I trained a model and get a submission file on my own computer. Then I make a new kernel on kaggle, read my file and save file as \"submission.csv\", commit is also ok, but submission always get an error? Submitting \"submission\" file only is not allowed?
", "messages": [ { "id": "838041", "postDate": "05/08/2020 08:45:55", "content": "I trained a model and get a submission file on my own computer. Then I make a new kernel on kaggle, read my file and save file as \"submission.csv\", commit is also ok, but submission always get an error? Submitting \"submission\" file only is not allowed?
", "rawMarkdown": "I trained a model and get a submission file on my own computer. Then I make a new kernel on kaggle, read my file and save file as \"submission.csv\", commit is also ok, but submission always get an error? Submitting \"submission\" file only is not allowed?", "votes": null }, { "id": "838518", "postDate": "05/08/2020 16:08:46", "content": "Well what is the submission error? The CSV isn't found? The kernel crashes?
", "rawMarkdown": "Well what is the submission error? The CSV isn't found? The kernel crashes?", "votes": null }, { "id": "838541", "postDate": "05/08/2020 16:26:17", "content": "That doesn't work for this competition because there is a hidden test set. You should predict in your kernel.
", "rawMarkdown": "That doesn't work for this competition because there is a hidden test set. You should predict in your kernel.", "votes": null }, { "id": "838682", "postDate": "05/08/2020 18:04:49", "content": "I encountered the same issue. When I submit sample_submission.csv (write it as submission.csv) as dummy, I can successfully submit it. But when I upload my submission from \"add data\" and write it as submission.csv, it doesn't work. I am not sure whether the problem is caused by the fact I don't make any model prediction on test set locally in my submission kernel.\n@bamps53 could you explain a bit on hidden test set? It seems the test set (for public LB) only has 10 samples which we can access. And when I saw the kernel notebooks from others, their submission.csv also only contain those 10 samples. \n@matthewmasters the error I received was Submission Scoring Error
", "rawMarkdown": "I encountered the same issue. When I submit sample_submission.csv (write it as submission.csv) as dummy, I can successfully submit it. But when I upload my submission from \"add data\" and write it as submission.csv, it doesn't work. I am not sure whether the problem is caused by the fact I don't make any model prediction on test set locally in my submission kernel.\n@bamps53 could you explain a bit on hidden test set? It seems the test set (for public LB) only has 10 samples which we can access. And when I saw the kernel notebooks from others, their submission.csv also only contain those 10 samples. \n@matthewmasters the error I received was Submission Scoring Error", "votes": null }, { "id": "838757", "postDate": "05/08/2020 19:10:14", "content": "you should debug your inference kernel. you can use try/except for searching bug. I recommend start from your main prediction method for one sample, for example:
\n\npython\nfor image_id, image in dataset:\n try:\n ... = make_my_prediction(image)\n except Exception:\n results.append({\n 'image_id': image_id,\n 'PredictionString': ''\n })\n continue\n
If your submission is success after this, your should search bug into \"make my prediction\" method or continue search anywhere
\n\nHere you can see successful inference kernel, I hope it helps you!
\n\nWelcome!
", "rawMarkdown": "you should debug your inference kernel. you can use try/except for searching bug. I recommend start from your main prediction method for one sample, for example:\n\n```python\nfor image_id, image in dataset:\n try:\n ... = make_my_prediction(image)\n except Exception:\n results.append({\n 'image_id': image_id,\n 'PredictionString': ''\n })\n continue\n```\n\nIf your submission is success after this, your should search bug into \"make my prediction\" method or continue search anywhere\n\n[Here](https://www.kaggle.com/pestipeti/pytorch-starter-fasterrcnn-inference) you can see successful inference kernel, I hope it helps you!\n\nWelcome!", "votes": null }, { "id": "839226", "postDate": "05/09/2020 07:17:23", "content": "Thanks @shonenkov for your tips.\nI tried submission with the inference kernel you referred and it worked.\nIt seems to make a successful submission, one must upload a model and give prediction in the kernel.\nHowever, it is still a puzzle to me on why directly copy and paste the \"sample_submssion.csv\" as \"submission.csv\" also works as a successful submission.
", "rawMarkdown": "Thanks @shonenkov for your tips.\nI tried submission with the inference kernel you referred and it worked.\nIt seems to make a successful submission, one must upload a model and give prediction in the kernel.\nHowever, it is still a puzzle to me on why directly copy and paste the \"sample_submssion.csv\" as \"submission.csv\" also works as a successful submission.", "votes": null }, { "id": "839339", "postDate": "05/09/2020 10:16:15", "content": "@alexlwh It is code competition with hidden test. It means that when you run \"submit\", test is being replaced during submission. Sample submission has only public test ids without hidden test. So this approach doesn't work here
", "rawMarkdown": "alexlwh It is code competition with hidden test. It means that when you run \"submit\", test is being replaced during submission. Sample submission has only public test ids without hidden test. So this approach doesn't work here", "votes": null }, { "id": "839350", "postDate": "05/09/2020 10:37:04", "content": "@shonenkov \nOh I see! Thanks so much. It clears my confusion!
", "rawMarkdown": "shonenkov \nOh I see! Thanks so much. It clears my confusion!", "votes": null }, { "id": "841737", "postDate": "05/11/2020 01:54:31", "content": "It's odd. Tweet Sentiment Extraction is also a code competition but a quick submission is enabled there.
", "rawMarkdown": "It's odd. Tweet Sentiment Extraction is also a code competition but a quick submission is enabled there.", "votes": null } ], "comments": [ { "id": 838518, "author_name": "matthewmasters", "author_url": "", "post_date": "05/08/2020 16:08:46", "content": "Well what is the submission error? The CSV isn't found? The kernel crashes?
", "votes": null, "replies": [] }, { "id": 838541, "author_name": "bamps53", "author_url": "", "post_date": "05/08/2020 16:26:17", "content": "That doesn't work for this competition because there is a hidden test set. You should predict in your kernel.
", "votes": null, "replies": [] }, { "id": 838682, "author_name": "alexlwh", "author_url": "", "post_date": "05/08/2020 18:04:49", "content": "I encountered the same issue. When I submit sample_submission.csv (write it as submission.csv) as dummy, I can successfully submit it. But when I upload my submission from \"add data\" and write it as submission.csv, it doesn't work. I am not sure whether the problem is caused by the fact I don't make any model prediction on test set locally in my submission kernel.\n@bamps53 could you explain a bit on hidden test set? It seems the test set (for public LB) only has 10 samples which we can access. And when I saw the kernel notebooks from others, their submission.csv also only contain those 10 samples. \n@matthewmasters the error I received was Submission Scoring Error
", "votes": null, "replies": [] }, { "id": 838757, "author_name": "shonenkov", "author_url": "", "post_date": "05/08/2020 19:10:14", "content": "you should debug your inference kernel. you can use try/except for searching bug. I recommend start from your main prediction method for one sample, for example:
\n\npython\nfor image_id, image in dataset:\n try:\n ... = make_my_prediction(image)\n except Exception:\n results.append({\n 'image_id': image_id,\n 'PredictionString': ''\n })\n continue\n
If your submission is success after this, your should search bug into \"make my prediction\" method or continue search anywhere
\n\nHere you can see successful inference kernel, I hope it helps you!
\n\nWelcome!
", "votes": null, "replies": [ { "id": 839226, "author_name": "alexlwh", "author_url": "", "post_date": "05/09/2020 07:17:23", "content": "Thanks @shonenkov for your tips.\nI tried submission with the inference kernel you referred and it worked.\nIt seems to make a successful submission, one must upload a model and give prediction in the kernel.\nHowever, it is still a puzzle to me on why directly copy and paste the \"sample_submssion.csv\" as \"submission.csv\" also works as a successful submission.
", "votes": null, "replies": [] }, { "id": 839339, "author_name": "shonenkov", "author_url": "", "post_date": "05/09/2020 10:16:15", "content": "@alexlwh It is code competition with hidden test. It means that when you run \"submit\", test is being replaced during submission. Sample submission has only public test ids without hidden test. So this approach doesn't work here
", "votes": null, "replies": [] }, { "id": 839350, "author_name": "alexlwh", "author_url": "", "post_date": "05/09/2020 10:37:04", "content": "@shonenkov \nOh I see! Thanks so much. It clears my confusion!
", "votes": null, "replies": [] } ] }, { "id": 841737, "author_name": "cooleel", "author_url": "", "post_date": "05/11/2020 01:54:31", "content": "It's odd. Tweet Sentiment Extraction is also a code competition but a quick submission is enabled there.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "838041": "I trained a model and get a submission file on my own computer. Then I make a new kernel on kaggle, read my file and save file as \"submission.csv\", commit is also ok, but submission always get an error? Submitting \"submission\" file only is not allowed?", "838518": "Well what is the submission error? The CSV isn't found? The kernel crashes?", "838541": "That doesn't work for this competition because there is a hidden test set. You should predict in your kernel.", "838682": "I encountered the same issue. When I submit sample_submission.csv (write it as submission.csv) as dummy, I can successfully submit it. But when I upload my submission from \"add data\" and write it as submission.csv, it doesn't work. I am not sure whether the problem is caused by the fact I don't make any model prediction on test set locally in my submission kernel.\n@bamps53 could you explain a bit on hidden test set? It seems the test set (for public LB) only has 10 samples which we can access. And when I saw the kernel notebooks from others, their submission.csv also only contain those 10 samples. \n@matthewmasters the error I received was Submission Scoring Error", "838757": "you should debug your inference kernel. you can use try/except for searching bug. I recommend start from your main prediction method for one sample, for example:\n\n```python\nfor image_id, image in dataset:\n try:\n ... = make_my_prediction(image)\n except Exception:\n results.append({\n 'image_id': image_id,\n 'PredictionString': ''\n })\n continue\n```\n\nIf your submission is success after this, your should search bug into \"make my prediction\" method or continue search anywhere\n\n[Here](https://www.kaggle.com/pestipeti/pytorch-starter-fasterrcnn-inference) you can see successful inference kernel, I hope it helps you!\n\nWelcome!", "839226": "Thanks @shonenkov for your tips.\nI tried submission with the inference kernel you referred and it worked.\nIt seems to make a successful submission, one must upload a model and give prediction in the kernel.\nHowever, it is still a puzzle to me on why directly copy and paste the \"sample_submssion.csv\" as \"submission.csv\" also works as a successful submission.", "839339": "alexlwh It is code competition with hidden test. It means that when you run \"submit\", test is being replaced during submission. Sample submission has only public test ids without hidden test. So this approach doesn't work here", "839350": "shonenkov \nOh I see! Thanks so much. It clears my confusion!", "841737": "It's odd. Tweet Sentiment Extraction is also a code competition but a quick submission is enabled there." }, "source": "meta" }