My every notebook gets scored except the second we bring in an external notebook( which is a csv file with patient name and a metric). The notebook runs perfectly fine as long as we do not take this csv data into the model. but when we do it gives \"error scoring submission\". The dataset has public access. The number of weeks etc are perfectly right. The notebook runs and gives \"submission.csv\" with proper values(from week -12 to 133). I am completely stuck at this point with more than 10 submissions with this. the only difference between a working scoring and one that doesnt is the change in the csv data that comes in as an extra variables for the model. Anyone is free to take a look as I am completely lost here as to what I can do further.(runtime= 138.5 seconds)
", "messages": [ { "id": 957940, "postDate": "2020-08-04T16:32:54.027Z", "content": "You have to compute those external features for private test set. Are you doing that?
", "rawMarkdown": "You have to compute those external features for private test set. Are you doing that?", "votes": 1, "replies": [ { "id": 957958, "postDate": "2020-08-04T16:46:59.783Z", "content": "I computed the external features on the training set, however, the external csv is based only on the CT scans, which are common for both test and train set. So i merge the test set on the patients ( as the patients in the test set are present in those of the train set for this problem for this initial stage of the leaderboard). So essentially yes. Unless I am mistaken in some way.
", "rawMarkdown": "I computed the external features on the training set, however, the external csv is based only on the CT scans, which are common for both test and train set. So i merge the test set on the patients ( as the patients in the test set are present in those of the train set for this problem for this initial stage of the leaderboard). So essentially yes. Unless I am mistaken in some way." }, { "id": 957966, "postDate": "2020-08-04T16:52:06.407Z", "content": "It looks like you are computing them for the dummy test set. The real test set is hidden so you have to compute them in runtime.
", "rawMarkdown": "It looks like you are computing them for the dummy test set. The real test set is hidden so you have to compute them in runtime.", "votes": 1 }, { "id": 957971, "postDate": "2020-08-04T16:58:09.330Z", "content": "Hmm.. i didnt realize that! so wait, we cannot work with the ct scans externally for the test set huh? Wow, interesting :P
", "rawMarkdown": "Hmm.. i didnt realize that! so wait, we cannot work with the ct scans externally for the test set huh? Wow, interesting :P", "votes": 1 } ] }, { "id": 957730, "postDate": "2020-08-04T14:13:15.283Z", "content": "https://www.kaggle.com/melvin97n/base5\nheres the link to it if anyone is curious(iv made it public for this reason alone)
", "rawMarkdown": "https://www.kaggle.com/melvin97n/base5\nheres the link to it if anyone is curious(iv made it public for this reason alone)\n", "votes": -1 }, { "id": 957720, "postDate": "2020-08-04T14:06:43.443Z", "content": "My every notebook gets scored except the second we bring in an external notebook( which is a csv file with patient name and a metric). The notebook runs perfectly fine as long as we do not take this csv data into the model. but when we do it gives \"error scoring submission\". The dataset has public access. The number of weeks etc are perfectly right. The notebook runs and gives \"submission.csv\" with proper values(from week -12 to 133). I am completely stuck at this point with more than 10 submissions with this. the only difference between a working scoring and one that doesnt is the change in the csv data that comes in as an extra variables for the model. Anyone is free to take a look as I am completely lost here as to what I can do further.(runtime= 138.5 seconds)
", "rawMarkdown": "My every notebook gets scored except the second we bring in an external notebook( which is a csv file with patient name and a metric). The notebook runs perfectly fine as long as we do not take this csv data into the model. but when we do it gives \"error scoring submission\". The dataset has public access. The number of weeks etc are perfectly right. The notebook runs and gives \"submission.csv\" with proper values(from week -12 to 133). I am completely stuck at this point with more than 10 submissions with this. the only difference between a working scoring and one that doesnt is the change in the csv data that comes in as an extra variables for the model. Anyone is free to take a look as I am completely lost here as to what I can do further.(runtime= 138.5 seconds)" }, { "id": 957948, "postDate": "2020-08-04T16:40:19.330Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 957940, "author_name": "Gunes Evitan", "author_url": "", "post_date": "2020-08-04T16:32:54.027000", "content": "You have to compute those external features for private test set. Are you doing that?
", "votes": 1, "replies": [ { "id": 957958, "author_name": "mseb", "author_url": "", "post_date": "2020-08-04T16:46:59.783000", "content": "I computed the external features on the training set, however, the external csv is based only on the CT scans, which are common for both test and train set. So i merge the test set on the patients ( as the patients in the test set are present in those of the train set for this problem for this initial stage of the leaderboard). So essentially yes. Unless I am mistaken in some way.
", "votes": 0, "replies": [] }, { "id": 957966, "author_name": "Gunes Evitan", "author_url": "", "post_date": "2020-08-04T16:52:06.407000", "content": "It looks like you are computing them for the dummy test set. The real test set is hidden so you have to compute them in runtime.
", "votes": 1, "replies": [] }, { "id": 957971, "author_name": "mseb", "author_url": "", "post_date": "2020-08-04T16:58:09.330000", "content": "Hmm.. i didnt realize that! so wait, we cannot work with the ct scans externally for the test set huh? Wow, interesting :P
", "votes": 1, "replies": [] } ] }, { "id": 957730, "author_name": "mseb", "author_url": "", "post_date": "2020-08-04T14:13:15.283000", "content": "https://www.kaggle.com/melvin97n/base5\nheres the link to it if anyone is curious(iv made it public for this reason alone)
", "votes": -1, "replies": [] }, { "id": 957948, "author_name": "", "author_url": "", "post_date": "2020-08-04T16:40:19.330000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "957940": "You have to compute those external features for private test set. Are you doing that?", "957730": "https://www.kaggle.com/melvin97n/base5\nheres the link to it if anyone is curious(iv made it public for this reason alone)\n", "957720": "My every notebook gets scored except the second we bring in an external notebook( which is a csv file with patient name and a metric). The notebook runs perfectly fine as long as we do not take this csv data into the model. but when we do it gives \"error scoring submission\". The dataset has public access. The number of weeks etc are perfectly right. The notebook runs and gives \"submission.csv\" with proper values(from week -12 to 133). I am completely stuck at this point with more than 10 submissions with this. the only difference between a working scoring and one that doesnt is the change in the csv data that comes in as an extra variables for the model. Anyone is free to take a look as I am completely lost here as to what I can do further.(runtime= 138.5 seconds)", "957948": "" } }