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

#  ViT Imagenet1K fine tuning from Imagenet21K classification training:
#  This example trains with batch_size = 32 * 8 GPUs, total 256.
#  Training time on 8 x GeForce RTX A5000 is 52min / epoch.
#  ViT Large : 85.64 (Final averaged model)
#
#  Log and tensorboard at s3://deci-pretrained-models/vit_large_cutmix_randaug_v2_lr=0.03/

# 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_vit_large


defaults:
  - imagenet_vit_base
  - _self_

dataset_params:
  train_dataloader_params:
    batch_size: 32

training_hyperparams:
  initial_lr: 0.06
  average_best_models: True

architecture: vit_large

experiment_name: vit_large_imagenet1k
multi_gpu: DDP
num_gpus: 8
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}}