Long training Time

[Alam4545]

I am trying to train swin_nano with 4 V100 GPUs. It's almost 20hrs but have not completed one epoch yet. I have followed the setup instructions stated in this repo. My setup is as foliows:
Package Version

---

certifi 2022.6.15
charset-normalizer 2.1.0
cycler 0.11.0
einops 0.4.1
fonttools 4.33.3
idna 3.3
kiwisolver 1.4.3
matplotlib 3.5.2
MultiScaleDeformableAttention 1.0
numpy 1.21.6
onnx 1.10.0
onnxruntime 1.4.0
opencv-python 4.1.1.26
packaging 21.3
Pillow 9.2.0
pip 19.0.3
protobuf 3.20.1
pycocotools 2.0.4
pyparsing 3.0.9
python-dateutil 2.8.2
requests 2.28.1
scipy 1.7.3
setuptools 40.8.0
six 1.16.0
timm 0.5.4
torch 1.8.0+cu111
torchaudio 0.8.0
torchvision 0.9.0+cu111
typing-extensions 4.3.0
urllib3 1.26.9

With the same setup DeformableDETR takes 1hr and 30 mins to complete one epoch on COCO 2017 dataset.
Could anyone identify the problem?