MRI Data 3D Brain Convolution

Setup

Make a softlink for biobank

conda create -n age python=3.10 -y
conda activate age
pip install -r requirements.txt

Commands

Confusing configs.

• Note that module.load_model_ckpt and module.load_from_checkpoint is different.
  • The former only calls models weights
  • While the latter recalls every configurations including optimizers and others.

Train

HYDRA_FULL_ERROR=1 CUDA_VISIBLE_DEVICES=0 python train.py logger.name="ResNet10t-masked"

Inference

HYDRA_FULL_ERROR=1 CUDA_VISIBLE_DEVICES=0 python inference.py module.load_model_ckpt=AA.CKPT

Resume Training

HYDRA_FULL_ERROR=1 CUDA_VISIBLE_DEVICES=0 python train.py **+logger.version=HASH +logger.resume=must** module.load_from_checkpoint=AA.CKPT