Skip to content

Latest commit

 

History

History

resnet50

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
LICENSE
LICENSE
 
 
README.md
README.md
 
 
valprep.sh
valprep.sh
 
 

README.md

  1. Start docker 启动命令可参考: README.md

  2. Prepare model

# download resnet50
wget https://download.openmmlab.com/mmclassification/v0/resnet/resnet50_8xb32_in1k_20210831-ea4938fc.pth

#wget https://download.openmmlab.com/mmclassification/v0/resnet/resnet50_b16x8_cifar10_20210528-f54bfad9.pth
#wget https://download.openmmlab.com/mmclassification/v0/resnet/resnet50_b16x8_cifar100_20210528-67b58a1b.pth
#wget https://download.openmmlab.com/mmclassification/v0/resnet/resnet50_8xb8_cub_20220307-57840e60.pth
  1. Prepare dataset
# You need to register an account at OpenXlab official website and login by CLI.
# install OpenXlab CLI tools
pip install -U openxlab
# log in OpenXLab
openxlab login  # [Access Key ID]: wqwrnxmr1qgovj5yppvm [Secret Access Key]: 4q8a1kj7grvjmpeggvelmvoxnw3owdz2ymbxlnxb
# download and preprocess by MIM, better to execute in $MMPreTrain directory.

# download ImageNet-1k
openxlab dataset get --dataset-repo OpenDataLab/ImageNet-1K && cd OpenDataLab___ImageNet-1K/raw/ && cat ImageNet-1K.tar.gz.000* > ImageNet-1K.tar.gz && tar -xvf ImageNet-1K.tar.gz
cp valprep.sh OpenDataLab___ImageNet-1K/raw/ImageNet-1K/val
bash valprep.sh

# download cifar-10
openxlab dataset get --dataset-repo OpenDataLab/CIFAR-10 && cd OpenDataLab___CIFAR-10/raw/ && tar -zxvf cifar-10-python.tar.gz

# download cifar-100
openxlab dataset get --dataset-repo OpenDataLab/CIFAR-100 && cd OpenDataLab___CIFAR-100/raw/ && tar -zxvf cifar-100-python.tar.gz
  1. Prepare scripts
## CentOS
yum install -y mesa-libGL
## Ubuntu
apt install -y libgl1-mesa-dev

cd package
# 目前只是编译了py310 使用驱动版本为rc3.1.0
pip install mmcv-2.1.0-cp310-cp310-linux_x86_64.whl

git clone -b 9124ebf7a285a https://github.com/open-mmlab/mmengine.git
cd mmengine
pip install -r requirements.txt
python setup.py install

git clone -b 17a886cb582 https://github.com/open-mmlab/mmpretrain.git 
cd mmpretrain
pip install -e .

vim configs/_base_/datasets/imagenet_bs32.py # 修改data_root 的值为数据所在地
vim configs/_base_/default_runtime.py        # 修改load_from 的值，放在''中
  1. Train
# single GPU
python tools/train.py configs/resnet/resnet50_8xb32_in1k.py --auto-scale-lr --amp
                                     resnet50_8xb16_cifar100.py
                                     resnet50_8xb16_cifar10.py


# Multi-GPU DDP
export MUSA_KERNEL_TIMEOUT=3600000
bash tools/dist_train.sh configs/resnet/resnet50_8xb32_in1k.py 8 --auto-scale-lr --amp
  1. Inference
python tools/test.py configs/resnet/resnet50_8xb32_in1k.py resnet50_8xb32_in1k_20210831-ea4938fc.pth
                                    resnet50_8xb16_cifar100.py
                                    resnet50_8xb16_cifar10.py