Since competition is coming to an end. I would like pen down about paint point that almost all code competitions are associated with that is Nothing but
\n Sub Csv not found , Sub Scoring Errors.
Reason : At high level we know it is caused by OOM or Cuda Memory error but what is causing that one has to Keep Guessing If unable to replicate it with train or sample test data
\nKaggle is continuously making an improvement to its platform i appreciate that but this area continues to remain area of Big improvement . Every competition kagglers spend hale lot of time,GPU resources in debugging the issues of this sort ,also lose the submission during critical ending days.
\nCan we expect from kaggle team that they come up with some solution to this long running problem like providing option to submit in Debug Mode which mandates to not put any Print statement so one can start misusing to make Private set data probe ,if Sub error out then one can check Error stack trace that helps in determining what causes the above said issue.
\nI think this would be big relief to every one and will also save time and useful GPU resources for Kaggle.
Regards
\nJaideep…
Since competition is coming to an end. I would like pen down about paint point that almost all code competitions are associated with that is Nothing but
\n Sub Csv not found , Sub Scoring Errors.
Reason : At high level we know it is caused by OOM or Cuda Memory error but what is causing that one has to Keep Guessing If unable to replicate it with train or sample test data
\nKaggle is continuously making an improvement to its platform i appreciate that but this area continues to remain area of Big improvement . Every competition kagglers spend hale lot of time,GPU resources in debugging the issues of this sort ,also lose the submission during critical ending days.
\nCan we expect from kaggle team that they come up with some solution to this long running problem like providing option to submit in Debug Mode which mandates to not put any Print statement so one can start misusing to make Private set data probe ,if Sub error out then one can check Error stack trace that helps in determining what causes the above said issue.
\nI think this would be big relief to every one and will also save time and useful GPU resources for Kaggle.
Regards
\nJaideep…
It would have been a lot easier if the largest image had been in the public test set not the private, so if you succeed in committing you should succeed in submission.
", "rawMarkdown": "It would have been a lot easier if the largest image had been in the public test set not the private, so if you succeed in committing you should succeed in submission.", "votes": null }, { "id": "1294388", "postDate": "05/05/2021 15:51:52", "content": "I struggled too !!!!! maybe I will even give up some of my ideas due to the private set BUGS
", "rawMarkdown": "I struggled too !!!!! maybe I will even give up some of my ideas due to the private set **BUGS**", "votes": null }, { "id": "1294407", "postDate": "05/05/2021 16:10:29", "content": "I totally agree with this. Kaggle should spend some time on giving proper error message or logs for the errors which are generated when we submit. It can be a single-line message saying you ran out of RAM or ran out of Cuda memory. But there should be some message. I would really appreciate it if they will add this feature in future.
", "rawMarkdown": "I totally agree with this. Kaggle should spend some time on giving proper error message or logs for the errors which are generated when we submit. It can be a single-line message saying you ran out of RAM or ran out of Cuda memory. But there should be some message. I would really appreciate it if they will add this feature in future.", "votes": null }, { "id": "1294498", "postDate": "05/05/2021 17:33:24", "content": "@philculliton i hope plea reaches to ears of kaggle management and they ought to do some thing about that so these things dont discourage participants in participating in code Competitions .
", "rawMarkdown": "philculliton i hope plea reaches to ears of kaggle management and they ought to do some thing about that so these things dont discourage participants in participating in code Competitions .", "votes": null } ], "comments": [ { "id": 1293793, "author_name": "jamesphoward", "author_url": "", "post_date": "05/05/2021 07:05:27", "content": "It would have been a lot easier if the largest image had been in the public test set not the private, so if you succeed in committing you should succeed in submission.
", "votes": null, "replies": [] }, { "id": 1294388, "author_name": "southsakura", "author_url": "", "post_date": "05/05/2021 15:51:52", "content": "I struggled too !!!!! maybe I will even give up some of my ideas due to the private set BUGS
", "votes": null, "replies": [] }, { "id": 1294407, "author_name": "urvishp80", "author_url": "", "post_date": "05/05/2021 16:10:29", "content": "I totally agree with this. Kaggle should spend some time on giving proper error message or logs for the errors which are generated when we submit. It can be a single-line message saying you ran out of RAM or ran out of Cuda memory. But there should be some message. I would really appreciate it if they will add this feature in future.
", "votes": null, "replies": [] }, { "id": 1294498, "author_name": "jaideepvalani", "author_url": "", "post_date": "05/05/2021 17:33:24", "content": "@philculliton i hope plea reaches to ears of kaggle management and they ought to do some thing about that so these things dont discourage participants in participating in code Competitions .
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1293663": "Since competition is coming to an end. I would like pen down about paint point that almost all code competitions are associated with that is Nothing but \n Sub Csv not found , Sub Scoring Errors.\n\n**Reason : At high level we know it is caused by OOM or Cuda Memory error but what is causing that one has to Keep Guessing If unable to replicate it with train or sample test data**\n\nKaggle is continuously making an improvement to its platform i appreciate that but this area continues to remain area of Big improvement . Every competition kagglers spend hale lot of time,GPU resources in debugging the issues of this sort ,also lose the submission during critical ending days. \n\nCan we expect from kaggle team that they come up with some solution to this long running problem like providing option to submit in Debug Mode which mandates to not put any Print statement so one can start misusing to make Private set data probe ,if Sub error out then one can check Error stack trace that helps in determining what causes the above said issue. \nI think this would be big relief to every one and will also save time and useful GPU resources for Kaggle.\n\nRegards\nJaideep...", "1293793": "It would have been a lot easier if the largest image had been in the public test set not the private, so if you succeed in committing you should succeed in submission.", "1294388": "I struggled too !!!!! maybe I will even give up some of my ideas due to the private set **BUGS**", "1294407": "I totally agree with this. Kaggle should spend some time on giving proper error message or logs for the errors which are generated when we submit. It can be a single-line message saying you ran out of RAM or ran out of Cuda memory. But there should be some message. I would really appreciate it if they will add this feature in future.", "1294498": "philculliton i hope plea reaches to ears of kaggle management and they ought to do some thing about that so these things dont discourage participants in participating in code Competitions ." }, "source": "meta" }