super-gradients/src/super_gradients/recipes/imagenet_repvgg.yaml

#  RepVGGA0 Imagenet classification training:
#  This example trains with batch_size = 64 * 4 GPUs, total 256.
#  Training time on 4 X GeForce RTX 3090 Ti is 10min / epoch, total time ~ 20h 22m (DistributedDataParallel).
#  Reach => 72.05 Top1 accuracy.
#
#  Log and tensorboard at s3://deci-pretrained-models/repvggg-a0-imagenet-tensorboard/
#
# Instructions:
#   0. Make sure that the data is stored in dataset_params.dataset_dir or add "dataset_params.data_dir=<PATH-TO-DATASET>" at the end of the command below (feel free to check ReadMe)
#   1. Move to the project root (where you will find the ReadMe and src folder)
#   2. Run the command:
#       python src/super_gradients/examples/train_from_recipe_example/train_from_recipe.py --config-name=imagenet_repvgg

defaults:
  - training_hyperparams: imagenet_repvgg_train_params
  - dataset_params: imagenet_dataset_params
  - arch_params: repvgg_arch_params
  - checkpoint_params: default_checkpoint_params
  - _self_

arch_params:
  num_classes: 1000
  build_residual_branches: True

train_dataloader: imagenet_train
val_dataloader: imagenet_val



resume: False
training_hyperparams:
  resume: ${resume}

experiment_name: repvgg_a0_imagenet_reproduce_fix

multi_gpu: DDP
num_gpus: 4

architecture: repvgg_a0
ckpt_root_dir:

# THE FOLLOWING PARAMS ARE DIRECTLY USED BY HYDRA
hydra:
  run:
    # Set the output directory (i.e. where .hydra folder that logs all the input params will be generated)
    dir: ${hydra_output_dir:${ckpt_root_dir}, ${experiment_name}}