saitej123/emlov2-session-03-saitej123

Jupyter Notebook

Lightning-Hydra-Template with CIFAR Dataset with Optuna + tensorboard

Experiment Results

Google drive Link for dvc - https://drive.google.com/drive/folders/1xV0qTuvvboYnyRspw5aeMos06WOKs5Mg?usp=sharing
Tensorboard Link for dvc - https://tensorboard.dev/experiment/WuPX9JlwRvi4IPdoqlntww/#scalars

Training stopped in-between because of CPU training (with less epochs) & failed in colab instance

Best HyperParameter from experiment

name: optuna
best_params:
   model.optimizer._target_: torch.optim.Adam
   model.optimizer.lr: 0.080057
   datamodule.batch_size: 8

DVC Steps

dvc add data
dvc add logs
dvc push -r gdrive

Link : https://drive.google.com/drive/folders/1xV0qTuvvboYnyRspw5aeMos06WOKs5Mg?usp=sharing

Hyper-parameter search Optuna

Set hyper-parameters for experiment tracking
Find the best batch_size and learning rate, and optimizer
Optimizers have to be one of Adam, SGD, RMSProp

Tensorboard Results

https://tensorboard.dev/experiment/HZTVAzfkSDC8EfVfAW99NQ/