Created by: Aurelius84

PR types

Function optimization

PR changes

APIs

Describe

enhance error message for StaticLayer.concrete_program.

Before

ValueError: No valid transformed program for function: foo_func(a,b,c,d), input_spec: None.

It will also trigger another error before this in len(flatten_input_spec) > 0. In this PR, I fix it.

TypeError: object of type 'NoneType' has no len()

After

ValueError: No valid transformed program for function: foo_func(a,b,c,d), input_spec: None.
            Please specific `input_spec` in `@paddle.jit.to_static` or feed input tensor to call the decorated function at once.