![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2305018%2F5257f6471171fdd03cc03f8fa15a0a0f%2FBildschirmfoto%20vom%202019-12-25%2020-45-27.png?generation=1577304139113851&alt=media\")
Sorry, quite new here. I have problems scoring my predictions. I first got the error \"Submission CSV Not Found\", which I think was related to memory issues (I loaded all 4 test files at once). Now I get the message \"Submission Scoring Error\" after approx 5 minutes. Have not found anything useful in other threads. I am pretty sure that the output file is in the correct format. Grateful for hints, thanks.
\n\n
Sorry, quite new here. I have problems scoring my predictions. I first got the error \"Submission CSV Not Found\", which I think was related to memory issues (I loaded all 4 test files at once). Now I get the message \"Submission Scoring Error\" after approx 5 minutes. Have not found anything useful in other threads. I am pretty sure that the output file is in the correct format. Grateful for hints, thanks.
\n\n
I don't know if you noticed but the Test dataset provided by the competition authors is only a small subsets of the actual dataset used for the public test as explained here: https://www.kaggle.com/c/bengaliai-cv19/discussion/122420
\n\nInstead, you should build the \"rowid\" column by using the imageid information hold in the 4 parquets file as the 4 parquet files used when evaluating your notebook for the public score contains way more than just 12 images.
\n\nIf the format of your .csv is incorrect (e.g not enough lines..) , I think it outputs that it is not found. It happened to me a lot of time and wasted quite a few submissions… When I did this mistake (by hardcoding the use of only 12 images to create the submission file and not the full 4 parquets), the \"Submission CSV Not Found\" error message appeared.
\n\nHowever, now all is fixed, I achieved to have a deplorable low score but I figured out an error and when fixing it, I now encounter the \"Submission Scoring Error\" which is very frustrating as the notebook is commited and is said to be functional by Kaggle..
", "rawMarkdown": "I don't know if you noticed but the Test dataset provided by the competition authors is only a small subsets of the actual dataset used for the public test as explained here: https://www.kaggle.com/c/bengaliai-cv19/discussion/122420\n\nInstead, you should build the \"rowid\" column by using the imageid information hold in the 4 parquets file as the 4 parquet files used when evaluating your notebook for the public score contains way more than just 12 images.\n\nIf the format of your .csv is incorrect (e.g not enough lines..) , I think it outputs that it is not found. It happened to me a lot of time and wasted quite a few submissions… When I did this mistake (by hardcoding the use of only 12 images to create the submission file and not the full 4 parquets), the \"***Submission CSV Not Found***\" error message appeared.\n\nHowever, now all is fixed, I achieved to have a deplorable low score but I figured out an error and when fixing it, I now encounter the \"***Submission Scoring Error***\" which is very frustrating as the notebook is commited and is said to be functional by Kaggle..", "votes": null }, { "id": "705318", "postDate": "12/28/2019 19:12:40", "content": "I found that a \"Submission Scoring Error\" is linked to an error happening with the code and in my case, it is a MemoryError when doing an operation converting a full parquet of images already loaded as a dataframe of uint8 into a dataframe of floats... \nI discovered this by running my inference notebook on the train parquets, which sizes are equivalent to the ones of the actual test parquet
", "rawMarkdown": "I found that a \"***Submission Scoring Error***\" is linked to an error happening with the code and in my case, it is a MemoryError when doing an operation converting a full parquet of images already loaded as a dataframe of uint8 into a dataframe of floats... \nI discovered this by running my inference notebook on the train parquets, which sizes are equivalent to the ones of the actual test parquet", "votes": null }, { "id": "705942", "postDate": "12/29/2019 17:33:42", "content": "Thanks for your help. Finally I got it working, I think it was really related to memory issues.
", "rawMarkdown": "Thanks for your help. Finally I got it working, I think it was really related to memory issues.", "votes": null }, { "id": "706504", "postDate": "12/30/2019 13:33:33", "content": "@dimartinot I discovered this by running my inference notebook on the train parquets, which sizes are equivalent to the ones of the actual test parquet \nI tried the same but turns out \"Submission Scoring Error\" is not linked to memory issue in my case.
I am getting a 'notebook exceeded compute error'. What might be the cause?
", "rawMarkdown": "I am getting a 'notebook exceeded compute error'. What might be the cause?", "votes": null }, { "id": "711832", "postDate": "01/06/2020 15:07:47", "content": "It happened to me too where I saturated the RAM provided by the notebook. It did not result in a Python MemoryError as it seemed to have directly crashed the notebook's code at that time. I diagnosed this issue with again the same strategy: trying to run inference on the train_parquet files.
", "rawMarkdown": "It happened to me too where I saturated the RAM provided by the notebook. It did not result in a Python MemoryError as it seemed to have directly crashed the notebook's code at that time. I diagnosed this issue with again the same strategy: trying to run inference on the train_parquet files.", "votes": null } ], "comments": [ { "id": 705290, "author_name": "dimartinot", "author_url": "", "post_date": "12/28/2019 18:04:57", "content": "I don't know if you noticed but the Test dataset provided by the competition authors is only a small subsets of the actual dataset used for the public test as explained here: https://www.kaggle.com/c/bengaliai-cv19/discussion/122420
\n\nInstead, you should build the \"rowid\" column by using the imageid information hold in the 4 parquets file as the 4 parquet files used when evaluating your notebook for the public score contains way more than just 12 images.
\n\nIf the format of your .csv is incorrect (e.g not enough lines..) , I think it outputs that it is not found. It happened to me a lot of time and wasted quite a few submissions… When I did this mistake (by hardcoding the use of only 12 images to create the submission file and not the full 4 parquets), the \"Submission CSV Not Found\" error message appeared.
\n\nHowever, now all is fixed, I achieved to have a deplorable low score but I figured out an error and when fixing it, I now encounter the \"Submission Scoring Error\" which is very frustrating as the notebook is commited and is said to be functional by Kaggle..
", "votes": null, "replies": [ { "id": 705318, "author_name": "dimartinot", "author_url": "", "post_date": "12/28/2019 19:12:40", "content": "I found that a \"Submission Scoring Error\" is linked to an error happening with the code and in my case, it is a MemoryError when doing an operation converting a full parquet of images already loaded as a dataframe of uint8 into a dataframe of floats... \nI discovered this by running my inference notebook on the train parquets, which sizes are equivalent to the ones of the actual test parquet
", "votes": null, "replies": [] }, { "id": 705942, "author_name": "terneat", "author_url": "", "post_date": "12/29/2019 17:33:42", "content": "Thanks for your help. Finally I got it working, I think it was really related to memory issues.
", "votes": null, "replies": [] }, { "id": 706504, "author_name": "namanj27", "author_url": "", "post_date": "12/30/2019 13:33:33", "content": "@dimartinot I discovered this by running my inference notebook on the train parquets, which sizes are equivalent to the ones of the actual test parquet \nI tried the same but turns out \"Submission Scoring Error\" is not linked to memory issue in my case.
I am getting a 'notebook exceeded compute error'. What might be the cause?
", "votes": null, "replies": [] }, { "id": 711832, "author_name": "dimartinot", "author_url": "", "post_date": "01/06/2020 15:07:47", "content": "It happened to me too where I saturated the RAM provided by the notebook. It did not result in a Python MemoryError as it seemed to have directly crashed the notebook's code at that time. I diagnosed this issue with again the same strategy: trying to run inference on the train_parquet files.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "703215": "Sorry, quite new here. I have problems scoring my predictions. I first got the error \"Submission CSV Not Found\", which I think was related to memory issues (I loaded all 4 test files at once). Now I get the message \"Submission Scoring Error\" after approx 5 minutes. Have not found anything useful in other threads. I am pretty sure that the output file is in the correct format. Grateful for hints, thanks.\n\n", "705290": "I don't know if you noticed but the Test dataset provided by the competition authors is only a small subsets of the actual dataset used for the public test as explained here: https://www.kaggle.com/c/bengaliai-cv19/discussion/122420\n\nInstead, you should build the \"rowid\" column by using the imageid information hold in the 4 parquets file as the 4 parquet files used when evaluating your notebook for the public score contains way more than just 12 images.\n\nIf the format of your .csv is incorrect (e.g not enough lines..) , I think it outputs that it is not found. It happened to me a lot of time and wasted quite a few submissions… When I did this mistake (by hardcoding the use of only 12 images to create the submission file and not the full 4 parquets), the \"***Submission CSV Not Found***\" error message appeared.\n\nHowever, now all is fixed, I achieved to have a deplorable low score but I figured out an error and when fixing it, I now encounter the \"***Submission Scoring Error***\" which is very frustrating as the notebook is commited and is said to be functional by Kaggle..", "705318": "I found that a \"***Submission Scoring Error***\" is linked to an error happening with the code and in my case, it is a MemoryError when doing an operation converting a full parquet of images already loaded as a dataframe of uint8 into a dataframe of floats... \nI discovered this by running my inference notebook on the train parquets, which sizes are equivalent to the ones of the actual test parquet", "705942": "Thanks for your help. Finally I got it working, I think it was really related to memory issues.", "706504": "dimartinot `I discovered this by running my inference notebook on the train parquets, which sizes are equivalent to the ones of the actual test parquet` \nI tried the same but turns out *\"Submission Scoring Error\"* is not linked to memory issue in my case.", "710340": "I am getting a 'notebook exceeded compute error'. What might be the cause?", "711832": "It happened to me too where I saturated the RAM provided by the notebook. It did not result in a Python MemoryError as it seemed to have directly crashed the notebook's code at that time. I diagnosed this issue with again the same strategy: trying to run inference on the train_parquet files." }, "source": "meta" }