Testing MCUnet - problem with testing python script

[antragoudaras]

Hello, I am completely new to the project. I've installed all the depedencies needed for the mcunet repo given (

• Python3.7
• Pytorch 1.11
• Tensorflow 1.15
• Horovod ) What I am trying to do is to run the jobs/eval_tflite.py script, but all my attempts failed. I've tried the following. I have downloaded the IMAGENET validation dataset and extracted its contents(.JPEG pictures) into a 'val' folder needed for the --val-dir argument to be passed when invoking the _evaltflite python script. Since I am a complete newbie to this project am I missing something here? Am I supposed to pass a differnet path to the --val-dir argument, I am a little confused since I opened the eval_tflite.py and saw these lines of code (parser.add_argument('--val-dir', default='/dataset/imagenet/val', help='path to validation data')) . Also the --val-dir PATH/TO/IMAGENET/val as provided in the run instructions seems confusing to me, since I am not sure which is the suggested path. Any suggestion/help/hint will be invalueable!

--Antonios

[antragoudaras]

For anyone intrested on how to build the validation folder containing the IMAGENET pictures with labeled subfolders take a look here https://github.com/pytorch/examples/tree/main/imagenet & here also https://docs.nvidia.com/deeplearning/dali/archives/dali_08_beta/dali-developer-guide/docs/examples/pytorch/resnet50/pytorch-resnet50.html#requirements