kaggle-curriculum-solution
kaggle-curriculum-solution copied to clipboard
psinger
Kaggle - Learning Equality - Curriculum Recommendations: 4th place solution code
Data preparation
Run the following commands to create a folder data and to download and extract the competition data:
mkdir data
cd data
kaggle competitions download -c learning-equality-curriculum-recommendations
unzip learning-equality-curriculum-recommendations.zip
To prepare the training dataframes, run the jupyter notebook prep_data_final_v1.ipynb, prep_data_final_v2.ipynb and prep_data_final_v3.ipynb.
Training
Running training is very straight forward, just run the following command:
python train.py -C yaml_accuracy/cfg_name.yaml
To train all models sequentially, run:
for cfg in yaml_accuracy/*; do python train.py -C $cfg; done
Inference & Evaluation
For inference and validation, please refer to the inference kernel on Kaggle.
Simply replace the corresponding checkpoints. All configs in this training code match the ones of the inference kernel. You can just run the inference kernel as-is without re-training, all datasets are shared.
Efficiency solution
To train all models sequentially, run:
for cfg in yaml_efficiency/*; do python train.py -C $cfg; done
psinger