The kernel runs and commits successfully and generates the proper submission.csv but on submitting results to competition there's an error with \"no information about this error\" ! mark.
", "messages": [ { "id": "630792", "postDate": "09/20/2019 18:12:18", "content": "The kernel runs and commits successfully and generates the proper submission.csv but on submitting results to competition there's an error with \"no information about this error\" ! mark.
", "rawMarkdown": "The kernel runs and commits successfully and generates the proper submission.csv but on submitting results to competition there's an error with \"no information about this error\" ! mark.", "votes": null }, { "id": "630980", "postDate": "09/21/2019 05:33:38", "content": "I had the same problem, I think it is becaouse there is the actual index of the dataframe, and the 'id' column, the first one looked shifted. What I did was import the sample submission on my notebook and then substitute the prediction column with my own predictions.
", "rawMarkdown": "I had the same problem, I think it is becaouse there is the actual index of the dataframe, and the 'id' column, the first one looked shifted. What I did was import the sample submission on my notebook and then substitute the prediction column with my own predictions.", "votes": null }, { "id": "632876", "postDate": "09/24/2019 07:10:08", "content": "What do you mean by shifted? The id column starts from 0, so it's 2nd index in the DataFrame (which starts indexing from 1 not 0). I have got the same problem now, but I do not get what's wrong.
Thanks @bustam, using the sample submission as a template fixed it.
", "rawMarkdown": "Thanks @bustam, using the sample submission as a template fixed it.", "votes": null }, { "id": "633271", "postDate": "09/24/2019 16:43:17", "content": "You welcome :)
", "rawMarkdown": "You welcome :)", "votes": null }, { "id": "633272", "postDate": "09/24/2019 16:44:36", "content": "It's the resulting dataframe from predictions. The index starts with 0 and the corresponidng row is the columns names. It's weird because I think the output looks right.
", "rawMarkdown": "It's the resulting dataframe from predictions. The index starts with 0 and the corresponidng row is the columns names. It's weird because I think the output looks right.", "votes": null }, { "id": "634090", "postDate": "09/25/2019 20:21:24", "content": "Thanks @bustam I had the same problem, I did what you said above and it worked! Now to improve my score. 😁
", "rawMarkdown": "Thanks @bustam I had the same problem, I did what you said above and it worked! Now to improve my score. 😁", "votes": null }, { "id": "634249", "postDate": "09/26/2019 04:56:11", "content": "@sanjos10 you welcome!
", "rawMarkdown": "sanjos10 you welcome!", "votes": null }, { "id": "634252", "postDate": "09/26/2019 05:08:13", "content": "I didn't have to shift any indices but I made it work anyway, I don't know how though
", "rawMarkdown": "I didn't have to shift any indices but I made it work anyway, I don't know how though", "votes": null }, { "id": "634422", "postDate": "09/26/2019 09:04:26", "content": "hey its easier to just reimport and append to the dataset as I did here:\nhttps://www.kaggle.com/marcthalen/mnist-cnn-99-top-30-using-keras
", "rawMarkdown": "hey its easier to just reimport and append to the dataset as I did here:\nhttps://www.kaggle.com/marcthalen/mnist-cnn-99-top-30-using-keras", "votes": null }, { "id": "634426", "postDate": "09/26/2019 09:09:39", "content": "Indeed, I saw this approach in another topic here.
", "rawMarkdown": "Indeed, I saw this approach in another topic here.", "votes": null }, { "id": "635840", "postDate": "09/28/2019 10:32:39", "content": "Hello to everyone. Is someone still have submission error? \nI tried to replace 'label' column in sample_submission.csv with my predicted values, but it doesn't work.\nNext I tried to save 'id' column from test.csv and then use it to to create 'id' column in submission file, because I thought that test.csv that kaggle uses when evaluating your kernel has different shape, but it's not working too.
\n\nAm I doing something wrong or this is just some problems on Kaggle?
", "rawMarkdown": "Hello to everyone. Is someone still have submission error? \nI tried to replace 'label' column in sample_submission.csv with my predicted values, but it doesn't work.\nNext I tried to save 'id' column from test.csv and then use it to to create 'id' column in submission file, because I thought that test.csv that kaggle uses when evaluating your kernel has different shape, but it's not working too.\n\nAm I doing something wrong or this is just some problems on Kaggle?", "votes": null }, { "id": "636401", "postDate": "09/29/2019 13:09:53", "content": "Solved https://www.kaggle.com/c/Kannada-MNIST/discussion/110586
", "rawMarkdown": "Solved https://www.kaggle.com/c/Kannada-MNIST/discussion/110586", "votes": null }, { "id": "639304", "postDate": "10/03/2019 03:41:10", "content": "Some people might point out that this potentially comes from the dataset(test data) which our code should have tested in the kernel. Did you use CNN? I used convolutional neural network. since image data is flattened, it has to be reshaped using some reshape functions, and I explicitly typed the number of test size which might not be compatible with the other test data set we do not know what it is. Although I did not try yet, I will modify my reshape function as reshape(-1,28,28,1) where -1 means automatically fitted by the size of dataset ,and it worked my personal kernel. :)
", "rawMarkdown": "Some people might point out that this potentially comes from the dataset(test data) which our code should have tested in the kernel. Did you use CNN? I used convolutional neural network. since image data is flattened, it has to be reshaped using some reshape functions, and I explicitly typed the number of test size which might not be compatible with the other test data set we do not know what it is. Although I did not try yet, I will modify my reshape function as reshape(-1,28,28,1) where -1 means automatically fitted by the size of dataset ,and it worked my personal kernel. :)", "votes": null } ], "comments": [ { "id": 630980, "author_name": "bustam", "author_url": "", "post_date": "09/21/2019 05:33:38", "content": "I had the same problem, I think it is becaouse there is the actual index of the dataframe, and the 'id' column, the first one looked shifted. What I did was import the sample submission on my notebook and then substitute the prediction column with my own predictions.
", "votes": null, "replies": [ { "id": 632876, "author_name": "mtszkw", "author_url": "", "post_date": "09/24/2019 07:10:08", "content": "What do you mean by shifted? The id column starts from 0, so it's 2nd index in the DataFrame (which starts indexing from 1 not 0). I have got the same problem now, but I do not get what's wrong.
Thanks @bustam, using the sample submission as a template fixed it.
", "votes": null, "replies": [] }, { "id": 633271, "author_name": "bustam", "author_url": "", "post_date": "09/24/2019 16:43:17", "content": "You welcome :)
", "votes": null, "replies": [] }, { "id": 633272, "author_name": "bustam", "author_url": "", "post_date": "09/24/2019 16:44:36", "content": "It's the resulting dataframe from predictions. The index starts with 0 and the corresponidng row is the columns names. It's weird because I think the output looks right.
", "votes": null, "replies": [] }, { "id": 634090, "author_name": "sanjos10", "author_url": "", "post_date": "09/25/2019 20:21:24", "content": "Thanks @bustam I had the same problem, I did what you said above and it worked! Now to improve my score. 😁
", "votes": null, "replies": [] }, { "id": 634249, "author_name": "bustam", "author_url": "", "post_date": "09/26/2019 04:56:11", "content": "@sanjos10 you welcome!
", "votes": null, "replies": [] }, { "id": 634252, "author_name": "mtszkw", "author_url": "", "post_date": "09/26/2019 05:08:13", "content": "I didn't have to shift any indices but I made it work anyway, I don't know how though
", "votes": null, "replies": [] } ] }, { "id": 634422, "author_name": "marcthalen", "author_url": "", "post_date": "09/26/2019 09:04:26", "content": "hey its easier to just reimport and append to the dataset as I did here:\nhttps://www.kaggle.com/marcthalen/mnist-cnn-99-top-30-using-keras
", "votes": null, "replies": [ { "id": 634426, "author_name": "mtszkw", "author_url": "", "post_date": "09/26/2019 09:09:39", "content": "Indeed, I saw this approach in another topic here.
", "votes": null, "replies": [] } ] }, { "id": 635840, "author_name": "trolukovich", "author_url": "", "post_date": "09/28/2019 10:32:39", "content": "Hello to everyone. Is someone still have submission error? \nI tried to replace 'label' column in sample_submission.csv with my predicted values, but it doesn't work.\nNext I tried to save 'id' column from test.csv and then use it to to create 'id' column in submission file, because I thought that test.csv that kaggle uses when evaluating your kernel has different shape, but it's not working too.
\n\nAm I doing something wrong or this is just some problems on Kaggle?
", "votes": null, "replies": [ { "id": 636401, "author_name": "trolukovich", "author_url": "", "post_date": "09/29/2019 13:09:53", "content": "Solved https://www.kaggle.com/c/Kannada-MNIST/discussion/110586
", "votes": null, "replies": [] } ] }, { "id": 639304, "author_name": "hanjoonchoe", "author_url": "", "post_date": "10/03/2019 03:41:10", "content": "Some people might point out that this potentially comes from the dataset(test data) which our code should have tested in the kernel. Did you use CNN? I used convolutional neural network. since image data is flattened, it has to be reshaped using some reshape functions, and I explicitly typed the number of test size which might not be compatible with the other test data set we do not know what it is. Although I did not try yet, I will modify my reshape function as reshape(-1,28,28,1) where -1 means automatically fitted by the size of dataset ,and it worked my personal kernel. :)
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "630792": "The kernel runs and commits successfully and generates the proper submission.csv but on submitting results to competition there's an error with \"no information about this error\" ! mark.", "630980": "I had the same problem, I think it is becaouse there is the actual index of the dataframe, and the 'id' column, the first one looked shifted. What I did was import the sample submission on my notebook and then substitute the prediction column with my own predictions.", "632876": "What do you mean by shifted? The `id` column starts from 0, so it's 2nd index in the DataFrame (which starts indexing from 1 not 0). I have got the same problem now, but I do not get what's wrong.", "633256": "Thanks @bustam, using the sample submission as a template fixed it.", "633271": "You welcome :)", "633272": "It's the resulting dataframe from predictions. The index starts with 0 and the corresponidng row is the columns names. It's weird because I think the output looks right.", "634090": "Thanks @bustam I had the same problem, I did what you said above and it worked! Now to improve my score. 😁", "634249": "sanjos10 you welcome!", "634252": "I didn't have to shift any indices but I made it work anyway, I don't know how though", "634422": "hey its easier to just reimport and append to the dataset as I did here:\nhttps://www.kaggle.com/marcthalen/mnist-cnn-99-top-30-using-keras", "634426": "Indeed, I saw this approach in another topic here.", "635840": "Hello to everyone. Is someone still have submission error? \nI tried to replace 'label' column in sample_submission.csv with my predicted values, but it doesn't work.\nNext I tried to save 'id' column from test.csv and then use it to to create 'id' column in submission file, because I thought that test.csv that kaggle uses when evaluating your kernel has different shape, but it's not working too.\n\nAm I doing something wrong or this is just some problems on Kaggle?", "636401": "Solved https://www.kaggle.com/c/Kannada-MNIST/discussion/110586", "639304": "Some people might point out that this potentially comes from the dataset(test data) which our code should have tested in the kernel. Did you use CNN? I used convolutional neural network. since image data is flattened, it has to be reshaped using some reshape functions, and I explicitly typed the number of test size which might not be compatible with the other test data set we do not know what it is. Although I did not try yet, I will modify my reshape function as reshape(-1,28,28,1) where -1 means automatically fitted by the size of dataset ,and it worked my personal kernel. :)" }, "source": "meta" }