Preparing .npy Data of a TensorFlow Model

Prerequisites

Prepare a .pb model (ResNet-50 is used in this example).

Procedure

1. Go to the directory where the .pb file is stored.
2. Create the dump script file tools_tensorflow_dump.py (code reference).
3. Dump data.
   Specify the .pb file, input .bin file for model inference, and input node name.

   python3 tools_tensorflow_dump.py resnet50_tensorflow_1.7.pb -i test.bin -n 'Placeholder:0' -s [1,224,224,3] -a

   

   • For details about the parameters, see Parameter Description.
   • For details about how to prepare the input .bin file for model inference, see Preparation of the Input .bin File for Model Inference.
4. Generate the result.

   The following figure shows the generated .npy file.