Installment

Please install the requirement for deit and detr.

deit : https://github.com/facebookresearch/deit
detr : https://github.com/facebookresearch/detr


replicate of Container-light on ImageNet Top-1 82.1% accuracy
Go to folder, container_light

Replicate results by running the following command line:
python -m torch.distributed.launch --nproc_per_node=8 --master_port=2324 --use_env main.py --model depthwise_small --batch-size 128 --data-path  your_dataset --output_dir container_light

expected results

82.1 Top-1 for 300 epochs

Replicate of Container-DETR

Move to folder : detr

Replicate results by running the following command line:
python -m torch.distributed.launch --nproc_per_node=8 --master_port 3242 --use_env main.py --coco_path your_dataset --lr_drop 40 --epochs 50

expected results

39.0 mAP for 50 epoches

Log please check ./container_light_detr_50_epoch/log.txt

