![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2236338%2Fc6b074349bf3276d57c082834da95f49%2F2019-07-02_165215.jpg?generation=1562082805042767&alt=media\")
Anyone else got this problem? It's been trying to calculate the public score on a submission for about two hours now.
", "messages": [ { "id": 565030, "postDate": "2019-06-30T10:21:51.443Z", "content": "Anyone else got this problem? It's been trying to calculate the public score on a submission for about two hours now.
", "rawMarkdown": "Anyone else got this problem? It's been trying to calculate the public score on a submission for about two hours now.", "votes": 8 }, { "id": 565238, "postDate": "2019-06-30T16:44:24.573Z", "content": "My Kernel commit took ~30 min, but after submitting it took ~5 hours! The reason can be found under Data Description,
\n\n\n", "rawMarkdown": "My Kernel commit took ~30 min, but after submitting it took ~5 hours! The reason can be found under Data Description,\n> When Kaggle runs your Kernel privately, it substitutes the private test set and sample submission in place of the public ones. You can plan on the private test set consisting of 20GB of data across 13,000 images (approximately).", "votes": 3, "replies": [ { "id": 565248, "postDate": "2019-06-30T16:57:15.027Z", "content": "When Kaggle runs your Kernel privately, it substitutes the private test set and sample submission in place of the public ones. You can plan on the private test set consisting of 20GB of data across 13,000 images (approximately).
\n
Thanks:) Still though, been running for 7 hours now...
", "rawMarkdown": "Thanks:) Still though, been running for 7 hours now..." } ] }, { "id": 565894, "postDate": "2019-07-01T14:04:00.707Z", "content": "Private test image is around 20GB while public is about 1.5 GB. So expect around 10 times time it will take for Private inference. Better is to make a separate inference kernel by uploading or selecting model from your training notebook.
", "rawMarkdown": "Private test image is around 20GB while public is about 1.5 GB. So expect around 10 times time it will take for Private inference. Better is to make a separate inference kernel by uploading or selecting model from your training notebook.", "votes": 1, "replies": [ { "id": 566304, "postDate": "2019-07-02T03:06:44.987Z", "content": "Thanks for this info!
", "rawMarkdown": "Thanks for this info!" }, { "id": 566796, "postDate": "2019-07-02T15:53:27.640Z", "content": "In this kernel, I just do inference using 5 pretrained models.
\n\n
Try reducing batch size and use batch-wise prediciton. Also delete the model once prediction is done. Clear GPU Cache. Then load next model. Maybe GPU mem and RAM is overflowing.
", "rawMarkdown": "Try reducing batch size and use batch-wise prediciton. Also delete the model once prediction is done. Clear GPU Cache. Then load next model. Maybe GPU mem and RAM is overflowing.", "votes": 1 } ] }, { "id": 565220, "postDate": "2019-06-30T16:03:26.277Z", "content": "Same here, no idea what is going on.
", "rawMarkdown": "Same here, no idea what is going on.", "votes": 1 }, { "id": 574337, "postDate": "2019-07-13T16:38:02.407Z", "content": "Someone raise an alarm. Feels like 3 years have passed by now.
", "rawMarkdown": "Someone raise an alarm. Feels like 3 years have passed by now.", "votes": 2 }, { "id": 565367, "postDate": "2019-06-30T20:32:50.460Z", "content": "I don't know why they don't score on the public test data only to give immediate results to competitors, then run the whole kernel in the background and provide a note later on whether the submission was a valid one.
", "rawMarkdown": "I don't know why they don't score on the public test data only to give immediate results to competitors, then run the whole kernel in the background and provide a note later on whether the submission was a valid one.", "votes": 2 }, { "id": 569508, "postDate": "2019-07-06T19:20:35.490Z", "content": "Hi, @mariusbrataas same here, probably around three or more hours,
", "rawMarkdown": "Hi, @mariusbrataas same here, probably around three or more hours," }, { "id": 568002, "postDate": "2019-07-04T08:50:24.437Z", "content": "Hi Marius,
\n\nI had the same issue and I believe that it has something to do with the resource allocation and the total submissions that are being calculated at the given time. I got a commit in under 15 minutes and the submission took around 25 minutes. So try to submit at a different time and see if you find a better performance.
", "rawMarkdown": "Hi Marius,\n\nI had the same issue and I believe that it has something to do with the resource allocation and the total submissions that are being calculated at the given time. I got a commit in under 15 minutes and the submission took around 25 minutes. So try to submit at a different time and see if you find a better performance." }, { "id": 566230, "postDate": "2019-07-01T23:51:34.460Z", "content": "I'm facing the same problem. calculation of public score takes a lot of time.
", "rawMarkdown": "I'm facing the same problem. calculation of **public score** takes a lot of time." }, { "id": 565354, "postDate": "2019-06-30T20:06:32.017Z", "content": "Having the same problem ! Is it re-running the kernel on submission ?
", "rawMarkdown": "Having the same problem ! Is it re-running the kernel on submission ?", "replies": [ { "id": 565896, "postDate": "2019-07-01T14:07:26.840Z", "content": "Yes. Its a Synchronous Kernel competition which reruns on full private test.
", "rawMarkdown": "Yes. Its a Synchronous Kernel competition which reruns on full private test." }, { "id": 565927, "postDate": "2019-07-01T15:08:12.877Z", "content": "Thanks !
", "rawMarkdown": "Thanks ! " } ] }, { "id": 565295, "postDate": "2019-06-30T18:02:54.287Z", "content": "Yes and I think it is longer today for some reason. With more or less same kernel I had public result ready within 1 hour yesterday, today it is running for about 3 hours and still is not finished.
", "rawMarkdown": "Yes and I think it is longer today for some reason. With more or less same kernel I had public result ready within 1 hour yesterday, today it is running for about 3 hours and still is not finished." }, { "id": 565103, "postDate": "2019-06-30T12:45:48.080Z", "content": "Same here, it is already over an hour and still going on, but the commit only took 30 minutes. Edit: It did got finish
", "rawMarkdown": "Same here, it is already over an hour and still going on, but the commit only took 30 minutes. Edit: It did got finish" }, { "id": 595287, "postDate": "2019-08-09T04:08:01.610Z", "content": "Same question!
", "rawMarkdown": "Same question!", "isDeleted": true }, { "id": 566066, "postDate": "2019-07-01T19:01:09.907Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 565238, "author_name": "Kiran Kunapuli", "author_url": "", "post_date": "2019-06-30T16:44:24.573000", "content": "My Kernel commit took ~30 min, but after submitting it took ~5 hours! The reason can be found under Data Description,
\n\n\n", "votes": 3, "replies": [ { "id": 565248, "author_name": "Marius Brataas", "author_url": "", "post_date": "2019-06-30T16:57:15.027000", "content": "When Kaggle runs your Kernel privately, it substitutes the private test set and sample submission in place of the public ones. You can plan on the private test set consisting of 20GB of data across 13,000 images (approximately).
\n
Thanks:) Still though, been running for 7 hours now...
", "votes": 0, "replies": [] } ] }, { "id": 565894, "author_name": "Murlikrishnan Tripathi", "author_url": "", "post_date": "2019-07-01T14:04:00.707000", "content": "Private test image is around 20GB while public is about 1.5 GB. So expect around 10 times time it will take for Private inference. Better is to make a separate inference kernel by uploading or selecting model from your training notebook.
", "votes": 1, "replies": [ { "id": 566304, "author_name": "jason", "author_url": "", "post_date": "2019-07-02T03:06:44.987000", "content": "Thanks for this info!
", "votes": 0, "replies": [] }, { "id": 566796, "author_name": "JM100", "author_url": "", "post_date": "2019-07-02T15:53:27.640000", "content": "In this kernel, I just do inference using 5 pretrained models.
\n\n
Try reducing batch size and use batch-wise prediciton. Also delete the model once prediction is done. Clear GPU Cache. Then load next model. Maybe GPU mem and RAM is overflowing.
", "votes": 1, "replies": [] } ] }, { "id": 565220, "author_name": "Carlo", "author_url": "", "post_date": "2019-06-30T16:03:26.277000", "content": "Same here, no idea what is going on.
", "votes": 1, "replies": [] }, { "id": 574337, "author_name": "torch", "author_url": "", "post_date": "2019-07-13T16:38:02.407000", "content": "Someone raise an alarm. Feels like 3 years have passed by now.
", "votes": 2, "replies": [] }, { "id": 565367, "author_name": "Darryl", "author_url": "", "post_date": "2019-06-30T20:32:50.460000", "content": "I don't know why they don't score on the public test data only to give immediate results to competitors, then run the whole kernel in the background and provide a note later on whether the submission was a valid one.
", "votes": 2, "replies": [] }, { "id": 569508, "author_name": "C4rl05/V", "author_url": "", "post_date": "2019-07-06T19:20:35.490000", "content": "Hi, @mariusbrataas same here, probably around three or more hours,
", "votes": 0, "replies": [] }, { "id": 568002, "author_name": "Rishabh Sharma", "author_url": "", "post_date": "2019-07-04T08:50:24.437000", "content": "Hi Marius,
\n\nI had the same issue and I believe that it has something to do with the resource allocation and the total submissions that are being calculated at the given time. I got a commit in under 15 minutes and the submission took around 25 minutes. So try to submit at a different time and see if you find a better performance.
", "votes": 0, "replies": [] }, { "id": 566230, "author_name": "JM100", "author_url": "", "post_date": "2019-07-01T23:51:34.460000", "content": "I'm facing the same problem. calculation of public score takes a lot of time.
", "votes": 0, "replies": [] }, { "id": 565354, "author_name": "Filemon", "author_url": "", "post_date": "2019-06-30T20:06:32.017000", "content": "Having the same problem ! Is it re-running the kernel on submission ?
", "votes": 0, "replies": [ { "id": 565896, "author_name": "Murlikrishnan Tripathi", "author_url": "", "post_date": "2019-07-01T14:07:26.840000", "content": "Yes. Its a Synchronous Kernel competition which reruns on full private test.
", "votes": 0, "replies": [] }, { "id": 565927, "author_name": "Filemon", "author_url": "", "post_date": "2019-07-01T15:08:12.877000", "content": "Thanks !
", "votes": 0, "replies": [] } ] }, { "id": 565295, "author_name": "Anna Novikova", "author_url": "", "post_date": "2019-06-30T18:02:54.287000", "content": "Yes and I think it is longer today for some reason. With more or less same kernel I had public result ready within 1 hour yesterday, today it is running for about 3 hours and still is not finished.
", "votes": 0, "replies": [] }, { "id": 565103, "author_name": "Cold", "author_url": "", "post_date": "2019-06-30T12:45:48.080000", "content": "Same here, it is already over an hour and still going on, but the commit only took 30 minutes. Edit: It did got finish
", "votes": 0, "replies": [] }, { "id": 595287, "author_name": "", "author_url": "", "post_date": "2019-08-09T04:08:01.610000", "content": "Same question!
", "votes": 0, "replies": [] }, { "id": 566066, "author_name": "", "author_url": "", "post_date": "2019-07-01T19:01:09.907000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "565030": "Anyone else got this problem? It's been trying to calculate the public score on a submission for about two hours now.", "565238": "My Kernel commit took ~30 min, but after submitting it took ~5 hours! The reason can be found under Data Description,\n> When Kaggle runs your Kernel privately, it substitutes the private test set and sample submission in place of the public ones. You can plan on the private test set consisting of 20GB of data across 13,000 images (approximately).", "565894": "Private test image is around 20GB while public is about 1.5 GB. So expect around 10 times time it will take for Private inference. Better is to make a separate inference kernel by uploading or selecting model from your training notebook.", "565220": "Same here, no idea what is going on.", "574337": "Someone raise an alarm. Feels like 3 years have passed by now.", "565367": "I don't know why they don't score on the public test data only to give immediate results to competitors, then run the whole kernel in the background and provide a note later on whether the submission was a valid one.", "569508": "Hi, @mariusbrataas same here, probably around three or more hours,", "568002": "Hi Marius,\n\nI had the same issue and I believe that it has something to do with the resource allocation and the total submissions that are being calculated at the given time. I got a commit in under 15 minutes and the submission took around 25 minutes. So try to submit at a different time and see if you find a better performance.", "566230": "I'm facing the same problem. calculation of **public score** takes a lot of time.", "565354": "Having the same problem ! Is it re-running the kernel on submission ?", "565295": "Yes and I think it is longer today for some reason. With more or less same kernel I had public result ready within 1 hour yesterday, today it is running for about 3 hours and still is not finished.", "565103": "Same here, it is already over an hour and still going on, but the commit only took 30 minutes. Edit: It did got finish", "595287": "Same question!", "566066": "" } }