Deci-AI
/
super-gradients
connected to https://github.com/Deci-AI/super-gradients.git


  
1

	
2

	
3

	
4

	
5

	
6

	
7

	
8

	
9

	
10

	
11

	
12

	
13

	
14

	
15

	
16

	
17

	
18

	
19

	
20

	
21

	
22

	
23

	
24

	
25

	
26

	
27

	
28

	
29

	
30

	
31

	
32

	
33

	
34

	
35

	
36

	
37

	
38

	
39

	
40

	
41

	
42

	
43

	
            #  ResNet50 Imagenet classification training:
#  This example trains with batch_size = 64 * 4 GPUs, total 256.
#  Training time on 4 X GeForce RTX 3090 Ti is 25min / epoch, total time ~ 46 hours.
#  Reach => 76.2 Top1 accuracy.
#  BE AWARE THAT THIS RECIPE USE DATA_PARALLEL, WHEN USING DDP FOR DISTRIBUTED TRAINING THIS RECIPE REACH ONLY 75.4 TOP1
#  ACCURACY.
#
#  Log and tensorboard at s3://deci-pretrained-models/resnet50_dataparallel/

# Instructions:
# running from the command line, set the PYTHONPATH environment variable: (Replace "YOUR_LOCAL_PATH" with the path to the downloaded repo):
#   export PYTHONPATH="YOUR_LOCAL_PATH"/super_gradients/
# Then:
#   python train_from_recipe_example/train_from_recipe.py --config-name=imagenet_resnet50


defaults:
  - training_hyperparams: imagenet_resnet50_train_params
  - dataset_params: imagenet_dataset_params
  - arch_params: resnet50_arch_params

dataset_interface:
  _target_: super_gradients.training.datasets.dataset_interfaces.dataset_interface.ImageNetDatasetInterface
  dataset_params: ${dataset_params}
  data_dir: /data/Imagenet

data_loader_num_workers: 8

model_checkpoints_location: local
load_checkpoint: False

experiment_name: resnet50_imagenet

multi_gpu:
  _target_: super_gradients.training.sg_model.MultiGPUMode
  value: 'DP'

deci_model:
  _target_: super_gradients.SgModel
  experiment_name: ${experiment_name}
  multi_gpu: ${multi_gpu}

architecture: resnet50