Fail to load onnx model after conversion from .etlt

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

Please provide the following information when requesting support.

• Hardware - GeForce RTX 3050
• Network Type - FaceDetect
• TLT Version - nvcr.io/nvidia/tao/tao-toolkit:5.0.0-tf1.15.5
• Training spec file (If have, please share here)
• How to reproduce the issue? (This is for errors. Please share the command line and the detailed log here.)

Steps to reproduce

  1. Download FaceDetect pruned_v2.0 from FaceDetect | NVIDIA NGC
  2. Create decode_etlt.py for .etlt to .onnx conversion
import argparse
import struct
from nvidia_tao_tf1.encoding import encoding

parser = argparse.ArgumentParser(description='ETLT Decode Tool')
parser.add_argument('-m',
                    '--model',
                    type=str,
                    required=True,
                    help='Path to the etlt file.')
parser.add_argument('-o',
                    '--uff',
                    required=True,
                    type=str,
                    help='The path to the uff file.')
parser.add_argument('-k',
                    '--key',
                    required=True,
                    type=str,
                    help='encryption key.')
args = parser.parse_args()
print(args)

with open(args.uff, 'wb') as temp_file, open(args.model, 'rb') as encoded_file:
    size = encoded_file.read(4)
    size = struct.unpack("<i", size)[0]
    input_node_name = encoded_file.read(size)
    encoding.decode(encoded_file, temp_file, args.key.encode())
print("Decode successfully.")
  1. docker run --runtime=nvidia -it --rm -v <local_dir>:<mapped_dir> nvcr.io/nvidia/tao/tao-toolkit:5.0.0-tf1.15.5 /bin/bash
  2. python decode_etlt.py -m model.etlt -o model.onnx -k nvidia_tlt
  3. Exit docker
  4. Load onnx model
import onnx

onnx_model = onnx.load("/home/tao_tutorials/model.onnx")

Error:

image
image1809×367 50 KB

3

In ngc, the facenet .etlt model is trained with detectnet_v2 network. For detectnet_v2 etlt models, after decoding, it is a .uff model instead of an onnx model.

4

Thank you for your prompt reply. I will then convert the .etlt to .uff using tao_deploy/internal/decrypt_onnx.py at main · NVIDIA/tao_deploy (github.com)
. Could you please share with me how to convert .uff to .onnx?

Thanks!

5

You can download the trainable facenet .tlt model and use the TAO 5.0 tf1 docker to export to onnx.
For example,
$ detectnet_v2 export --model /path/to/model.tlt --key nvidia_tlt --output /path/to/model.onnx