HuQiong
/
mae-cnn
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[upd] typo

Browse Source
main
Keyu Tian 1 year ago committed by GitHub
parent f2ffa9b9cb
commit 74552bc23e
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 1 additions and 1 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 2
      pretrain/README.md

2
pretrain/README.md
Unescape View File

@ -20,7 +20,7 @@ Then you can use `--model=your_convnet` in the pretraining script.
## Tutorial for pretraining your own dataset
See the comment of function `build_dataset_to_pretrain` in [line55 of /pretrain/utils/imagenet.py](/pretrain/utils/imagenet.py#L55) to yours.
See the comment of function `build_dataset_to_pretrain` in [line55 of /pretrain/utils/imagenet.py](/pretrain/utils/imagenet.py#L55).
Define a subclass of `torch.utils.data.Dataset` for your own unlabeled dataset, to replace our `ImageNetDataset`. Use `args.data_path` and `args.input_size` to help build your dataset, with `--data_path=... --input_size=...` to specify them.
Note the batch size `--bs` is the total batch size of all GPU, which may need to be adjusted based on your dataset size. For instance, we use `--bs=4096` for ImageNet, which contains 1.28 million images.

Write Preview
Loading…
Cancel
Save