我正在运行TensorFlow附带的deepdream.ipnyb
示例。第render_naive(T(layer)[:,:,:,channel])
行给了我CopyCPUTensorToGPU: GPU Memcpy failed
。
一种可能性是,来自TensorFlow HEAD的笔记本电脑与我最近发布的TensorFlow安装电脑不匹配。
但如果是这种情况,我会期望更多的人有同样的问题。这似乎是一个非常罕见的错误,因为CopyCPUTensorToGPU: GPU Memcpy failed
的{{3}}会产生总共5个结果。
第二种可能性是,这是我运行的特定机器的问题,EC2的g2.2xlarge。
Stack:EC2的g2.2xlarge机器,Python 2,Jupyter
完整追踪:
---------------------------------------------------------------------------
InternalError Traceback (most recent call last)
<ipython-input-4-401ce587fe66> in <module>()
35 showarray(visstd(img))
36
---> 37 render_naive(T(layer)[:,:,:,channel])
<ipython-input-4-401ce587fe66> in render_naive(t_obj, img0, iter_n, step)
27 img = img0.copy()
28 for i in xrange(iter_n):
---> 29 g, score = sess.run([t_grad, t_score], {t_input:img})
30 # normalizing the gradient, so the same step size should work
31 g /= g.std()+1e-8 # for different layers and networks
/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc in run(self, fetches, feed_dict)
347
348 # Run request and get response.
--> 349 results = self._do_run(target_list, unique_fetch_targets, feed_dict_string)
350
351 # User may have fetched the same tensor multiple times, but we
/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc in _do_run(self, target_list, fetch_list, feed_dict)
421 # pylint: disable=protected-access
422 raise errors._make_specific_exception(node_def, op, e.error_message,
--> 423 e.code)
424 # pylint: enable=protected-access
425 raise e_type, e_value, e_traceback
InternalError: CopyCPUTensorToGPU: GPU Memcpy failed
[[Node: import/mixed3b/_197 = _Recv[client_terminated=false, recv_device="/job:localhost/replica:0/task:0/gpu:0", send_device="/job:localhost/replica:0/task:0/cpu:0", send_device_incarnation=1, tensor_name="edge_1001_import/mixed3b", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/gpu:0"]()]]