curriculum-recommendations

1st Place Solution to the Curriculum Recommendations on Kaggle

Model Training Codes

MiniLM

• Run Model_v36mini.ipynb
• Run Model_v36mini_full.ipynb for submission

BERT

• Run Model_v36bert.ipynb
• Run Model_v36bert_full.ipynb for submission

xlm-roberta-base

• Run Model_v36.ipynb
• Run Model_v36_full.ipynb for submission

xlm-roberta-large

• Run Model_v29_large.ipynb first, then Model_v29_finetune_large.ipynb
• Run Model_v29_large_full.ipynb first, then Model_v29_finetune_large_full.ipynb for submission

Best Efficiency Submission

• Requires MiniLM to be ready
• Run Efficiency_718.ipynb

Fast Alternative to Efficiency Solution

• Run Efficiency_530.ipynb

Main Track

• Run Solution_752exp_diverse.ipynb
• Run Solution_758.ipynb
• Solution_764b.ipynb is only here for the latest postprocess function.

Submission/Inference

• Actual inference notebook (v15 selected): https://www.kaggle.com/code/aerdem4/lecr-ensemble-v03
• Best Efficiency notebook: https://www.kaggle.com/code/aerdem4/lecr-efficiency-minilm
• An alternative Efficiency solution: https://www.kaggle.com/code/aerdem4/lecr-efficiency-nobert