Classification pipeline for Kaggle Pogchamp Series competition built using TF/Keras and W&B.
Download dataset:
kaggle competitions download -c kaggle-pog-series-s01e03
python train.py --config configs/baseline --wandb --log_model
The training and validation metrics will be logged to Weights and Biases. You can find my training logs in this Weights and Biases workspace: https://wandb.ai/ayush-thakur/pogchamp
The model checkpoints will be logged to the same and you can visit the W&B run page > artifacts to select the model you want to use for creating submission.csv
.
The <path/to/model/artifact>
is the Full Name shown in the artifact page screenshot below:
python test.py --config configs/baseline --model_artifact_path <path/to/model/artifact>
This will write the predictions to a submission.py
file. Download it and upload it for submission.