Template for my personal Experiment Tracking hyperparameters
This project base on this NeptuneAI's blog about the Experiment Tracking
Using Hydra for reading config files and WANDB for tracking experiments
This will help reduce repeated taskes and tracking your experiment more effective.
python demo_hydra.py
Output:
project: ORGANIZATION/experiment-tracking
name: experiment-tracking-default-risk
wandb:
WANDB_API_KEY: YOUR_WANDB_API
entity: YOUR_WANDB_API
project: experiment-tracking
name: TEST
parameters:
n_cv_splits: 5
validation_size: 0.2
stratified_cv: true
shuffle: 1
rf__n_estimators: 2000
rf__criterion: gini
rf__max_depth: 40
rf__class_weight: balanced
rf__max_features: 0.3
You can change your hyperparams in cli like this
python demo_hydra.py parameters.n_cv_splits=10 parameters.validation_size=0.9