{ "id": 403055, "title": "Model Inference Example and \"Guide\"", "url": "/competitions/predict-student-performance-from-game-play/discussion/403055", "author_name": "", "post_date": "2023-04-20T23:38:56.072373500Z", "votes": 1, "comment_count": 1, "views": 0, "content": "

Hey everybody,

\n

I was struggling with loading 18 models into a a competition notebook in order to get it to score the test competition data, so I figured I would post my model inference notebook here . Hopefully this provides a good helping point for anyone else struggling with loading their models in after using Chris' 2 notebook method . Also if you are curious how I saved my models in the training environment, it is saved in the \"XGBoost Train\" section of Chris' baseline XGboost notebook here . I just made the following change (added clf.save_model(f'XGB_question{t}.xgb') ) to his notebook in the #train model part:

\n
    # TRAIN MODEL        \n    clf =  XGBClassifier(**xgb_params)\n    clf.fit(train_x[FEATURES].astype('float32'), train_y['correct'],\n            eval_set=[ (valid_x[FEATURES].astype('float32'), valid_y['correct']) ],\n            verbose=0)\n    print(f'{t}({clf.best_ntree_limit}), ',end='')\n    clf.save_model(f'XGB_question{t}.xgb')  \n
\n

Hopefully this helps some of you get the infrastructure corrected so that you guys can train and upload your own models! I essentially just split Chris' notebook into two parts so we all have an example.
\n )

", "messages": [ { "id": "2228922", "postDate": "04/20/2023 23:38:56", "content": "

Hey everybody,

\n

I was struggling with loading 18 models into a a competition notebook in order to get it to score the test competition data, so I figured I would post my model inference notebook here . Hopefully this provides a good helping point for anyone else struggling with loading their models in after using Chris' 2 notebook method . Also if you are curious how I saved my models in the training environment, it is saved in the \"XGBoost Train\" section of Chris' baseline XGboost notebook here . I just made the following change (added clf.save_model(f'XGB_question{t}.xgb') ) to his notebook in the #train model part:

\n
    # TRAIN MODEL        \n    clf =  XGBClassifier(**xgb_params)\n    clf.fit(train_x[FEATURES].astype('float32'), train_y['correct'],\n            eval_set=[ (valid_x[FEATURES].astype('float32'), valid_y['correct']) ],\n            verbose=0)\n    print(f'{t}({clf.best_ntree_limit}), ',end='')\n    clf.save_model(f'XGB_question{t}.xgb')  \n
\n

Hopefully this helps some of you get the infrastructure corrected so that you guys can train and upload your own models! I essentially just split Chris' notebook into two parts so we all have an example.
\n )

", "rawMarkdown": "Hey everybody,\n\nI was struggling with loading 18 models into a a competition notebook in order to get it to score the test competition data, so I figured I would post my model inference notebook [here](https://www.kaggle.com/code/paulmerica/model-inference) . Hopefully this provides a good helping point for anyone else struggling with loading their models in after using[ Chris' 2 notebook method](https://www.kaggle.com/competitions/predict-student-performance-from-game-play/discussion/396979) . Also if you are curious how I saved my models in the training environment, it is saved in the \"XGBoost Train\" section of Chris' baseline XGboost notebook [here](https://www.kaggle.com/code/cdeotte/xgboost-baseline-0-680) . I just made the following change (added clf.save_model(f'XGB_question{t}.xgb') ) to his notebook in the #train model part: \n\n\n # TRAIN MODEL \n clf = XGBClassifier(**xgb_params)\n clf.fit(train_x[FEATURES].astype('float32'), train_y['correct'],\n eval_set=[ (valid_x[FEATURES].astype('float32'), valid_y['correct']) ],\n verbose=0)\n print(f'{t}({clf.best_ntree_limit}), ',end='')\n clf.save_model(f'XGB_question{t}.xgb') \n\n\n\nHopefully this helps some of you get the infrastructure corrected so that you guys can train and upload your own models! I essentially just split Chris' notebook into two parts so we all have an example.\n )", "votes": null }, { "id": "2241559", "postDate": "05/01/2023 15:42:30", "content": "

18 models is too much.

", "rawMarkdown": "18 models is too much.", "votes": null } ], "comments": [ { "id": 2241559, "author_name": "littlstar123", "author_url": "", "post_date": "05/01/2023 15:42:30", "content": "

18 models is too much.

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2228922": "Hey everybody,\n\nI was struggling with loading 18 models into a a competition notebook in order to get it to score the test competition data, so I figured I would post my model inference notebook [here](https://www.kaggle.com/code/paulmerica/model-inference) . Hopefully this provides a good helping point for anyone else struggling with loading their models in after using[ Chris' 2 notebook method](https://www.kaggle.com/competitions/predict-student-performance-from-game-play/discussion/396979) . Also if you are curious how I saved my models in the training environment, it is saved in the \"XGBoost Train\" section of Chris' baseline XGboost notebook [here](https://www.kaggle.com/code/cdeotte/xgboost-baseline-0-680) . I just made the following change (added clf.save_model(f'XGB_question{t}.xgb') ) to his notebook in the #train model part: \n\n\n # TRAIN MODEL \n clf = XGBClassifier(**xgb_params)\n clf.fit(train_x[FEATURES].astype('float32'), train_y['correct'],\n eval_set=[ (valid_x[FEATURES].astype('float32'), valid_y['correct']) ],\n verbose=0)\n print(f'{t}({clf.best_ntree_limit}), ',end='')\n clf.save_model(f'XGB_question{t}.xgb') \n\n\n\nHopefully this helps some of you get the infrastructure corrected so that you guys can train and upload your own models! I essentially just split Chris' notebook into two parts so we all have an example.\n )", "2241559": "18 models is too much." }, "source": "meta" }