"FrontEnd API failed with InitializationFailure" when converting YOLOv8 from Ultralytics

[msmiatac]

When running a script to convert the YoloV8 model from Ultralytics (https://huggingface.co/Ultralytics/YOLOv8/resolve/main/yolov8n.pt), I encounter the following error on one of the machines:

OpenVINO: export failure ❌ 10.8s: Check 'false' failed at src/frontends/common/src/manager.cpp:64:
FrontEnd API failed with InitializationFailure:
FrontEnd for Framework pytorch is not found

However, on other machines with seemingly identical configurations, the conversion completes successfully. Has anyone encountered this issue or have any suggestions on how to resolve it?

Here is the conversion script:

#!/bin/bash
#Set the name of the virtual environment directory
VENV_DIR="yolo_venv"

#Create a Python virtual environment if it doesn't exist
if [ ! -d "$VENV_DIR" ]; then
echo "Creating virtual environment in $VENV_DIR..."
python3 -m venv "$VENV_DIR"
fi

#Activate the virtual environment
source "$VENV_DIR/bin/activate"

#Upgrade pip in the virtual environment
pip install --upgrade pip

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu

pip install openvino
pip install ultralytics

#Function to export YOLO model
export_yolo_model() {
local weights_path=$1 # User-provided weights path

if [ ! -f "$weights_path" ]; then
  echo "Weights file not found: ${weights_path}"
  return 1
fi

echo "Using weights from: ${weights_path}"

python3 - <<EOF "$weights_path"

import sys
from ultralytics import YOLO
import openvino

weights_path = sys.argv[1]
model_type = "YOLOv8"

try:
model = YOLO(weights_path)
model.info()
converted_path = model.export(format='openvino')

except Exception as e:
print(f"An error occurred during model conversion: {e}")
EOF
}

#Check if a weights path is provided as an argument
if [ $# -eq 0 ]; then
echo "Usage: $0 <path_to_weights_file>"
exit 1
fi

#Call the function with the user-provided weights path
export_yolo_model "$1"