{ "id": 411346, "title": "Submitting", "url": "/competitions/birdclef-2023/discussion/411346", "author_name": "", "post_date": "2023-05-18T18:38:06.100322900Z", "votes": 1, "comment_count": 5, "views": 0, "content": "

Hi, I'm new to Kaggle and a bit confused about the submitting procedure. So from what I understand, you can train for as long as you like with GPU, then its the submission that's limited to 120 minutes. Are you supposed to train, save the model, upload to another notebook, and then submit? I have tried looking a little bit but please forgive me if this information is mentioned somewhere already. Thanks in advance

", "messages": [ { "id": "2264816", "postDate": "05/18/2023 18:38:06", "content": "

Hi, I'm new to Kaggle and a bit confused about the submitting procedure. So from what I understand, you can train for as long as you like with GPU, then its the submission that's limited to 120 minutes. Are you supposed to train, save the model, upload to another notebook, and then submit? I have tried looking a little bit but please forgive me if this information is mentioned somewhere already. Thanks in advance

", "rawMarkdown": "Hi, I'm new to Kaggle and a bit confused about the submitting procedure. So from what I understand, you can train for as long as you like with GPU, then its the submission that's limited to 120 minutes. Are you supposed to train, save the model, upload to another notebook, and then submit? I have tried looking a little bit but please forgive me if this information is mentioned somewhere already. Thanks in advance", "votes": null }, { "id": "2264986", "postDate": "05/18/2023 22:54:18", "content": "

No problem! Let me clarify the submission process for you.

\n

In Kaggle competitions, the 120-minute limit typically applies to the total runtime of your submission. This includes both training and inference time. During training, you can utilize resources like GPUs to train your model, but it's important to keep track of the time spent. Once you have trained your model and are ready to make predictions on the test dataset, you need to ensure that the inference time for your submission doesn't exceed the 120-minute limit.

\n

Here's a general outline of the submission process:

\n

Training: Develop your machine learning model in a Kaggle notebook, taking advantage of GPU resources if needed. This step involves loading and preprocessing the training data, training your model, and optimizing its parameters. Save the trained model to disk.

\n

Prediction/Inference: Create a separate notebook or section in your existing notebook dedicated to making predictions on the test data. In this notebook, load the saved trained model, preprocess the test data in the same way as the training data, and generate predictions using the trained model.

\n

Submission: Once you have generated the predictions, you need to format them correctly according to the competition's submission requirements. This usually involves saving the predictions to a CSV file or a similar format. Follow the competition guidelines on how to structure your submission file.

\n

Submitting: Go to the competition submission page, where you can upload your submission file. Make sure to review the submission instructions carefully to ensure you submit in the correct format and adhere to any additional rules or restrictions.

\n

Remember that the 120-minute limit is for the entire runtime of your submission, so you need to consider the time it takes for both training and prediction. If your training takes a considerable amount of time, it may be a good idea to save the trained model to disk and load it during the inference step to avoid repeating the training process.

\n

Always refer to the specific competition's rules and guidelines for any additional instructions or limitations regarding submissions. Good luck with your Kaggle competitions!

", "rawMarkdown": "No problem! Let me clarify the submission process for you.\n\nIn Kaggle competitions, the 120-minute limit typically applies to the total runtime of your submission. This includes both training and inference time. During training, you can utilize resources like GPUs to train your model, but it's important to keep track of the time spent. Once you have trained your model and are ready to make predictions on the test dataset, you need to ensure that the inference time for your submission doesn't exceed the 120-minute limit.\n\nHere's a general outline of the submission process:\n\nTraining: Develop your machine learning model in a Kaggle notebook, taking advantage of GPU resources if needed. This step involves loading and preprocessing the training data, training your model, and optimizing its parameters. Save the trained model to disk.\n\nPrediction/Inference: Create a separate notebook or section in your existing notebook dedicated to making predictions on the test data. In this notebook, load the saved trained model, preprocess the test data in the same way as the training data, and generate predictions using the trained model.\n\nSubmission: Once you have generated the predictions, you need to format them correctly according to the competition's submission requirements. This usually involves saving the predictions to a CSV file or a similar format. Follow the competition guidelines on how to structure your submission file.\n\nSubmitting: Go to the competition submission page, where you can upload your submission file. Make sure to review the submission instructions carefully to ensure you submit in the correct format and adhere to any additional rules or restrictions.\n\nRemember that the 120-minute limit is for the entire runtime of your submission, so you need to consider the time it takes for both training and prediction. If your training takes a considerable amount of time, it may be a good idea to save the trained model to disk and load it during the inference step to avoid repeating the training process.\n\nAlways refer to the specific competition's rules and guidelines for any additional instructions or limitations regarding submissions. Good luck with your Kaggle competitions!", "votes": null }, { "id": "2265344", "postDate": "05/19/2023 07:22:18", "content": "

Thank you for the info :)

", "rawMarkdown": "Thank you for the info :)", "votes": null }, { "id": "2265349", "postDate": "05/19/2023 07:30:25", "content": "

@ericka42
\n\"\"

", "rawMarkdown": "ericka42 \n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F12870466%2F9e6f63dfe32bf2571e2bf348a35486f6%2FScreenshot_94.jpg?generation=1684481397767533&alt=media)", "votes": null }, { "id": "2265363", "postDate": "05/19/2023 07:39:28", "content": "

i mean it did sound like gpt but it also answered the question

", "rawMarkdown": "i mean it did sound like gpt but it also answered the question", "votes": null }, { "id": "2269392", "postDate": "05/22/2023 12:53:01", "content": "

Thanks for your answer,I want to know if saving the trained model to disk and loading it during the inference step will not count as training time when submitting

", "rawMarkdown": "Thanks for your answer,I want to know if saving the trained model to disk and loading it during the inference step will not count as training time when submitting", "votes": null } ], "comments": [ { "id": 2264986, "author_name": "ericka42", "author_url": "", "post_date": "05/18/2023 22:54:18", "content": "

No problem! Let me clarify the submission process for you.

\n

In Kaggle competitions, the 120-minute limit typically applies to the total runtime of your submission. This includes both training and inference time. During training, you can utilize resources like GPUs to train your model, but it's important to keep track of the time spent. Once you have trained your model and are ready to make predictions on the test dataset, you need to ensure that the inference time for your submission doesn't exceed the 120-minute limit.

\n

Here's a general outline of the submission process:

\n

Training: Develop your machine learning model in a Kaggle notebook, taking advantage of GPU resources if needed. This step involves loading and preprocessing the training data, training your model, and optimizing its parameters. Save the trained model to disk.

\n

Prediction/Inference: Create a separate notebook or section in your existing notebook dedicated to making predictions on the test data. In this notebook, load the saved trained model, preprocess the test data in the same way as the training data, and generate predictions using the trained model.

\n

Submission: Once you have generated the predictions, you need to format them correctly according to the competition's submission requirements. This usually involves saving the predictions to a CSV file or a similar format. Follow the competition guidelines on how to structure your submission file.

\n

Submitting: Go to the competition submission page, where you can upload your submission file. Make sure to review the submission instructions carefully to ensure you submit in the correct format and adhere to any additional rules or restrictions.

\n

Remember that the 120-minute limit is for the entire runtime of your submission, so you need to consider the time it takes for both training and prediction. If your training takes a considerable amount of time, it may be a good idea to save the trained model to disk and load it during the inference step to avoid repeating the training process.

\n

Always refer to the specific competition's rules and guidelines for any additional instructions or limitations regarding submissions. Good luck with your Kaggle competitions!

", "votes": null, "replies": [ { "id": 2265344, "author_name": "alexyi0321", "author_url": "", "post_date": "05/19/2023 07:22:18", "content": "

Thank you for the info :)

", "votes": null, "replies": [] }, { "id": 2265349, "author_name": "janmpia", "author_url": "", "post_date": "05/19/2023 07:30:25", "content": "

@ericka42
\n\"\"

", "votes": null, "replies": [ { "id": 2265363, "author_name": "alexyi0321", "author_url": "", "post_date": "05/19/2023 07:39:28", "content": "

i mean it did sound like gpt but it also answered the question

", "votes": null, "replies": [] } ] }, { "id": 2269392, "author_name": "areno9516", "author_url": "", "post_date": "05/22/2023 12:53:01", "content": "

Thanks for your answer,I want to know if saving the trained model to disk and loading it during the inference step will not count as training time when submitting

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "2264816": "Hi, I'm new to Kaggle and a bit confused about the submitting procedure. So from what I understand, you can train for as long as you like with GPU, then its the submission that's limited to 120 minutes. Are you supposed to train, save the model, upload to another notebook, and then submit? I have tried looking a little bit but please forgive me if this information is mentioned somewhere already. Thanks in advance", "2264986": "No problem! Let me clarify the submission process for you.\n\nIn Kaggle competitions, the 120-minute limit typically applies to the total runtime of your submission. This includes both training and inference time. During training, you can utilize resources like GPUs to train your model, but it's important to keep track of the time spent. Once you have trained your model and are ready to make predictions on the test dataset, you need to ensure that the inference time for your submission doesn't exceed the 120-minute limit.\n\nHere's a general outline of the submission process:\n\nTraining: Develop your machine learning model in a Kaggle notebook, taking advantage of GPU resources if needed. This step involves loading and preprocessing the training data, training your model, and optimizing its parameters. Save the trained model to disk.\n\nPrediction/Inference: Create a separate notebook or section in your existing notebook dedicated to making predictions on the test data. In this notebook, load the saved trained model, preprocess the test data in the same way as the training data, and generate predictions using the trained model.\n\nSubmission: Once you have generated the predictions, you need to format them correctly according to the competition's submission requirements. This usually involves saving the predictions to a CSV file or a similar format. Follow the competition guidelines on how to structure your submission file.\n\nSubmitting: Go to the competition submission page, where you can upload your submission file. Make sure to review the submission instructions carefully to ensure you submit in the correct format and adhere to any additional rules or restrictions.\n\nRemember that the 120-minute limit is for the entire runtime of your submission, so you need to consider the time it takes for both training and prediction. If your training takes a considerable amount of time, it may be a good idea to save the trained model to disk and load it during the inference step to avoid repeating the training process.\n\nAlways refer to the specific competition's rules and guidelines for any additional instructions or limitations regarding submissions. Good luck with your Kaggle competitions!", "2265344": "Thank you for the info :)", "2265349": "ericka42 \n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F12870466%2F9e6f63dfe32bf2571e2bf348a35486f6%2FScreenshot_94.jpg?generation=1684481397767533&alt=media)", "2265363": "i mean it did sound like gpt but it also answered the question", "2269392": "Thanks for your answer,I want to know if saving the trained model to disk and loading it during the inference step will not count as training time when submitting" }, "source": "meta" }