Hello,\nI have a little problem, I trained my model on my computer, so I predicted also on my computer.\nI import my submission file on a Kaggle notebook, but after saving version with \"Save and Run\", I can't submit my prediction file (I have my output file on the output section)\nDo you have an idea?
", "messages": [ { "id": 884616, "postDate": "2020-06-13T13:46:38.050Z", "content": "Part of the requirement of this competition is that your inference kernel should NOT have internet access. If you need data from elsewhere to be able to perform the inference (e.g. a saved model checkpoint), you could do so by having a separate training kernel that saves the model checkpoints to file. In your inference kernel, you can then load this model checkpoint as an input using the add input option.
", "rawMarkdown": "Part of the requirement of this competition is that your inference kernel should NOT have internet access. If you need data from elsewhere to be able to perform the inference (e.g. a saved model checkpoint), you could do so by having a separate training kernel that saves the model checkpoints to file. In your inference kernel, you can then load this model checkpoint as an input using the add input option." }, { "id": 884232, "postDate": "2020-06-13T08:54:34.610Z", "content": "You must create an inference kernel. The test set is only 10 images to give you an idea of the data format. The real test set is hidden from us and is much larger. You can look at some of the public notebooks to get an idea how to make it.
", "rawMarkdown": "You must create an inference kernel. The test set is only 10 images to give you an idea of the data format. The real test set is hidden from us and is much larger. You can look at some of the public notebooks to get an idea how to make it.", "replies": [ { "id": 884245, "postDate": "2020-06-13T09:02:34.120Z", "content": "Thanks you for you answer. I also have my submission.csv file in /kaggle/working/. But I don't now how I can create an inference kernel. Can you help me for this ?
", "rawMarkdown": "Thanks you for you answer. I also have my submission.csv file in /kaggle/working/. But I don't now how I can create an inference kernel. Can you help me for this ?" }, { "id": 884279, "postDate": "2020-06-13T09:12:10.193Z", "content": "That's why I say you should look at existing kernels. You need to perform the prediction on the test set in the kernel itself, just as you did on your computer. This is because it must be re-run on the private test set.
", "rawMarkdown": "That's why I say you should look at existing kernels. You need to perform the prediction on the test set in the kernel itself, just as you did on your computer. This is because it must be re-run on the private test set." }, { "id": 884280, "postDate": "2020-06-13T09:12:47.893Z", "content": "ok thanks
", "rawMarkdown": "ok thanks" }, { "id": 884286, "postDate": "2020-06-13T09:16:27.407Z", "rawMarkdown": "", "isDeleted": true }, { "id": 884337, "postDate": "2020-06-13T09:42:39.640Z", "content": "I find the submit button, but now he says : this notebook don t have internet access. I see another discussion, but after checking if I have Internet in this notebook, I have the same error.
", "rawMarkdown": "I find the submit button, but now he says : this notebook don t have internet access. I see another discussion, but after checking if I have Internet in this notebook, I have the same error." } ] }, { "id": 884196, "postDate": "2020-06-13T08:33:28.460Z", "content": "Hello,\nI have a little problem, I trained my model on my computer, so I predicted also on my computer.\nI import my submission file on a Kaggle notebook, but after saving version with \"Save and Run\", I can't submit my prediction file (I have my output file on the output section)\nDo you have an idea?
", "rawMarkdown": "Hello,\nI have a little problem, I trained my model on my computer, so I predicted also on my computer.\nI import my submission file on a Kaggle notebook, but after saving version with \"Save and Run\", I can't submit my prediction file (I have my output file on the output section)\nDo you have an idea?" } ], "comments": [ { "id": 884616, "author_name": "CookCodeSleep", "author_url": "", "post_date": "2020-06-13T13:46:38.050000", "content": "Part of the requirement of this competition is that your inference kernel should NOT have internet access. If you need data from elsewhere to be able to perform the inference (e.g. a saved model checkpoint), you could do so by having a separate training kernel that saves the model checkpoints to file. In your inference kernel, you can then load this model checkpoint as an input using the add input option.
", "votes": 0, "replies": [] }, { "id": 884232, "author_name": "Matt", "author_url": "", "post_date": "2020-06-13T08:54:34.610000", "content": "You must create an inference kernel. The test set is only 10 images to give you an idea of the data format. The real test set is hidden from us and is much larger. You can look at some of the public notebooks to get an idea how to make it.
", "votes": 0, "replies": [ { "id": 884245, "author_name": "Adrien I", "author_url": "", "post_date": "2020-06-13T09:02:34.120000", "content": "Thanks you for you answer. I also have my submission.csv file in /kaggle/working/. But I don't now how I can create an inference kernel. Can you help me for this ?
", "votes": 0, "replies": [] }, { "id": 884279, "author_name": "Matt", "author_url": "", "post_date": "2020-06-13T09:12:10.193000", "content": "That's why I say you should look at existing kernels. You need to perform the prediction on the test set in the kernel itself, just as you did on your computer. This is because it must be re-run on the private test set.
", "votes": 0, "replies": [] }, { "id": 884280, "author_name": "Adrien I", "author_url": "", "post_date": "2020-06-13T09:12:47.893000", "content": "ok thanks
", "votes": 0, "replies": [] }, { "id": 884286, "author_name": "", "author_url": "", "post_date": "2020-06-13T09:16:27.407000", "content": "", "votes": 0, "replies": [] }, { "id": 884337, "author_name": "Adrien I", "author_url": "", "post_date": "2020-06-13T09:42:39.640000", "content": "I find the submit button, but now he says : this notebook don t have internet access. I see another discussion, but after checking if I have Internet in this notebook, I have the same error.
", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "884616": "Part of the requirement of this competition is that your inference kernel should NOT have internet access. If you need data from elsewhere to be able to perform the inference (e.g. a saved model checkpoint), you could do so by having a separate training kernel that saves the model checkpoints to file. In your inference kernel, you can then load this model checkpoint as an input using the add input option.", "884232": "You must create an inference kernel. The test set is only 10 images to give you an idea of the data format. The real test set is hidden from us and is much larger. You can look at some of the public notebooks to get an idea how to make it.", "884196": "Hello,\nI have a little problem, I trained my model on my computer, so I predicted also on my computer.\nI import my submission file on a Kaggle notebook, but after saving version with \"Save and Run\", I can't submit my prediction file (I have my output file on the output section)\nDo you have an idea?" } }