{ "id": 440814, "title": "How to do submission via notebook if I train my model locally?", "url": "/competitions/rsna-2023-abdominal-trauma-detection/discussion/440814", "author_name": "Shabbir Marzban", "post_date": "2023-09-16T08:29:27.323000", "votes": 0, "comment_count": 3, "views": 0, "content": "

New to kaggle here, I cant seem to figure how to make submission if I had trained the model locally.
\nDo i need to:

\n\n

Thanks

", "messages": [ { "id": 2441594, "postDate": "2023-09-16T09:53:51.047Z", "content": "

You can upload your model weights to a Dataset you create for them.
\nIf there are any additional packages needed e.g. for processing dicoms, to create the model without weights - you can check Datasets to see if what you need is there already (or check existing Notebooks that do submissions) or add your own Datasets.
\nCreate your notebook for submission - use Add Dataset to add your model weights and any others needed. Make sure internet is switched off. Do your inference on your model and create the submission.csv in a Save and Run All commit. When finished go to Output and there should be a submit button to use next to submission.csv .
\nOr use the Right panel of the notebook section Submit to Competition use Submit and it will first do the Save and Run All commit then submission in sequence. If the notebook has errors it will use up one submission though.

", "rawMarkdown": "You can upload your model weights to a Dataset you create for them. \nIf there are any additional packages needed e.g. for processing dicoms, to create the model without weights - you can check Datasets to see if what you need is there already (or check existing Notebooks that do submissions) or add your own Datasets.\nCreate your notebook for submission - use Add Dataset to add your model weights and any others needed. Make sure internet is switched off. Do your inference on your model and create the submission.csv in a Save and Run All commit. When finished go to Output and there should be a submit button to use next to submission.csv . \nOr use the Right panel of the notebook section Submit to Competition use Submit and it will first do the Save and Run All commit then submission in sequence. If the notebook has errors it will use up one submission though. ", "replies": [ { "id": 2441633, "postDate": "2023-09-16T10:44:47.573Z", "content": "

you for details :)

", "rawMarkdown": "you for details :)", "replies": [ { "id": 2441660, "postDate": "2023-09-16T11:01:17.527Z", "content": "

No worries, you had the idea for what was needed. Good Luck!

", "rawMarkdown": "No worries, you had the idea for what was needed. Good Luck!" } ] } ] }, { "id": 2441523, "postDate": "2023-09-16T08:29:27.323Z", "content": "

New to kaggle here, I cant seem to figure how to make submission if I had trained the model locally.
\nDo i need to:

\n\n

Thanks

", "rawMarkdown": "New to kaggle here, I cant seem to figure how to make submission if I had trained the model locally.\nDo i need to:\n- upload trained model (but we dont have access to internet, so as a \"dataset\" item)\n- upload neede python python packages (wheels) as \"dataset\" item\n- write submission notebook script that can load the packages and trained weights.\n\nThanks " } ], "comments": [ { "id": 2441594, "author_name": "something4kag", "author_url": "", "post_date": "2023-09-16T09:53:51.047000", "content": "

You can upload your model weights to a Dataset you create for them.
\nIf there are any additional packages needed e.g. for processing dicoms, to create the model without weights - you can check Datasets to see if what you need is there already (or check existing Notebooks that do submissions) or add your own Datasets.
\nCreate your notebook for submission - use Add Dataset to add your model weights and any others needed. Make sure internet is switched off. Do your inference on your model and create the submission.csv in a Save and Run All commit. When finished go to Output and there should be a submit button to use next to submission.csv .
\nOr use the Right panel of the notebook section Submit to Competition use Submit and it will first do the Save and Run All commit then submission in sequence. If the notebook has errors it will use up one submission though.

", "votes": 0, "replies": [ { "id": 2441633, "author_name": "Shabbir Marzban", "author_url": "", "post_date": "2023-09-16T10:44:47.573000", "content": "

you for details :)

", "votes": 0, "replies": [ { "id": 2441660, "author_name": "something4kag", "author_url": "", "post_date": "2023-09-16T11:01:17.527000", "content": "

No worries, you had the idea for what was needed. Good Luck!

", "votes": 0, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "2441594": "You can upload your model weights to a Dataset you create for them. \nIf there are any additional packages needed e.g. for processing dicoms, to create the model without weights - you can check Datasets to see if what you need is there already (or check existing Notebooks that do submissions) or add your own Datasets.\nCreate your notebook for submission - use Add Dataset to add your model weights and any others needed. Make sure internet is switched off. Do your inference on your model and create the submission.csv in a Save and Run All commit. When finished go to Output and there should be a submit button to use next to submission.csv . \nOr use the Right panel of the notebook section Submit to Competition use Submit and it will first do the Save and Run All commit then submission in sequence. If the notebook has errors it will use up one submission though. ", "2441523": "New to kaggle here, I cant seem to figure how to make submission if I had trained the model locally.\nDo i need to:\n- upload trained model (but we dont have access to internet, so as a \"dataset\" item)\n- upload neede python python packages (wheels) as \"dataset\" item\n- write submission notebook script that can load the packages and trained weights.\n\nThanks " } }