I'm new to Kaggle Competetions and its not clear how I am supposed to submit.
\nThe competition rules require \"no TPU\" and \"no internet connection\". I searched around and found that I can train on TPU and run inference on GPU. After some efforts, I was able to load my saved model into a new notebook and run the loaded model with the test set given by the competition. I even generated a submission.csv file.
\nBut after submitting to the competition, I get the error \"Submission CSV not found\". Am I doing something wrong in loading the private test dataset ? Any help is appreciated.
I'm new to Kaggle Competetions and its not clear how I am supposed to submit.
\nThe competition rules require \"no TPU\" and \"no internet connection\". I searched around and found that I can train on TPU and run inference on GPU. After some efforts, I was able to load my saved model into a new notebook and run the loaded model with the test set given by the competition. I even generated a submission.csv file.
\nBut after submitting to the competition, I get the error \"Submission CSV not found\". Am I doing something wrong in loading the private test dataset ? Any help is appreciated.
Submission CSV not found indicate you are not save the submission table to submission.csv. btw, Private test dataset are hidden, only run in backend after you submit to the leaderboard.
\nYou can check some public notebooks which have succeed submission and have a public leaderboard.
I think I may have solved it. I was hardcoding the test path in the notebook and now I changed it to a regex (as given in the starter notebook). The uncertainty is because now it's showing me a different error \"Submission scoring error\". Any idea on what this error might be ?
", "rawMarkdown": "I think I may have solved it. I was hardcoding the test path in the notebook and now I changed it to a regex (as given in the starter notebook). The uncertainty is because now it's showing me a different error \"Submission scoring error\". Any idea on what this error might be ?", "votes": null }, { "id": "1163264", "postDate": "01/21/2021 15:58:49", "content": "@siddalore This happens when the notebook encounters an error during the re-run after you submit. Make sure your notebook does not run out of GPU memory when predicting. Also even a small syntax error can cause this error because the notebook stops halfway while executing, so the CSV file is not produced for scoring.
\nEdit: I just saw you solved that error but now face a new one. Make sure your predicted label is an integer and not float and set index to False when saving the CSV
", "rawMarkdown": "siddalore This happens when the notebook encounters an error during the re-run after you submit. Make sure your notebook does not run out of GPU memory when predicting. Also even a small syntax error can cause this error because the notebook stops halfway while executing, so the CSV file is not produced for scoring.\n\nEdit: I just saw you solved that error but now face a new one. Make sure your predicted label is an integer and not float and set index to False when saving the CSV", "votes": null }, { "id": "1163345", "postDate": "01/21/2021 16:43:22", "content": "I compared my submission file with the sample submission csv. Both have the same data types for the columns (object for image_id and int64 for label). I'm not sure what the error is
@siddalore can you tell me how you solve the problem I am facing the same problem.
", "rawMarkdown": "siddalore can you tell me how you solve the problem I am facing the same problem.", "votes": null } ], "comments": [ { "id": 1163193, "author_name": "steamedsheep", "author_url": "", "post_date": "01/21/2021 15:36:34", "content": "Submission CSV not found indicate you are not save the submission table to submission.csv. btw, Private test dataset are hidden, only run in backend after you submit to the leaderboard.
\nYou can check some public notebooks which have succeed submission and have a public leaderboard.
I think I may have solved it. I was hardcoding the test path in the notebook and now I changed it to a regex (as given in the starter notebook). The uncertainty is because now it's showing me a different error \"Submission scoring error\". Any idea on what this error might be ?
", "votes": null, "replies": [] }, { "id": 1206543, "author_name": "riyajm", "author_url": "", "post_date": "02/17/2021 12:25:09", "content": "@siddalore can you tell me how you solve the problem I am facing the same problem.
", "votes": null, "replies": [] } ] }, { "id": 1163264, "author_name": "yovinyahathugoda", "author_url": "", "post_date": "01/21/2021 15:58:49", "content": "@siddalore This happens when the notebook encounters an error during the re-run after you submit. Make sure your notebook does not run out of GPU memory when predicting. Also even a small syntax error can cause this error because the notebook stops halfway while executing, so the CSV file is not produced for scoring.
\nEdit: I just saw you solved that error but now face a new one. Make sure your predicted label is an integer and not float and set index to False when saving the CSV
", "votes": null, "replies": [ { "id": 1163345, "author_name": "siddalore", "author_url": "", "post_date": "01/21/2021 16:43:22", "content": "I compared my submission file with the sample submission csv. Both have the same data types for the columns (object for image_id and int64 for label). I'm not sure what the error is