vgg model is killed

[hayoyo12]

Issue Type

Bug, Feature Request, Others

OS

Ubuntu

OS architecture

x86_64

Programming Language

Python

Framework

OpenVINO, ONNX, TensorFlow

Download URL for ONNX / OpenVINO IR

wget https://s3.amazonaws.com/onnx-model-zoo/vgg/vgg16/vgg16.onnx

Convert Script

# load onnx model
onnx_model = onnx.load(onnx_path)

# convert from onnnx to openVINO IR
source ../openvino_env/bin/activate
mo --input_model $onnx --output_dir $save_dir
deactivate

# convert from openVINO IR to TensorFlow using openvino2tensorflow
openvino2tensorflow --model_path $input_path --model_output_path $output_path --output_saved_model

Description

While converting the openVino IR file to Tensorflow using 'openvino2tensorflow', it suddenly stopped nearly at the bottom node. It does not return any log.

Please check the log images:

Relevant Log Output

70361 Killed                  openvino2tensorflow --model_path $input_path --model_output_path $output_path --output_saved_model

The number, where is placed at '# killed', keeps changing.

Source code for simple inference testing code

No response

[PINTO0309]

Perhaps it is not a problem with the specifications of this tool. Is there enough RAM in your device? It seems that TensorFlow consumes 60GB of RAM in the process of generating saved_model in the backend.

I dare to suggest that you should download VGG16 in saved_model format from the beginning.

docker run -it --rm \
-v `pwd`:/home/user/workdir \
ghcr.io/pinto0309/openvino2tensorflow:latest

wget https://s3.amazonaws.com/onnx-model-zoo/vgg/vgg16/vgg16.onnx

MODEL=vgg16
H=224
W=224

onnxsim ${MODEL}.onnx ${MODEL}_${H}x${W}.onnx

$INTEL_OPENVINO_DIR/deployment_tools/model_optimizer/mo.py \
--input_model ${MODEL}_${H}x${W}.onnx \
--data_type FP32 \
--output_dir ${MODEL}_${H}x${W}/openvino/FP32 \
--model_name ${MODEL}_${H}x${W}

openvino2tensorflow \
--model_path ${MODEL}_${H}x${W}/openvino/FP32/${MODEL}_${H}x${W}.xml \
--output_saved_model \
--output_pb \
--output_no_quant_float32_tflite \
--non_verbose

pb_to_saved_model \
--pb_file_path saved_model/model_float32.pb \
--inputs inputs:0 \
--outputs model/tf.math.add_15/Add:0

mv saved_model_from_pb/* saved_model

[hayoyo12]

@PINTO0309 I will check the RAM again. Thank you!