mit-han-lab / once-for-all

[ICLR 2020] Once for All: Train One Network and Specialize it for Efficient Deployment
https://ofa.mit.edu/
MIT License
1.89k stars 333 forks source link

How to train the supernet? #64

Closed Hukongtao closed 2 years ago

Hukongtao commented 2 years ago

As written in the documentation,you can use the following command to train a OFA network:

horovodrun -np 32 -H <server1_ip>:8,<server2_ip>:8,<server3_ip>:8,<server4_ip>:8 \
    python train_ofa_net.py

But after reading the code, I feel that training a OFA network should take three steps:

python train_ofa_net.py --task kernel
python train_ofa_net.py --task depth
python train_ofa_net.py --task expand

I don't know if I understand it right

callmewenhao commented 2 months ago

i think so! 🤔