hi, i had done several subbions for this competion but after the close all they appaer on error.\nIt's strange beacuse before the close i can submit ad see my score .
\n\nthanks for every one that could help me
", "messages": [ { "id": "573674", "postDate": "07/12/2019 15:43:16", "content": "hi, i had done several subbions for this competion but after the close all they appaer on error.\nIt's strange beacuse before the close i can submit ad see my score .
\n\nthanks for every one that could help me
", "rawMarkdown": "hi, i had done several subbions for this competion but after the close all they appaer on error.\nIt's strange beacuse before the close i can submit ad see my score .\n\nthanks for every one that could help me", "votes": null }, { "id": "573822", "postDate": "07/12/2019 19:55:42", "content": "Same here. This is my first competition and it feels disapointing to not get an actual final score after all work I put in.
", "rawMarkdown": "Same here. This is my first competition and it feels disapointing to not get an actual final score after all work I put in.", "votes": null }, { "id": "573859", "postDate": "07/12/2019 21:00:11", "content": "Hi,
\n\nIt was a 2 stages competition: https://www.kaggle.com/c/freesound-audio-tagging-2019/overview/timeline
\n\nwhen competition moved from phase 1 to phase 2, all submissions were set in error and the two selected submissions (if none were selected, kaggle picked the two best of stage 1) were re ran with new test dataset.
\n\nYou should have between 0 and 2 submissions without error corresponding to successful re-runs. If you have 0, then I suggest you to open your 2 re ran submissions to see why they failed. As far as I understand, common mistakes are:\n- as stage 2 test dataset was about 3x the size of stage 1 test dataset, kernel did not met kernel limitations (less than one hour run, exceeded kernel memory). \n- use of hard coded values such as audio clips ID, or number of clips\n- use a dataset of Pre processed stage 1 test dataset (e.g. Mel spectrogram of audio clips). As the second stage used another test dataset, kernel did not have the right Pre processed data for stage 2 .
", "rawMarkdown": "Hi,\n\nIt was a 2 stages competition: https://www.kaggle.com/c/freesound-audio-tagging-2019/overview/timeline\n\nwhen competition moved from phase 1 to phase 2, all submissions were set in error and the two selected submissions (if none were selected, kaggle picked the two best of stage 1) were re ran with new test dataset.\n\nYou should have between 0 and 2 submissions without error corresponding to successful re-runs. If you have 0, then I suggest you to open your 2 re ran submissions to see why they failed. As far as I understand, common mistakes are:\n- as stage 2 test dataset was about 3x the size of stage 1 test dataset, kernel did not met kernel limitations (less than one hour run, exceeded kernel memory). \n- use of hard coded values such as audio clips ID, or number of clips\n- use a dataset of Pre processed stage 1 test dataset (e.g. Mel spectrogram of audio clips). As the second stage used another test dataset, kernel did not have the right Pre processed data for stage 2 .", "votes": null } ], "comments": [ { "id": 573822, "author_name": "klirfan", "author_url": "", "post_date": "07/12/2019 19:55:42", "content": "Same here. This is my first competition and it feels disapointing to not get an actual final score after all work I put in.
", "votes": null, "replies": [] }, { "id": 573859, "author_name": "ebouteillon", "author_url": "", "post_date": "07/12/2019 21:00:11", "content": "Hi,
\n\nIt was a 2 stages competition: https://www.kaggle.com/c/freesound-audio-tagging-2019/overview/timeline
\n\nwhen competition moved from phase 1 to phase 2, all submissions were set in error and the two selected submissions (if none were selected, kaggle picked the two best of stage 1) were re ran with new test dataset.
\n\nYou should have between 0 and 2 submissions without error corresponding to successful re-runs. If you have 0, then I suggest you to open your 2 re ran submissions to see why they failed. As far as I understand, common mistakes are:\n- as stage 2 test dataset was about 3x the size of stage 1 test dataset, kernel did not met kernel limitations (less than one hour run, exceeded kernel memory). \n- use of hard coded values such as audio clips ID, or number of clips\n- use a dataset of Pre processed stage 1 test dataset (e.g. Mel spectrogram of audio clips). As the second stage used another test dataset, kernel did not have the right Pre processed data for stage 2 .
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "573674": "hi, i had done several subbions for this competion but after the close all they appaer on error.\nIt's strange beacuse before the close i can submit ad see my score .\n\nthanks for every one that could help me", "573822": "Same here. This is my first competition and it feels disapointing to not get an actual final score after all work I put in.", "573859": "Hi,\n\nIt was a 2 stages competition: https://www.kaggle.com/c/freesound-audio-tagging-2019/overview/timeline\n\nwhen competition moved from phase 1 to phase 2, all submissions were set in error and the two selected submissions (if none were selected, kaggle picked the two best of stage 1) were re ran with new test dataset.\n\nYou should have between 0 and 2 submissions without error corresponding to successful re-runs. If you have 0, then I suggest you to open your 2 re ran submissions to see why they failed. As far as I understand, common mistakes are:\n- as stage 2 test dataset was about 3x the size of stage 1 test dataset, kernel did not met kernel limitations (less than one hour run, exceeded kernel memory). \n- use of hard coded values such as audio clips ID, or number of clips\n- use a dataset of Pre processed stage 1 test dataset (e.g. Mel spectrogram of audio clips). As the second stage used another test dataset, kernel did not have the right Pre processed data for stage 2 ." }, "source": "meta" }