search

FlyEgle / MAE-pytorch

Masked Autoencoders Are Scalable Vision Learners
245 stars 36 forks source link

run these programs #1

Closed MingfangDeng closed 2 years ago

MingfangDeng commented 2 years ago

how to run these programs?can you tell me details? the readme is not clearly.thank you very much.

FlyEgle commented 2 years ago
  1. You need prepare the imagenet train log like my readme format.
  2. And you can run my training bash code to train the pretrain model.
  3. If you only have one gpu, you need modify the CUDA_VISIBLE_DEVICES=0 python -W ignore -m torch.distributed.launch --nproc_per_node 1 train_mae.py
  4. My FineTune and Linear-prob may train as soon as possible.
  5. I think my readme is easy for read, if you have possible may read paper first, thx.