TensorFlow ObjectDetection 多 GPU

查看全部 · 发表在已解决 2018-5-15 18:10:28

快来加入 TensorFlowers 大家庭！

您需要登录才可以下载或查看，没有帐号？加入社区

x

本帖最后由 AngelZheng 于 2018-5-19 16:19 编辑

我在使用 Google TensorFlow Object Detection API 训练 ssd_mobileNet_v1 模型的时候遇到了个问题，我在只使用一个 1050Ti 的 GPU 训练时，训练没有出现问题，并且默认的 200000 步训练完毕。但当我使用两个 1050Ti 的 GPU 做训练时，训练到一定的步数会出现 Nan in summary histogram 的问题，请问是怎么回事呢？

TensorFlow-GPU 的版本是 1.6.0
Models 库今天（2018/05/15）刚刚从 Github 中 Clone 的 Master 分支

下面是错误的关键字：

INFO:tensorflow:Error reported to Coordinator: Nan in summary histogram for: ModelVars/FeatureExtractor/MobilenetV1/Conv2d_2_depthwise/BatchNorm
/moving_variance

下面是错误的完整内容

INFO:tensorflow:Error reported to Coordinator: Nan in summary histogram for: ModelVars/FeatureExtractor/MobilenetV1/Conv2d_2_depthwise/BatchNorm
/moving_variance
      [[Node: ModelVars/FeatureExtractor/MobilenetV1/Conv2d_2_depthwise/BatchNorm/moving_variance = HistogramSummary[T=DT_FLOAT, _device="/jo
b:localhost/replica:0/task:0/device:CPU:0"](ModelVars/FeatureExtractor/MobilenetV1/Conv2d_2_depthwise/BatchNorm/moving_variance/tag, FeatureExtr
actor/MobilenetV1/Conv2d_2_depthwise/BatchNorm/moving_variance/read)]]
      [[Node: Loss/assert_equal_5/Assert/Assert/_1184 = _Recv[client_terminated=false, recv_device="/job:localhost/replica:0/task:0/device:GP
U:0", send_device="/job:localhost/replica:0/task:0/device:CPU:0", send_device_incarnation=1, tensor_name="edge_6302_Loss/assert_equal_5/Assert/A
ssert", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:GPU:0"]()]]
Caused by op 'ModelVars/FeatureExtractor/MobilenetV1/Conv2d_2_depthwise/BatchNorm/moving_variance', defined at:
  File "object_detection/train.py", line 184, in <module>
tf.app.run()
  File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/platform/app.py", line 126, in run
_sys.exit(main(argv))
  File "object_detection/train.py", line 180, in main
graph_hook_fn=graph_rewriter_fn)
  File "/home/jovyan/Appendix/tensorflow_models/research/object_detection/trainer.py", line 338, in train
model_var.op.name, model_var))
  File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/summary/summary.py", line 193, in histogram
tag=tag, values=values, name=scope)
  File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/ops/gen_logging_ops.py", line 189, in _histogram_summary
"HistogramSummary", tag=tag, values=values, name=name)
  File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/op_def_library.py", line 787, in _apply_op_helper
op_def=op_def)
  File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/ops.py", line 3271, in create_op
op_def=op_def)
  File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/ops.py", line 1650, in __init__
self._traceback = self._graph._extract_stack()  # pylint: disable=protected-access
InvalidArgumentError (see above for traceback): Nan in summary histogram for: ModelVars/FeatureExtractor/MobilenetV1/Conv2d_2_depthwise/BatchNor
m/moving_variance
      [[Node: ModelVars/FeatureExtractor/MobilenetV1/Conv2d_2_depthwise/BatchNorm/moving_variance = HistogramSummary[T=DT_FLOAT, _device="/jo
b:localhost/replica:0/task:0/device:CPU:0"](ModelVars/FeatureExtractor/MobilenetV1/Conv2d_2_depthwise/BatchNorm/moving_variance/tag, FeatureExtr
actor/MobilenetV1/Conv2d_2_depthwise/BatchNorm/moving_variance/read)]]
      [[Node: Loss/assert_equal_5/Assert/Assert/_1184 = _Recv[client_terminated=false, recv_device="/job:localhost/replica:0/task:0/device:GP
U:0", send_device="/job:localhost/replica:0/task:0/device:CPU:0", send_device_incarnation=1, tensor_name="edge_6302_Loss/assert_equal_5/Assert/A
ssert", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:GPU:0"]()]]
Traceback (most recent call last):
  File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1361, in _do_call
return fn(*args)
  File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1340, in _run_fn
target_list, status, run_metadata)
  File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/errors_impl.py", line 516, in __exit__
c_api.TF_GetCode(self.status.status))
tensorflow.python.framework.errors_impl.InvalidArgumentError: Nan in summary histogram for: ModelVars/FeatureExtractor/MobilenetV1/Conv2d_2_dept
hwise/BatchNorm/moving_variance
      [[Node: ModelVars/FeatureExtractor/MobilenetV1/Conv2d_2_depthwise/BatchNorm/moving_variance = HistogramSummary[T=DT_FLOAT, _device="/jo
b:localhost/replica:0/task:0/device:CPU:0"](ModelVars/FeatureExtractor/MobilenetV1/Conv2d_2_depthwise/BatchNorm/moving_variance/tag, FeatureExtr
actor/MobilenetV1/Conv2d_2_depthwise/BatchNorm/moving_variance/read)]]
      [[Node: Loss/assert_equal_5/Assert/Assert/_1184 = _Recv[client_terminated=false, recv_device="/job:localhost/replica:0/task:0/device:GP
U:0", send_device="/job:localhost/replica:0/task:0/device:CPU:0", send_device_incarnation=1, tensor_name="edge_6302_Loss/assert_equal_5/Assert/A
ssert", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:GPU:0"]()]]

我在 Stack Overflow 上搜索过相关的问题，上面说可能是数据集过小或者标签文件。

经过测试，我在单 1050Ti GPU 上训练未出现问题，因此数据集是没问题的。

是不是 TensorFlow Object Detection API 不支持多 GPU 训练或者需要单独做设置呢？我看 models 库里很多模型介绍都是以单 GPU 为标准（虽然那个卡略强）。

我知道答案回答被采纳将会获得10 金币 + 5 金币已有2人回答