Make a softlink for biobank
conda create -n age python=3.10 -y
conda activate age
pip install -r requirements.txt
Confusing configs.
- Note that
module.load_model_ckpt
andmodule.load_from_checkpoint
is different.- The former only calls models weights
- While the latter recalls every configurations including optimizers and others.
HYDRA_FULL_ERROR=1 CUDA_VISIBLE_DEVICES=0 python train.py logger.name="ResNet10t-masked"
HYDRA_FULL_ERROR=1 CUDA_VISIBLE_DEVICES=0 python inference.py module.load_model_ckpt=AA.CKPT
HYDRA_FULL_ERROR=1 CUDA_VISIBLE_DEVICES=0 python train.py **+logger.version=HASH +logger.resume=must** module.load_from_checkpoint=AA.CKPT