In the code requirements it says \"Freely & publicly available external data is allowed, including pre-trained models\".
\nAre there any restrictions to that? If we work in a research group that will publish some pre-trained model during the competition, could we use it? If some pre-trained model is published let's say a week before the competition ends, could we still use it? Is it \"public enough\" to consider it public?
\nIt would be nice if the rules were really clear hear as they can be interpreted very subjectively. Thanks in advance for clarification.
In the code requirements it says \"Freely & publicly available external data is allowed, including pre-trained models\".
\nAre there any restrictions to that? If we work in a research group that will publish some pre-trained model during the competition, could we use it? If some pre-trained model is published let's say a week before the competition ends, could we still use it? Is it \"public enough\" to consider it public?
\nIt would be nice if the rules were really clear hear as they can be interpreted very subjectively. Thanks in advance for clarification.
I find it very confusing to figure out what is allowed in this competition, and what is not. I have just read in the discussions that we are allowed to train models privately in other notebooks and load them just for inference. Does it mean that we can also use other resources than Kaggle notebooks to train them? I would appreciate someone officially making it clear.
", "rawMarkdown": "I find it very confusing to figure out what is allowed in this competition, and what is not. I have just read in the discussions that we are allowed to train models privately in other notebooks and load them just for inference. Does it mean that we can also use other resources than Kaggle notebooks to train them? I would appreciate someone officially making it clear.", "votes": null }, { "id": "1355983", "postDate": "06/18/2021 16:57:37", "content": "If the pretrained model and/or external data has been made publicly available at no cost to other participants, it is eligible for use. The rule on this is straightforward.
\nYou are welcome to train locally. However because this is a code competition, your submission must be made via Kaggle notebooks, and that submission notebook is subject to the constraints listed under Code Requirements. It is common in this type of a competition for people to train locally, upload those models as external data to Kaggle as a dataset, and use them as an input to their modeling. Their submission is effectively reserve to perform inference to make predictions on the test set.
", "rawMarkdown": "If the pretrained model and/or external data has been made publicly available at no cost to other participants, it is eligible for use. The rule on this is straightforward.\n\nYou are welcome to train locally. However because this is a code competition, your submission must be made via Kaggle notebooks, and that submission notebook is subject to the constraints listed under [Code Requirements](https://www.kaggle.com/c/siim-covid19-detection/overview/code-requirements). It is common in this type of a competition for people to train locally, upload those models as external data to Kaggle as a dataset, and use them as an input to their modeling. Their submission is effectively reserve to perform inference to make predictions on the test set.", "votes": null }, { "id": "1356203", "postDate": "06/18/2021 19:58:16", "content": "Thanks a lot for clarification!
", "rawMarkdown": "Thanks a lot for clarification!", "votes": null } ], "comments": [ { "id": 1355582, "author_name": "mikecho", "author_url": "", "post_date": "06/18/2021 11:45:08", "content": "I find it very confusing to figure out what is allowed in this competition, and what is not. I have just read in the discussions that we are allowed to train models privately in other notebooks and load them just for inference. Does it mean that we can also use other resources than Kaggle notebooks to train them? I would appreciate someone officially making it clear.
", "votes": null, "replies": [] }, { "id": 1355983, "author_name": "juliaelliott", "author_url": "", "post_date": "06/18/2021 16:57:37", "content": "If the pretrained model and/or external data has been made publicly available at no cost to other participants, it is eligible for use. The rule on this is straightforward.
\nYou are welcome to train locally. However because this is a code competition, your submission must be made via Kaggle notebooks, and that submission notebook is subject to the constraints listed under Code Requirements. It is common in this type of a competition for people to train locally, upload those models as external data to Kaggle as a dataset, and use them as an input to their modeling. Their submission is effectively reserve to perform inference to make predictions on the test set.
", "votes": null, "replies": [ { "id": 1356203, "author_name": "mikecho", "author_url": "", "post_date": "06/18/2021 19:58:16", "content": "Thanks a lot for clarification!
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1332826": "In the code requirements it says \"Freely & publicly available external data is allowed, including pre-trained models\".\nAre there any restrictions to that? If we work in a research group that will publish some pre-trained model during the competition, could we use it? If some pre-trained model is published let's say a week before the competition ends, could we still use it? Is it \"public enough\" to consider it public?\nIt would be nice if the rules were really clear hear as they can be interpreted very subjectively. Thanks in advance for clarification.", "1355582": "I find it very confusing to figure out what is allowed in this competition, and what is not. I have just read in the discussions that we are allowed to train models privately in other notebooks and load them just for inference. Does it mean that we can also use other resources than Kaggle notebooks to train them? I would appreciate someone officially making it clear.", "1355983": "If the pretrained model and/or external data has been made publicly available at no cost to other participants, it is eligible for use. The rule on this is straightforward.\n\nYou are welcome to train locally. However because this is a code competition, your submission must be made via Kaggle notebooks, and that submission notebook is subject to the constraints listed under [Code Requirements](https://www.kaggle.com/c/siim-covid19-detection/overview/code-requirements). It is common in this type of a competition for people to train locally, upload those models as external data to Kaggle as a dataset, and use them as an input to their modeling. Their submission is effectively reserve to perform inference to make predictions on the test set.", "1356203": "Thanks a lot for clarification!" }, "source": "meta" }