2

  1. Created a simple dummy sequential model in tf.keras as shown below:

    model = tf.keras.Sequential()
model.add(layers.Dense(10, input_shape=(100, 100)))
model.add(layers.Conv1D(3, 2))
model.add(layers.Flatten())
model.add(layers.Dense(10, activation='softmax', name='predict_10'))

  2. Trained the model and saved it using tf.keras.models.saved_model.

  3. To get the input input and output node names used saved_model_cli.

    saved_model_cli show --dir "path/to/SavedModel" --all

    enter image description here

  4. Froze the saved model with freeze_graph.py utility.

    python freeze_graph.py --input_saved_model_dir=<path/to/SavedModel> --output_graph=<path/freeze.pb> --input_binary=True --output_node_names=StatefulPartitionedCall

    Model is frozen.

Now Here's the main issue:

  1. To load the frozen graph I've used this guide Migrate tf1.x to tf2.x (wrap_frozen_graph)
  2. Used 
    with tf.io.gfile.GFile("patf/to/freeze.pb", 'rb') as f:
   graph_def = tf.compat.v1.GraphDef()
   graph_def.ParseFromString(f.read())
load_frozen = wrap_frozen_graph(graph_def, inputs='dense_3_input:0', outputs='predict_10:0')
  3. Output error ValueError: Input 1 of node StatefulPartitionedCall was passed float from dense_3/kernel:0 incompatible with expected resource.

I'm getting same error when converting .pb to .dlc (Qualcomm). Actually I want to run original model on Qualcomm's Hexagon DSP or GPU.

desertnaut
  • 57,590
  • 26
  • 140
  • 166
Arvind
  • 198
  • 2
  • 11
  • Use `model = tf.keras.models.load_model('saved_model/my_model')` library to load the saved model. For more information take a look at [tensorflow site](https://www.tensorflow.org/tutorials/keras/save_and_load#savedmodel_format). Thanks! –  Mar 08 '21 at 10:14
  • 1
    @TFer My question was using frozen model and not using SavedModel which has separate files for weights and graph. `load_model` works for SavedModel. For edge devices it's good to have a single (pb) file. – Arvind Mar 08 '21 at 10:41
  • 1
    Does [this](https://stackoverflow.com/a/56384808/14290244) answer your question . Thanks! –  Mar 16 '21 at 03:34
  • @TFer Thanks. This means Qualcomm SNPE doesn't support TF2.0 yet. – Arvind Mar 20 '21 at 10:52

0 Answers0