You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Fluid was designed a trail towards a new DL programming language. It doesn't make sense for users to write a Python program that calls the Python interpreter. However, this is what we are doing right now:
The correct solution is to use the standard package atexit:
When we import paddle.fluid, we run the initialization module of Fluid, which should call atexit.register(run_fluid_program).
When the Python interpret runs the Fluid application program, it generates a ProgramDesc message.
When the application completes, Python calls run_fluid_program, which should call the C++ class Executor run interpret and run the generated ProgramDesc message.
Fluid was designed a trail towards a new DL programming language. It doesn't make sense for users to write a Python program that calls the Python interpreter. However, this is what we are doing right now:
Paddle/python/paddle/fluid/tests/book/test_fit_a_line.py
Line 104 in 25317bd
Let us remove the expose of Executor, and follow the design #7464.
The text was updated successfully, but these errors were encountered: