For a machine with 8 NVIDIA A100 GPUs, MobileNetv2 model can be trained by running the following command from the root directory:
cvnets-train --common.config-file config/classification/mobilenetv2.yaml --common.results-loc results_imagenet1k_mobilenetv2
Note that the default location of the ImageNet-1k training and validation sets in the configuration file are /mnt/imagenet/training and /mnt/imagenet/validation respectively. Please make changes accordingly in the configuration file.
Evaluation on the ImageNet-1k dataset can be achieved using cvnets-eval command. Let us assume that MobileNetv2-1.0 model weight file name is checkpoint_ema.pt and is stored in results_imagenet1k_mobilenetv2 folder. Also, this folder should also contain config.yaml file, which is nothing but a copy of the configuration file that was used during training.
Now, we can run the following command to evaluate the performance on GPU-0:
CUDA_VISIBLE_DEVICES=0 cvnets-eval --common.config-file results_imagenet1k_mobilenetv2/config.yaml --common.results-loc results_imagenet1k_mobilenetv2 --model.classification.pretrained results_imagenet1k_mobilenetv2/checkpoint_ema_best.pt
Here are the results along with a pre-trained model on the ImageNet-K dataset.
| Model | Parameters | Top-1 | Pretrained weights | Config file |
|---|---|---|---|---|
| MobileNetv2-1.0 | 3.5 M | 73.3 | Link | Link |
Note that run-to-run variance of +/- 0.3 is natural on the ImageNet-1k dataset and could arise because of many factors, including different GPUs, seeds, etc.