tensorflow.python.framework.errors_impl.NotFoundError: Resource __per_step_8/_tensor_arraysrpn2proposals/map/TensorArray_2_17/N10tensorflow11TensorArrayE does not exist.

[yizt]

当proposal层使用tf.map_fn时：报错如下：

  File "/root/anaconda3/envs/keras/lib/python3.6/site-packages/tensorflow/python/framework/errors_impl.py", line 519, in __exit__
    c_api.TF_GetCode(self.status.status))
tensorflow.python.framework.errors_impl.NotFoundError: Resource __per_step_8/_tensor_arraysrpn2proposals/map/TensorArray_2_17/N10tensorflow11TensorArrayE does not exist.
	 [[Node: training/SGD/gradients/rpn2proposals/map/TensorArrayStack/TensorArrayGatherV3_grad/TensorArrayGrad/TensorArrayGradV3 = TensorArrayGradV3[_class=["loc:@train...yScatterV3"], source="training/SGD/gradients", _device="/job:localhost/replica:0/task:0/device:CPU:0"](rpn2proposals/map/TensorArray_2/_2645, rpn2proposals/map/while/Exit_2/_2665)]]
	 [[Node: rpn2proposals/map_1/TensorArrayUnstack/range/_2906 = _HostSend[T=DT_INT32, client_terminated=false, recv_device="/job:localhost/replica:0/task:0/device:CPU:0", send_device="/job:localhost/replica:0/task:0/device:GPU:0", send_device_incarnation=1, tensor_name="edge_10460_rpn2proposals/map_1/TensorArrayUnstack/range", _device="/job:localhost/replica:0/task:0/device:GPU:0"](rpn2proposals/map_1/TensorArrayUnstack/range)]]

哪位知道怎么处理？类似的问题：https://stackoverflow.com/questions/52187269/tensorflow-map-fn-does-not-work-in-graph-mode-during-training-in-eager-mode-it;

因此proposal层，仍然使用tf_utils.batch_slice而不是tf.map_fn