Created by: liym27

PR types:

New features

PR changes:

dygraph to static graph

Describe:

1. New feature 1

Use function convert_while_loop to run the transformed for loop code dynamically:

• for stmts are transformed whether it depends on Tensor or not 1). for x in range(var[]|var.numpy()[]) 2). for x in var|var.numpy() 3). for i, x enumerate(var|var.numpy())

  Note: Here var is VarBase(Tensor) or python variable

• But in run-time, the code will be run in different ways (1) run paddle op like while_loop; (2) run python while. How it runs depends on whether it depends on Tensor.

2. New feature:

support that case : variable is used and created in loop, but used before created. For example, var_a should be in loop_var and we should create it out loop.

res = 0
for i, x in enumerate(x_array):
      if i > 2:
          x = func1(var_a)
      var_a = func2(x)

3. Fix bug:

Fix bug: raise None -> return None.

4. Code optimization:

Move convert_len to file convert_operators.py