Question

执行代码时，我收到此错误AttributeError: 'ZeroOut' object has no attribute 'eval'。我在tensorflow（https://www.tensorflow.org/extend/adding_an_op）

的标准教程中编译了它

TF_CFLAGS=( $(python -c 'import tensorflow as tf; print(" ".join(tf.sysconfig.get_compile_flags()))') )
TF_LFLAGS=( $(python -c 'import tensorflow as tf; print(" ".join(tf.sysconfig.get_link_flags()))') )
g++ -std=c++11 -shared zero_out.cc -o zero_out.so -fPIC ${TF_CFLAGS[@]} ${TF_LFLAGS[@]} -O2

main.py

import tensorflow as tf
import numpy as np

forward_module = tf.load_op_library('./zero_out.so')
with tf.Session(''):
    outgrad0 = np.arange(1,7).reshape(3,2).astype('float32')
    print(forward_module.zero_out(outgrad0).eval())

zero_out.cc文件：

#include "tensorflow/core/framework/op.h"
#include "tensorflow/core/framework/shape_inference.h"
#include "tensorflow/core/framework/op_kernel.h"
using namespace tensorflow;

REGISTER_OP("ZeroOut")
    .Input("to_zero: int32")
    .Output("zeroed: int32")
    .Output("indice: int32");

class ZeroOutOp : public OpKernel {
 public:
  explicit ZeroOutOp(OpKernelConstruction* context) : OpKernel(context) {
    }

  void Compute(OpKernelContext* context) override {
    const Tensor& input_tensor = context->input(0);
    auto input = input_tensor.flat<int32>();

    Tensor* output_tensor = NULL;
    Tensor* output_tensor_indice = NULL;
    TensorShape indice_shape;
    int dims[] = {1};
    TensorShapeUtils::MakeShape(dims, 1, &indice_shape);

    OP_REQUIRES_OK(context, context->allocate_output(0, input_tensor.shape(), &output_tensor));
    OP_REQUIRES_OK(context, context->allocate_output(1, indice_shape, &output_tensor_indice));
    auto output_flat = output_tensor->flat<int32>();
    auto indice_flat = output_tensor_indice->flat<int32>();
    indice_flat(0) = 3;
  }
};

REGISTER_KERNEL_BUILDER(Name("ZeroOut").Device(DEVICE_CPU), ZeroOutOp);

正如你所看到的，我添加了另一个输出变量，因为我想要一个元组作为输出。但我总是得到这个愚蠢的错误，我不知道如何摆脱它。

当编译没有错误产生时，但是当我执行python代码时，如果没有eval（）它就不能进行评估它至少有2个输出。

编辑：

另一种方法是在C ++中：

void addOne(int &y) // y is a reference variable
{
    y = y + 1;
}

优点是我不需要任何return。对于tensorflow，这是怎么做的？

Answer 1

问题：

out0, out1 = zero_out_module.zero_out(volume0, img0_in, img1_in, np.array([0]), 1).eval()  ---> leads to the error in the title

解决方案：

out0, out1 = zero_out_module.zero_out(volume0, img0_in, img1_in, np.array([0]), 1) 

# use eval() seperately
print (out0.eval().shape)
print (out1.eval().shape)

AttributeError：'ZeroOut'对象没有属性'eval'

1 个答案: