FF_Only

Do You Even Need Attention? A Stack of Feed-Forward Layers Does

Surprisingly Well on ImageNet, arxiv

PaddlePaddle training/validation code and pretrained models for FF_Only.

The official pytorch implementation is here.

This implementation is developed by PaddleViT.

FF_Only Model Overview

Update

Update (2022-04-08): Code is updated. Update (2021-09-14): Code is released and ported weights are uploaded.

Models Zoo

Model	Acc@1	Acc@5	Image Size	Crop_pct	Interpolation	Link
ff_tiny	61.28	84.06	224	0.875	bicubic	google/baidu
ff_base	74.82	91.71	224	0.875	bicubic	google/baidu

*The results are evaluated on ImageNet2012 validation set.

Note: FF_Only weights are ported from here.

Data Preparation

ImageNet2012 dataset is used in the following file structure:

│imagenet/
├──train_list.txt
├──val_list.txt
├──train/
│  ├── n01440764
│  │   ├── n01440764_10026.JPEG
│  │   ├── n01440764_10027.JPEG
│  │   ├── ......
│  ├── ......
├──val/
│  ├── n01440764
│  │   ├── ILSVRC2012_val_00000293.JPEG
│  │   ├── ILSVRC2012_val_00002138.JPEG
│  │   ├── ......
│  ├── ......

• train_list.txt: list of relative paths and labels of training images. You can download it from: google/baidu
• val_list.txt: list of relative paths and labels of validation images. You can download it from: google/baidu

Usage

To use the model with pretrained weights, download the .pdparam weight file and change related file paths in the following python scripts. The model config files are located in ./configs/.

For example, assume weight file is downloaded in ./ff_only_tiny.pdparams, to use the ff_only_tiny model in python:

from config import get_config
from ffonly import build_ffonly as build_model
# config files in ./configs/
config = get_config('./configs/ff_only_tiny.yaml')
# build model
model = build_model(config)
# load pretrained weights
model_state_dict = paddle.load('./ff_only_tiny.pdparams')
model.set_state_dict(model_state_dict)

Evaluation

To evaluate model performance on ImageNet2012, run the following script using command line:

sh run_eval_multi.sh

or

CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
python main_multi_gpu.py \
-cfg='./configs/ff_only_tiny.yaml' \
-dataset='imagenet2012' \
-batch_size=256 \
-data_path='/dataset/imagenet' \
-eval \
-pretrained='./ff_only_tiny.pdparams' \
-amp

Note: if you have only 1 GPU, change device number to CUDA_VISIBLE_DEVICES=0 would run the evaluation on single GPU.

Training

To train the model on ImageNet2012, run the following script using command line:

sh run_train_multi.sh

or

CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
python main_multi_gpu.py \
-cfg='./configs/ff_only_tiny.yaml' \
-dataset='imagenet2012' \
-batch_size=256 \
-data_path='/dataset/imagenet' \
-amp

Note: it is highly recommanded to run the training using multiple GPUs / multi-node GPUs.

Reference

@article{melaskyriazi2021doyoueven,
  title={Do You Even Need Attention? A Stack of Feed-Forward Layers Does Surprisingly Well on ImageNet},
  author={Luke Melas-Kyriazi},
  journal=arxiv,
  year=2021
}