- wrap your train.py with
sagemaker.pytorch.PyTorch
and aws will train yourtrian.py
on sepcify ec2 type device - training entry point:
code/train.py
- reference pytorch-offical
code/inference.py
- deploy 的
entry_ponit
- deploy 的
- pytorch_sagemaker.ipynb
- example inculde train/deploy
- deploy from attached job(just trained): Success
- deploy from re-attached job: Fail
- wrap
sagemaker.pytorch.PyTorch
intosagemaker.tuner.HyperparameterTuner
and aws will tune yourtrain.py
on sepcify ec2 type device
- pytorch_sagemaker_tunner.ipynb
- example include Train/Tune/Analyze/Deploy
- deploy from best tunning job: Fail
- deploy from specify tunning jib: Fail