conda create -n ivt python=3.7
conda activate ivt
conda install pytorch==1.10.0 torchvision==0.11.0 torchaudio==0.10.0 cudatoolkit=10.2 -c pytorch

git clone https://github.com/TencentYoutuResearch/PersonRetrieval-IVT.git
cd PersonRetrieval-IVT
pip install -r requirements.txt

python train_pretrain.py 

# We leverage four V-100 GPUs for training on CUHK-PEDES and ICFG-PEDES datasets.
# training with multi-gpus
sh start.sh
# or, you could also train them with a single gpu, but slow speed, maybe better performance.
python train_cuhkpedes_gpu.py
python train_icfg_gpu.py
# As RSTPReid is small, we leverage only one V-100 GPU for training.
python train_rstp.py