Wrong net architecture in 3D Classification notebook

[iarganda]

When selecting EfficientNetB0 as model, the training throws the following error:

[07:11:28.777149] Training configuration finished.
[07:11:28.786321] Date: 2024-02-05 07:11:28
[07:11:28.786428] Arguments: Namespace(config='/content/my_3d_classification.yaml', result_dir='/content/output', name='my_3d_classification', run_id=1, gpu=0, world_size=1, local_rank=-1, dist_on_itp=False, dist_url='env://', dist_backend='nccl')
[07:11:28.787806] Job: my_3d_classification_1
[07:11:28.787876] Python       : 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0]
[07:11:28.789216] PyTorch:  2.1.0+cu121
[07:11:28.790429] Not using distributed mode
---------------------------------------------------------------------------
AssertionError                            Traceback (most recent call last)
[<ipython-input-28-a88867eca5b3>](https://vyjefdubo0j-496ff2e9c6d22116-0-colab.googleusercontent.com/outputframe.html?vrz=colab_20240201-060115_RC00_603318229#) in <cell line: 116>()
    114 
    115 # Run the code
--> 116 biapy = BiaPy(f'/content/{job_name}.yaml', result_dir=output_path, name=job_name, run_id=1, gpu=0)
    117 biapy.run_job()

1 frames
[/usr/local/lib/python3.10/dist-packages/biapy/engine/check_configuration.py](https://vyjefdubo0j-496ff2e9c6d22116-0-colab.googleusercontent.com/outputframe.html?vrz=colab_20240201-060115_RC00_603318229#) in check_configuration(cfg, jobname, check_data_paths)
    506     ### Model ###
    507     if cfg.MODEL.SOURCE == "biapy":
--> 508         assert model_arch in ['unet', 'resunet', 'resunet++', 'attention_unet', 'multiresunet', 'seunet', 'simple_cnn', 'efficientnet_b0', 
    509             'efficientnet_b1', 'efficientnet_b2', 'efficientnet_b3', 'efficientnet_b4','efficientnet_b5','efficientnet_b6','efficientnet_b7',
    510             'unetr', 'edsr', 'rcan', 'dfcan', 'wdsr', 'vit', 'mae'],\

AssertionError: MODEL.ARCHITECTURE not in ['unet', 'resunet', 'resunet++', 'attention_unet', 'multiresunet', 'seunet', 'simple_cnn', 'efficientnet_b[0-7]', 'unetr', 'edsr', 'rcan', 'dfcan', 'wdsr', 'vit', 'mae']

[danifranco]

Which name did you use? Notice that it's efficientnet_b0 and not EfficientNetB0. Once you use it with that name it leads to an error loading TorchVision weights as discussed here. Will try to implement the temporal solution they propose there.

[iarganda]

OK, I will change, it was written as EfficientNetB0