Hello, for anyone using **Colab **or a Local Environment. rather than the Kaggle Notebook, what's is usually the best workflow to submit predictions because this competition uses an API to generate predictions. I thought that this could work:
\nPlease let me know if this makes any sense or there is a better option, Thanks.
", "messages": [ { "id": "1634663", "postDate": "01/01/2022 01:04:08", "content": "Hello, for anyone using **Colab **or a Local Environment. rather than the Kaggle Notebook, what's is usually the best workflow to submit predictions because this competition uses an API to generate predictions. I thought that this could work:
\nPlease let me know if this makes any sense or there is a better option, Thanks.
", "rawMarkdown": "Hello, for anyone using **Colab **or a **Local Environment**. rather than the **Kaggle Notebook**, what's is usually the best workflow to submit predictions because this competition uses an API to generate predictions. I thought that this could work:\n- Train a model using a Local or External environment.\n- Save the trained model and load it to a Kaggle dataset.\n- Use the trained model to complete Inference and submit predictions.\n\nPlease let me know if this makes any sense or there is a better option, Thanks.", "votes": null }, { "id": "1634932", "postDate": "01/01/2022 10:16:49", "content": "One problem with this competition is how to deal with the additional data: basically we will get additional data that correspond to what happened during the time of the competition. If you train a model externally and submit you will miss this additional data. One option was to build the dataset and train the model last minute so that you can have most of this additional data. I believe this might get tricky, especially if you build moving average features. I think the most straightforward solution is to train the model in your submission.
", "rawMarkdown": "One problem with this competition is how to deal with the additional data: basically we will get additional data that correspond to what happened during the time of the competition. If you train a model externally and submit you will miss this additional data. One option was to build the dataset and train the model last minute so that you can have most of this additional data. I believe this might get tricky, especially if you build moving average features. I think the most straightforward solution is to train the model in your submission.", "votes": null }, { "id": "1635221", "postDate": "01/01/2022 15:47:50", "content": "The great thing about this competition is that we get two submissions, so both routes could be employed.
", "rawMarkdown": "The great thing about this competition is that we get two submissions, so both routes could be employed.", "votes": null }, { "id": "1635320", "postDate": "01/01/2022 17:01:57", "content": "Thanks for the comments, yes I will basically have to update the model every time they release new training data, I can solve this by connecting my Colab to the Kaggle API, the only problem I will need to update my Kaggle dataset every time but this can also be done with Colab, after this I will need to submit using a Kaggle Notebook. I will implement this and send an update.
\nLet me know if you think there is something better.
\nThanks.
", "rawMarkdown": "Thanks for the comments, yes I will basically have to update the model every time they release new training data, I can solve this by connecting my Colab to the Kaggle API, the only problem I will need to update my Kaggle dataset every time but this can also be done with Colab, after this I will need to submit using a Kaggle Notebook. I will implement this and send an update.\n\nLet me know if you think there is something better.\n\nThanks.", "votes": null }, { "id": "1651709", "postDate": "01/16/2022 02:00:56", "content": "@cv13j0 Be careful about your final submission. I'm pretty sure it can only access a version of a dataset that was made before the submission deadline.
\nWe could also use current data to pretrain models to save time and continue training them with new data in our final notebook.
", "rawMarkdown": "cv13j0 Be careful about your final submission. I'm pretty sure it can only access a version of a dataset that was made before the submission deadline. \n\nWe could also use current data to pretrain models to save time and continue training them with new data in our final notebook.", "votes": null }, { "id": "1651730", "postDate": "01/16/2022 02:50:46", "content": "Thanks, @chrisrichardmiles; I will consider this.
", "rawMarkdown": "Thanks, @chrisrichardmiles; I will consider this.", "votes": null } ], "comments": [ { "id": 1634932, "author_name": "lucasmorin", "author_url": "", "post_date": "01/01/2022 10:16:49", "content": "One problem with this competition is how to deal with the additional data: basically we will get additional data that correspond to what happened during the time of the competition. If you train a model externally and submit you will miss this additional data. One option was to build the dataset and train the model last minute so that you can have most of this additional data. I believe this might get tricky, especially if you build moving average features. I think the most straightforward solution is to train the model in your submission.
", "votes": null, "replies": [ { "id": 1635221, "author_name": "petersk20", "author_url": "", "post_date": "01/01/2022 15:47:50", "content": "The great thing about this competition is that we get two submissions, so both routes could be employed.
", "votes": null, "replies": [] }, { "id": 1635320, "author_name": "cv13j0", "author_url": "", "post_date": "01/01/2022 17:01:57", "content": "Thanks for the comments, yes I will basically have to update the model every time they release new training data, I can solve this by connecting my Colab to the Kaggle API, the only problem I will need to update my Kaggle dataset every time but this can also be done with Colab, after this I will need to submit using a Kaggle Notebook. I will implement this and send an update.
\nLet me know if you think there is something better.
\nThanks.
", "votes": null, "replies": [] }, { "id": 1651709, "author_name": "chrisrichardmiles", "author_url": "", "post_date": "01/16/2022 02:00:56", "content": "@cv13j0 Be careful about your final submission. I'm pretty sure it can only access a version of a dataset that was made before the submission deadline.
\nWe could also use current data to pretrain models to save time and continue training them with new data in our final notebook.
", "votes": null, "replies": [] }, { "id": 1651730, "author_name": "cv13j0", "author_url": "", "post_date": "01/16/2022 02:50:46", "content": "Thanks, @chrisrichardmiles; I will consider this.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1634663": "Hello, for anyone using **Colab **or a **Local Environment**. rather than the **Kaggle Notebook**, what's is usually the best workflow to submit predictions because this competition uses an API to generate predictions. I thought that this could work:\n- Train a model using a Local or External environment.\n- Save the trained model and load it to a Kaggle dataset.\n- Use the trained model to complete Inference and submit predictions.\n\nPlease let me know if this makes any sense or there is a better option, Thanks.", "1634932": "One problem with this competition is how to deal with the additional data: basically we will get additional data that correspond to what happened during the time of the competition. If you train a model externally and submit you will miss this additional data. One option was to build the dataset and train the model last minute so that you can have most of this additional data. I believe this might get tricky, especially if you build moving average features. I think the most straightforward solution is to train the model in your submission.", "1635221": "The great thing about this competition is that we get two submissions, so both routes could be employed.", "1635320": "Thanks for the comments, yes I will basically have to update the model every time they release new training data, I can solve this by connecting my Colab to the Kaggle API, the only problem I will need to update my Kaggle dataset every time but this can also be done with Colab, after this I will need to submit using a Kaggle Notebook. I will implement this and send an update.\n\nLet me know if you think there is something better.\n\nThanks.", "1651709": "cv13j0 Be careful about your final submission. I'm pretty sure it can only access a version of a dataset that was made before the submission deadline. \n\nWe could also use current data to pretrain models to save time and continue training them with new data in our final notebook.", "1651730": "Thanks, @chrisrichardmiles; I will consider this." }, "source": "meta" }