yolov8s模型转换成dla

wrzeng

搭建环境：
1:conda create -n yolov8 python=3.7
2:conda activate yolov8

# 2. 安装 YOLOv8
git clone https://github.com/ultralytics/yolov5.git
cd yolov5
git reset --hard 485da42

pip3 install -r requirements.txt
pip3 install torch==1.9.0 torchvision==0.10.0
pip3 install  ultralytics==8.0.151
wget https://github.com/ultralytics/a ... d/v8.2.0/yolov8s.pt
export.py里 export_torchscript  
  ts = torch.jit.trace(model, im, strict=False) 改成ts = torch.jit.trace(model, im, strict=False,check_trace=False)

python3 export.py --weights yolov8s.pt --img-size 640 640 --include torchscript

python3 convert_to_tflite_quantized.py

export LD_LIBRARY_PATH=/home/ubuntu/neuropilot/neuropilot-sdk-basic-8.0.11-build20260211/neuron_sdk/host/lib/ （你的具体文件路径）
../neuropilot-sdk-basic-8.0.11-build20260211/neuron_sdk/host/bin/ncc-tflite --arch=mdla5.3,edma3.6 yolov8s_quant.tflite （你的具体文件路径）

wrzeng

通过网盘分享的文件：neuropilot-sdk-basic-8.0.11-build20260211.tar.gz
链接: https://pan.baidu.com/s/1-CG-UuGddDc4Q3y8w_iR0A 提取码: gpsq

转换脚本如附件：convert_to_tflite_quantized.py

wrzeng

总结就如下4步
1：download model
2：Export the PyTorch model to TorchScript
3:Convert to TFLite format
4:Convert to DLA format