How Do I Convert an .npy File into a Dump File?

Converting the Caffe .npy File into a Dump File

After obtaining the Caffe .npy file, execute the dump_data_conversion.py script to convert it into a dump file in binary format. The command syntax is as follows:

python3 dump_data_conversion.py -type type -target target -i input_path -o output_path

• -type: (required) data type, selected from:
  • quant: quantized Caffe model data
  • tf: non-quantized TensorFlow model data
  • caffe: non-quantized Caffe model data
  • offline: offline model data
• -target: (required) target format, either numpy or dump
  • numpy: converts a dump file into a NumPy file.
  • dump: converts a NumPy file into a dump file.
• -i: (required) data folder/file path
  • Converting NumPy files into dump files:

    To pass a folder path to i, ensure that names of the files in the folder are in {op_name}.{output_index}.{timestamp}.npy format.

    If the i parameter is set to a file name, the file name must be in the {op_name}.{output_index}.{timestamp}.npy format. Only one file can be set at a time.

    op_name must comply with the A-Za-z0-9_- regular expression rule, timestamp is of 16 bits, and output_index is a digit in the range 0–9.

  • Converting dump files into NumPy files:

    The format of the folder or file name specified by the i parameter must comply with the naming conventions described in Dump File Naming Conventions.

• -o: (required) output path

  You are not advised to configure directories that are different from those of the current user to avoid privilege escalation risks.

The following is a command example for converting the .npy file into a dump file:

python3 dump_data_conversion.py -type caffe -target dump -i $HOME/caffenpyfile -o $HOME/caffedump

Converting the TensorFlow .npy File into a Dump File

After obtaining the TensorFlow .npy file, execute the dump_data_conversion.py script to convert it into a dump file in binary format. The command syntax is as follows:

python3 dump_data_conversion.py -type type -target target -i input_path -o output_path

The following is a command example for converting the .npy file into a dump file:

• -type: (required) data type, selected from:
  • quant: quantized Caffe model data
  • tf: non-quantized TensorFlow model data
  • caffe: non-quantized Caffe model data
  • offline: offline model data
• -target: (required) target format, either numpy or dump
  • numpy: converts a dump file into a NumPy file.
  • dump: converts a NumPy file into a dump file.
• -i: (required) data folder/file path
  • Converting NumPy files into dump files:

    To pass a folder path to i, ensure that names of the files in the folder are in {op_name}.{output_index}.{timestamp}.npy format.

    If the i parameter is set to a file name, the file name must be in the {op_name}.{output_index}.{timestamp}.npy format. Only one file can be set at a time.

    op_name must comply with the A-Za-z0-9_- regular expression rule, timestamp is of 16 bits, and output_index is a digit in the range 0–9.

  • Converting dump files into NumPy files:

    The format of the folder or file name specified by the i parameter must comply with the naming conventions described in Dump File Naming Conventions.

• -o: (required) output path

  You are not advised to configure directories that are different from those of the current user to avoid privilege escalation risks.

The following is a command example for converting the .npy file into a dump file:

python3 dump_data_conversion.py -type tf -target dump -i $HOME/tfnpyfile -o $HOME/tfdump