错误运行TensorFlow Lite示例

时间:2018-07-11 09:04:44

标签: tensorflow tensorflow-lite

我尝试构建并运行tensorflow lite示例(最小)。

构建正常,但是运行该程序时发生错误。

为什么会发生此错误?

使用的模型是对tensorflow进行自定义模型训练,并使用toco转换tensorflow lite。

root@90f212114f89:/tensorflow# bazel-bin/tensorflow/contrib/lite/examples/minimal/minimal /root/DNNSE/model/tflite_model.lite
tensorflow/contrib/lite/kernels/reduce.cc:107 current >= 0 && current < input_num_dims was not true.
tensorflow/contrib/lite/kernels/reduce.cc:107 current >= 0 && current < input_num_dims was not true.
tensorflow/contrib/lite/kernels/reduce.cc:107 current >= 0 && current < input_num_dims was not true.
tensorflow/contrib/lite/kernels/reduce.cc:107 current >= 0 && current < input_num_dims was not true.
Node 8 failed to prepare.

Error at tensorflow/contrib/lite/examples/minimal/minimal.cc:59

节点8无法准备。 <-在解释器PrepareOpsStartingAt()函数中抛出此错误。

我认为模型转换.tflite时会出错

如何调试此问题?

我在下面使用了构建选项。

 bazel build ... --compilation_mode=dbg -s

还有其他有用的构建选项吗?

0 个答案:

没有答案