我按照https://www.tensorflow.org/versions/r0.10/how_tos/distributed/中的示例创建了分布式TensorFlow，其中包含一个ps和两个worker . 我正在测试的机器中只有1个cpu和8个内核 .
InvalidArgumentError (see above for traceback): Cannot assign a device to node 'save/RestoreV2_14': Could not satisfy explicit device specification '/job:ps/task:0/device:CPU:0' because no devices matching that specification are registered in this process; available devices: /job:local/replica:0/task:0/cpu:0, /job:local/replica:0/task:1/cpu:0, /job:worker/replica:0/task:1/cpu:0
[[Node: save/RestoreV2_14 = RestoreV2[dtypes=[DT_INT32], _device="/job:ps/task:0/device:CPU:0"](save/Const, save/RestoreV2_14/tensor_names, save/RestoreV2_14/shape_and_slices)]]
我已经将server.target作为参数传递给sv.prepare_or_wait_for_session
sess = sv.prepare_or_wait_for_session（server.target）
不知道是什么原因造成的？
我的培训代码是
parser = argparse.ArgumentParser(description='tensorflow')
parser.add_argument('--job_name', dest='job_name')
parser.add_argument('--task_index', dest='task_index', default=0)
args = parser.parse_args()
ps_hosts = ['localhost:2222']
worker_hosts = ['localhost:2223', 'localhost:2224']
job_name = args.job_name
task_index = int(args.task_index)
# Create a cluster from the parameter server and worker hosts.
cluster = tf.train.ClusterSpec({"ps": ps_hosts, "worker": worker_hosts})
# Create and start a server for the local task.
server = tf.train.Server(cluster, job_name=job_name, task_index=task_index)
if job_name == "ps":
server.join()
elif job_name == "worker":
with tf.device(tf.train.replica_device_setter(
worker_device="/job:worker/task:%d" % task_index,
cluster=cluster)):
total_input_features = len(train_x[0])
x = tf.placeholder('float', [None, total_input_features])
y = tf.placeholder('float')
global_step = tf.Variable(0, name="global_step", trainable=False)
is_chief = (task_index == 0)
prediction = neural_network_model(x, total_input_features, n_nodes_hl1,
first_layer_activation,
n_nodes_hl2,
second_layer_activation)
total_loss = tf.reduce_mean(tf.square(prediction - y))
optimizer = tf.train.AdamOptimizer()
train_op = optimizer.minimize(total_loss, global_step=global_step)
init_op = tf.initialize_all_variables()
sv = tf.train.Supervisor(
is_chief=is_chief,
logdir="/tmp/train_logs",
init_op=init_op,
global_step=global_step)
print '******** ALL CREATED ********'
# The supervisor takes care of session initialization, restoring from
# a checkpoint, and closing when done or an error occurs.
with sv.managed_session(server.target) as sess:
# Loop until the supervisor shuts down or 1000000 steps have completed.
step = 0
while not sv.should_stop() and step < 1000000:
# Run a training step asynchronously.
# See `tf.train.SyncReplicasOptimizer` for additional details on how to
# perform *synchronous* training.
train_feed = {x: train_x, y: train_y}
_, step = sess.run([train_op, global_step], feed_dict=train_feed)
if step % 100 == 0:
print "Done step %d" % step
sv.stop()
完整的堆栈跟踪是：
Traceback（最近一次调用最后一次）：文件“.... / PycharmProjects / SparkProject / server_client_updated.py”，第162行，用sv.managed_session（server.target）作为sess：File“/ System / Library / Frameworks / Python.framework / Versions / 2.7 / lib / python2.7 / contextlib.py“，第17行，在enter中返回self.gen.next（）文件”.... / tensorflow / lib / python2.7 / site-packages /tensorflow/python/training/supervisor.py“，第973行，在managed_session self.stop（close_summary_writer = close_summary_writer）文件”.... / tensorflow / lib / python2.7 / site-packages / tensorflow / python / training / supervisor.py“，第801行，在stop stop_grace_period_secs = self._stop_grace_secs）文件”.... / tensorflow / lib / python2.7 / site-packages / tensorflow / python / training / coordinator.py“，第386行， join six.reraise（* self._exc_info_to_raise）file“.... / tensorflow / lib / python2.7 / site-packages / tensorflow / python / training / supervisor.py”，第962行，在managed_session start_standard_services = start_standard_services）文件” .... / tensorflow / lib中/ python2.7 /站点包s / tensorflow / python / training / supervisor.py“，第726行，在prepare_or_wait_for_session中max_wait_secs = max_wait_secs）文件”.... / tensorflow / lib / python2.7 / site-packages / tensorflow / python / training / session_manager.py “，第384行，在wait_for_session sess中）文件”.... / tensorflow / lib / python2.7 / site-packages / tensorflow / python / training / session_manager.py“，第467行，在_try_run_local_init_op sess.run中（自我 . local_init_op）文件“.... / tensorflow / lib / python2.7 / site-packages / tensorflow / python / client / session.py”，第767行，运行run_metadata_ptr）文件“/ Users / jabermo / tensorflow / lib / python2.7 / site-packages / tensorflow / python / client / session.py“，第965行，在_run feed_dict_string，options，run_metadata中）文件”.... / tensorflow / lib / python2.7 / site-packages / tensorflow /python/client/session.py“，第1015行，在_do_run target_list，options，run_metadata中）文件”.... / tensorflow / lib / python2.7 / site-packages / tensorflow / python / client / session.py“ ，第1035行，在_do_call中提升类型（e）（node_def，op，message）tensorflow.py thon.framework.errors_impl.InvalidArgumentError：无法将设备分配给节点'save / RestoreV2_14'：无法满足显式设备规范'/ job：ps / task：0 / device：CPU：0'因为没有注册匹配该规范的设备在这个过程中;可用设备：/ job：local / replica：0 / task：0 / cpu：0，/ job：local / replica：0 / task：1 / cpu：0，/ job：worker / replica：0 / task：1 / cpu：0 [[Node：save / RestoreV2_14 = RestoreV2 [dtypes = [DT_INT32]， device =“/ job：ps / task：0 / device：CPU：0”]（save / Const，save / RestoreV2_14 / tensor_names，save / RestoreV2_14 / shape_and_slices）]]由op u'save / RestoreV2_14'引起，定义于：文件“.... / PycharmProjects / SparkProject / server_client_updated.py”，第127行，在global_step = global_step）文件“.... /tensorflow/lib/python2.7/site-packages/tensorflow/python/training/supervisor.py“，第313行，在init self._init_saver（saver = saver）文件”.... / tensorflow / lib / python2 . 7 / site-packages / tensorflow / python / training / supervisor.py“，第459行，在_init_saver saver = saver_mod.Saver（）文件”.... / tensorflow / lib / python2.7 / site-packages / tensorflow / python / training / saver.py“，第1051行，在init self.build（）文件”.... / tensorflow / lib / python2.7 / site-packages / tensorflow / python / training / saver.py“，line 1081，在build restore_sequentially = self._restor中e_sequentially）文件“.... / tensorflow / lib / python2.7 / site-packages / tensorflow / python / training / saver.py”，第675行，在build restore_sequentially，reshape）文件“.... / tensorflow / lib / python2.7 / site-packages / tensorflow / python / training / saver.py“，第402行，_AddRestoreOps张量= self.restore_op（filename_tensor，saveable，preferred_shard）文件”.... / tensorflow / lib / python2 .7 / site-packages / tensorflow / python / training / saver.py“，第242行，在restore_op [spec.tensor.dtype]中）[0]）文件”.... / tensorflow / lib / python2.7 / site-packages / tensorflow / python / ops / gen_io_ops.py“，第668行，在restore_v2中dtypes = dtypes，name = name）文件”.... / tensorflow / lib / python2.7 / site-packages / tensorflow / python /framework/op_def_library.py“，第763行，在apply_op中op_def = op_def）文件“.... / tensorflow / lib / python2.7 / site-packages / tensorflow / python / framework / ops.py”，第2395行，在create_op original_op = self._default_original_op，op_def = op_def）文件“... ./tensorflow/lib/python2.7/site-packages/tensorflow/python/framework/ops.py“，第1264行，在init self._traceback = extract_stack（）InvalidArgumentError（参见上面的回溯）：无法将设备分配给节点'save / RestoreV2_14'：无法满足显式设备规范'/ job：ps / task：0 / device：CPU：0'，因为在此过程中没有注册与该规范匹配的设备;可用设备：/ job：local / replica：0 / task：0 / cpu：0，/ job：local / replica：0 / task：1 / cpu：0，/ job：worker / replica：0 / task：1 / cpu：0 [[Node：save / RestoreV2_14 = RestoreV2 [dtypes = [DT_INT32]， device =“/ job：ps / task：0 / device：CPU：0”]（save / Const，save / RestoreV2_14 / tensor_names，save / RestoreV2_14 / shape_and_slices）]]