Created by: zhizunbao-y

    student_program = fluid.Program()
    s_startup = fluid.Program()
    with fluid.program_guard(student_program, s_startup):
        with fluid.unique_name.guard():
            # model definition
           ...
    place = fluid.CUDAPlace(0) if args.use_gpu else fluid.CPUPlace()
    exe = fluid.Executor(place)
    load_model(exe, student_program, args.pretrained_model)  #prune.io
    val_program = student_program.clone(for_test=True)

    teacher_model = models.__dict__[args.teacher_model]()
    teacher_program = fluid.Program()
    t_startup = fluid.Program()
    with fluid.program_guard(teacher_program, t_startup):
        with fluid.unique_name.guard():
            # teacher model definition
           ...

    exe.run(t_startup)
    if args.teacher_pretrained_model:
        def if_exist(var):
            return os.path.exists(
                os.path.join(args.teacher_pretrained_model, var.name))
        fluid.io.load_vars(
            exe,
            args.teacher_pretrained_model,
            main_program=teacher_program,
            predicate=if_exist)

    data_name_map = {'images': 'images'}
    merge(teacher_program, student_program, data_name_map, place)
    with fluid.program_guard(student_program, s_startup):
        dist_loss = soft_label_loss("teacher_fc_0.tmp_0", "fc_0.tmp_0", student_program)
        loss = avg_cost + dist_loss
        lr, opt = create_optimizer(args)
        opt.minimize(loss)
    exe.run(s_startup)

会报以下错误：

Error: Param and Velocity of MomentumOp should have the same dimension.
  [Hint: Expected param_dim == ctx->GetInputDim("Velocity"), but received param_dim:256 != ctx->GetInputDim("Velocity"):128.] at (/paddle/paddle/fluid/operators/optimizers/momentum_op.h:79)
  [operator < momentum > error]

怀疑是exe.run(s_startup)这句代码覆盖了load_model ，但是optimizer又需要初始化，请问如何解决？