This is 42AI Lab Template.
This template is Research Ready.
It will allow for fast and high quality experiment, development, and results analysis.
There is still some work to do such that we can make it production ready.
For high quality inference these next steps seems wise:
- Adding ONNX support
- Docker compliant
- Fast API server
Install dependencies
# clone project
git clone https://github.com/42-AI/Template_Lab
cd Template_Lab
# Setup the repository
.42AI/init.sh
Train your model
python train.py
Train model with default configuration
# train on CPU
python train.py trainer.gpus=0
# train on GPU
python train.py trainer.gpus=1
Train model with chosen experiment configuration from [configs/experiment/](configs/experiment/)
python train.py experiment=experiment_name.yaml
You can override any parameter from command line like this
python train.py trainer.max_epochs=20 datamodule.batch_size=64
Multi run experiment
python train.py -m experiment=fashion_conv "++model.net.dropout=range(0,.5,.1)" "++model.lr=1e-2,1e-3"
Thanks ashleve/lightning-hydra-template for your great template.
There is more details on their Readme about this template usage