thanks for the reply, @FrozenGene!

The signatures of the two functions are:

def _alter_conv2d_layout(attrs, inputs, types, out_type):

def _qnn_conv2d_legalize_arm_cpu(attrs, inputs, types):

While they look similar, inputs in _alter_conv2d_layout contains actual Tensors while inputs in _qnn_conv2d_legalize_arm_cpu are of class tvm.relay.expr.Var.

Is there a way to get the tensors during legalization from variables of type tvm.relay.expr.Var?

Thanks a lot,

Giuseppe