TensorFlow lite .pb 转 .tflite 报错

Question

错误信息

 F tensorflow/contrib/lite/toco/toco.cc:76] Check failed: has_input_file != has_savedmodel_dir (0 vs. 0)Specify either input_file or savedmodel_directory flag.

frozen_graph https://storage.googleapis.com/download.tensorflow.org/models/mobilenet_v1_1.0_224_frozen.tgz

转换命令

bazel run toco  \
  input_file=/Users/dchealth/Desktop/mobilenet/frozen_graph.pb \
  output_file=/Users/dchealth/Desktop/mobilenet/frozen_graphnew.tflite \
  output_format=TFLITE \
  input_format=TENSORFLOW_GRAPHDEF \
  inference_type=FLOAT \
  input_type=FLOAT \
  input_shapes=1,128,128,3 \
  input_arrays=input \
  output_arrays=MobilenetV1/Predictions/Reshape_1

什么原因

enter image description here

Answer 1

我认为您需要为标志传递 '--'，如下所示。

bazel run --config=opt \
  tensorflow/contrib/lite/toco:toco -- \
  --input_file=/Users/dchealth/Desktop/mobilenet/frozen_graph.pb \
  --output_file=/Users/dchealth/Desktop/mobilenet/frozen_graphnew.tflite \
  --inference_type=FLOAT \
  --input_type=FLOAT \
  --input_shape=1,128,128,3 \
  --input_arrays=input \
  --output_arrays=MobilenetV1/Predictions/Reshape_1

TensorFlow lite .pb 转 .tflite 报错

TensorFlow lite .pb turn .tflite An error

android

ios

tensorflow

tensorflow-lite