I can see Submission.csv in Output and can download the file too.File looks okay to me.\nStill I am getting \"Submission CSV Not Found\" error and not able to do the submission.\nKindly help.
", "messages": [ { "id": "615281", "postDate": "09/01/2019 18:17:02", "content": "I can see Submission.csv in Output and can download the file too.File looks okay to me.\nStill I am getting \"Submission CSV Not Found\" error and not able to do the submission.\nKindly help.
", "rawMarkdown": "I can see Submission.csv in Output and can download the file too.File looks okay to me.\nStill I am getting \"Submission CSV Not Found\" error and not able to do the submission.\nKindly help.", "votes": null }, { "id": "615344", "postDate": "09/01/2019 19:39:37", "content": "Hey!
\n\nIt doesn't help to say, but I have the same problem, except my submission file dropped off in a \"../working\" folder. I use the \"test.csv\" file, do my job and do the same trick (namely, \"results.to_csv(\"submission.csv\", index=False)\"). Other people take the \"sample_submission.csv\", obtaining a dataframe, and then convert this one to a \"submission.csv\". I read from an other discussion that this way of doing it would work. Don't see the point, stricly the same. Did not make tons of submissions, but did not have to mind that, for what I can figure out...
\n\nAlain
", "rawMarkdown": "Hey!\n\n It doesn't help to say, but I have the same problem, except my submission file dropped off in a \"../working\" folder. I use the \"test.csv\" file, do my job and do the same trick (namely, \"results.to_csv(\"submission.csv\", index=False)\"). Other people take the \"sample_submission.csv\", obtaining a dataframe, and then convert this one to a \"submission.csv\". I read from an other discussion that this way of doing it would work. Don't see the point, stricly the same. Did not make tons of submissions, but did not have to mind that, for what I can figure out...\n\nAlain", "votes": null }, { "id": "615386", "postDate": "09/01/2019 21:14:24", "content": "... As long as you see the file at the Output tab, there shouldn't have any problem. I checked the format, I have (str, int64), no problem. No typos either. I'm stuck here...
", "rawMarkdown": "... As long as you see the file at the Output tab, there shouldn't have any problem. I checked the format, I have (str, int64), no problem. No typos either. I'm stuck here...", "votes": null }, { "id": "615426", "postDate": "09/01/2019 23:24:15", "content": "We're experiencing this issue today as well I'd like to test it more but I don't want to waste submissions ☹️
", "rawMarkdown": "We're experiencing this issue today as well I'd like to test it more but I don't want to waste submissions ☹️", "votes": null }, { "id": "615578", "postDate": "09/02/2019 05:48:54", "content": "same problem, I did not meet this problem with same kernel before.
", "rawMarkdown": "same problem, I did not meet this problem with same kernel before.", "votes": null }, { "id": "615854", "postDate": "09/02/2019 12:40:39", "content": "No response from kaggle team , today again i have wasted 2 hours and not able to submit.\nAtleast Kaggle team should provide some information.
", "rawMarkdown": "No response from kaggle team , today again i have wasted 2 hours and not able to submit.\nAtleast Kaggle team should provide some information.", "votes": null }, { "id": "615863", "postDate": "09/02/2019 12:51:33", "content": "must be a problem on your side - there are many submissions during today (I also did two)
\n\ntry forking and submitting this kernel: https://www.kaggle.com/danilds/fast-submission\n(should take a few minutes to see if submission works, you will need to upload your own submission.csv of course)
", "rawMarkdown": "must be a problem on your side - there are many submissions during today (I also did two)\n\ntry forking and submitting this kernel: https://www.kaggle.com/danilds/fast-submission\n(should take a few minutes to see if submission works, you will need to upload your own submission.csv of course)", "votes": null }, { "id": "615956", "postDate": "09/02/2019 14:41:53", "content": "hi!\n Never thought that the problem was not on my side, personally. Some suggested that it could be a memory issue. I was skeptical (boo-hoo! memory problem!), then I read one comment from a \"champ\" (ranking 3d, don't remember who that was, sorry), suggesting gc.collect, del machinTruc1, machinTruc2... same result. For instance, my try is not that big, a classical ConvNet, a few layers, 256x256 resolution (well, with a great deal of training parameters, this said). I quickly catched once: \"10% above memory\"... Can't draw a solid conclusion from that -- Ding! Dong!
", "rawMarkdown": "hi!\n Never thought that the problem was not on my side, personally. Some suggested that it could be a memory issue. I was skeptical (boo-hoo! memory problem!), then I read one comment from a \"champ\" (ranking 3d, don't remember who that was, sorry), suggesting gc.collect, del machinTruc1, machinTruc2... same result. For instance, my try is not that big, a classical ConvNet, a few layers, 256x256 resolution (well, with a great deal of training parameters, this said). I quickly catched once: \"10% above memory\"... Can't draw a solid conclusion from that -- Ding! Dong!", "votes": null }, { "id": "616127", "postDate": "09/02/2019 18:02:50", "content": "Thanks Rose.\nIts working with fast-submission.I can see the score.\nDoes it mean its issue with my code?\nThis is first time I am participating in Kaggle. Is it okay to always submit with fast-submission?\nCan you please give me few more inputs how I can debug my code?
", "rawMarkdown": "Thanks Rose.\nIts working with fast-submission.I can see the score.\nDoes it mean its issue with my code?\nThis is first time I am participating in Kaggle. Is it okay to always submit with fast-submission?\nCan you please give me few more inputs how I can debug my code?", "votes": null }, { "id": "616134", "postDate": "09/02/2019 18:12:55", "content": "yes, it's very likely an issue with your code, unfortunately it's hard to debug errors on submission (since we have no logs)
\n\nif your code is working 100% correctly on commit, and you see the submission.csv file, make sure that:\n1. if you are cropping the images - make sure that you handle empty crops (i.e. after the crop you are left with image with 0x0 size)\n2. you don't use anything hardcoded - i.e. you read the image id's from the test.csv (or from the sample submission)\n3. you don't read everything at once into memory (with 13000 images you would run out of memory)\n4. you don't use too big batch size for predictions (I use 32 and it's fine)
\n\nyou can have a look at some of the notebooks - for example this one is nice: https://www.kaggle.com/ahoukang/aptos-vote\n(it handles the empty crops which might be your problem)
", "rawMarkdown": "yes, it's very likely an issue with your code, unfortunately it's hard to debug errors on submission (since we have no logs)\n\nif your code is working 100% correctly on commit, and you see the submission.csv file, make sure that:\n1. if you are cropping the images - make sure that you handle empty crops (i.e. after the crop you are left with image with 0x0 size)\n2. you don't use anything hardcoded - i.e. you read the image id's from the test.csv (or from the sample submission)\n3. you don't read everything at once into memory (with 13000 images you would run out of memory)\n4. you don't use too big batch size for predictions (I use 32 and it's fine)\n\nyou can have a look at some of the notebooks - for example this one is nice: https://www.kaggle.com/ahoukang/aptos-vote\n(it handles the empty crops which might be your problem)", "votes": null } ], "comments": [ { "id": 615344, "author_name": "geezontop", "author_url": "", "post_date": "09/01/2019 19:39:37", "content": "Hey!
\n\nIt doesn't help to say, but I have the same problem, except my submission file dropped off in a \"../working\" folder. I use the \"test.csv\" file, do my job and do the same trick (namely, \"results.to_csv(\"submission.csv\", index=False)\"). Other people take the \"sample_submission.csv\", obtaining a dataframe, and then convert this one to a \"submission.csv\". I read from an other discussion that this way of doing it would work. Don't see the point, stricly the same. Did not make tons of submissions, but did not have to mind that, for what I can figure out...
\n\nAlain
", "votes": null, "replies": [] }, { "id": 615386, "author_name": "geezontop", "author_url": "", "post_date": "09/01/2019 21:14:24", "content": "... As long as you see the file at the Output tab, there shouldn't have any problem. I checked the format, I have (str, int64), no problem. No typos either. I'm stuck here...
", "votes": null, "replies": [] }, { "id": 615426, "author_name": "lextoumbourou", "author_url": "", "post_date": "09/01/2019 23:24:15", "content": "We're experiencing this issue today as well I'd like to test it more but I don't want to waste submissions ☹️
", "votes": null, "replies": [] }, { "id": 615578, "author_name": "garybios", "author_url": "", "post_date": "09/02/2019 05:48:54", "content": "same problem, I did not meet this problem with same kernel before.
", "votes": null, "replies": [] }, { "id": 615854, "author_name": "nj2020", "author_url": "", "post_date": "09/02/2019 12:40:39", "content": "No response from kaggle team , today again i have wasted 2 hours and not able to submit.\nAtleast Kaggle team should provide some information.
", "votes": null, "replies": [ { "id": 615863, "author_name": "steelrose", "author_url": "", "post_date": "09/02/2019 12:51:33", "content": "must be a problem on your side - there are many submissions during today (I also did two)
\n\ntry forking and submitting this kernel: https://www.kaggle.com/danilds/fast-submission\n(should take a few minutes to see if submission works, you will need to upload your own submission.csv of course)
", "votes": null, "replies": [] }, { "id": 616127, "author_name": "rajnijain7782", "author_url": "", "post_date": "09/02/2019 18:02:50", "content": "Thanks Rose.\nIts working with fast-submission.I can see the score.\nDoes it mean its issue with my code?\nThis is first time I am participating in Kaggle. Is it okay to always submit with fast-submission?\nCan you please give me few more inputs how I can debug my code?
", "votes": null, "replies": [] }, { "id": 616134, "author_name": "steelrose", "author_url": "", "post_date": "09/02/2019 18:12:55", "content": "yes, it's very likely an issue with your code, unfortunately it's hard to debug errors on submission (since we have no logs)
\n\nif your code is working 100% correctly on commit, and you see the submission.csv file, make sure that:\n1. if you are cropping the images - make sure that you handle empty crops (i.e. after the crop you are left with image with 0x0 size)\n2. you don't use anything hardcoded - i.e. you read the image id's from the test.csv (or from the sample submission)\n3. you don't read everything at once into memory (with 13000 images you would run out of memory)\n4. you don't use too big batch size for predictions (I use 32 and it's fine)
\n\nyou can have a look at some of the notebooks - for example this one is nice: https://www.kaggle.com/ahoukang/aptos-vote\n(it handles the empty crops which might be your problem)
", "votes": null, "replies": [] } ] }, { "id": 615956, "author_name": "geezontop", "author_url": "", "post_date": "09/02/2019 14:41:53", "content": "hi!\n Never thought that the problem was not on my side, personally. Some suggested that it could be a memory issue. I was skeptical (boo-hoo! memory problem!), then I read one comment from a \"champ\" (ranking 3d, don't remember who that was, sorry), suggesting gc.collect, del machinTruc1, machinTruc2... same result. For instance, my try is not that big, a classical ConvNet, a few layers, 256x256 resolution (well, with a great deal of training parameters, this said). I quickly catched once: \"10% above memory\"... Can't draw a solid conclusion from that -- Ding! Dong!
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "615281": "I can see Submission.csv in Output and can download the file too.File looks okay to me.\nStill I am getting \"Submission CSV Not Found\" error and not able to do the submission.\nKindly help.", "615344": "Hey!\n\n It doesn't help to say, but I have the same problem, except my submission file dropped off in a \"../working\" folder. I use the \"test.csv\" file, do my job and do the same trick (namely, \"results.to_csv(\"submission.csv\", index=False)\"). Other people take the \"sample_submission.csv\", obtaining a dataframe, and then convert this one to a \"submission.csv\". I read from an other discussion that this way of doing it would work. Don't see the point, stricly the same. Did not make tons of submissions, but did not have to mind that, for what I can figure out...\n\nAlain", "615386": "... As long as you see the file at the Output tab, there shouldn't have any problem. I checked the format, I have (str, int64), no problem. No typos either. I'm stuck here...", "615426": "We're experiencing this issue today as well I'd like to test it more but I don't want to waste submissions ☹️", "615578": "same problem, I did not meet this problem with same kernel before.", "615854": "No response from kaggle team , today again i have wasted 2 hours and not able to submit.\nAtleast Kaggle team should provide some information.", "615863": "must be a problem on your side - there are many submissions during today (I also did two)\n\ntry forking and submitting this kernel: https://www.kaggle.com/danilds/fast-submission\n(should take a few minutes to see if submission works, you will need to upload your own submission.csv of course)", "615956": "hi!\n Never thought that the problem was not on my side, personally. Some suggested that it could be a memory issue. I was skeptical (boo-hoo! memory problem!), then I read one comment from a \"champ\" (ranking 3d, don't remember who that was, sorry), suggesting gc.collect, del machinTruc1, machinTruc2... same result. For instance, my try is not that big, a classical ConvNet, a few layers, 256x256 resolution (well, with a great deal of training parameters, this said). I quickly catched once: \"10% above memory\"... Can't draw a solid conclusion from that -- Ding! Dong!", "616127": "Thanks Rose.\nIts working with fast-submission.I can see the score.\nDoes it mean its issue with my code?\nThis is first time I am participating in Kaggle. Is it okay to always submit with fast-submission?\nCan you please give me few more inputs how I can debug my code?", "616134": "yes, it's very likely an issue with your code, unfortunately it's hard to debug errors on submission (since we have no logs)\n\nif your code is working 100% correctly on commit, and you see the submission.csv file, make sure that:\n1. if you are cropping the images - make sure that you handle empty crops (i.e. after the crop you are left with image with 0x0 size)\n2. you don't use anything hardcoded - i.e. you read the image id's from the test.csv (or from the sample submission)\n3. you don't read everything at once into memory (with 13000 images you would run out of memory)\n4. you don't use too big batch size for predictions (I use 32 and it's fine)\n\nyou can have a look at some of the notebooks - for example this one is nice: https://www.kaggle.com/ahoukang/aptos-vote\n(it handles the empty crops which might be your problem)" }, "source": "meta" }