In the dataset for this competition, -1011141123527297803_long is the first example_id in sample_submission.csv. This perfectly matches the test set in the 4G download of the dataset. However, in the dataset attached to a newly created notebook kernel, simplified-nq-test.jsonl has different examples. The closest example_id to the one above is -1011141123527297792_long.
So the ids in simplified-nq-test.jsonl do not match the ids in sample_submission.csv for a newly created notebook. This causes confusion and submission errors.
In the dataset for this competition, -1011141123527297803_long is the first example_id in sample_submission.csv. This perfectly matches the test set in the 4G download of the dataset. However, in the dataset attached to a newly created notebook kernel, simplified-nq-test.jsonl has different examples. The closest example_id to the one above is -1011141123527297792_long.
So the ids in simplified-nq-test.jsonl do not match the ids in sample_submission.csv for a newly created notebook. This causes confusion and submission errors.
Hi, Ken! Good to see you here.
\n\nThis shall not be a problem, the eval script operates with dictionaries, and the order of example ids doesn’t matter much.
", "rawMarkdown": "Hi, Ken! Good to see you here. \n\nThis shall not be a problem, the eval script operates with dictionaries, and the order of example ids doesn’t matter much.", "votes": null }, { "id": "706514", "postDate": "12/30/2019 13:50:04", "content": "Hi Yury. Good to hear from you again. I hope you are well.
\n\nI am not concerned about the order of the examples. I am concerned that the examples in simplified-nq-test.jsonl are different to the examples in sample_submission.csv.
They are actually the same. Looks like you read them somewhere with Pandas, and the int type is not read correctly. Specify string type then for example ids.
", "rawMarkdown": "They are actually the same. Looks like you read them somewhere with Pandas, and the int type is not read correctly. Specify string type then for example ids.", "votes": null }, { "id": "706540", "postDate": "12/30/2019 14:33:27", "content": "Hi Yury. I am indeed using Pandas, but even if I explicitly convert to string they are different as this very short kernel illustrates.
", "rawMarkdown": "Hi Yury. I am indeed using Pandas, but even if I explicitly convert to string they are different as [this](https://www.kaggle.com/kenkrige/dataset-conflict) very short kernel illustrates.", "votes": null }, { "id": "706625", "postDate": "12/30/2019 16:19:06", "content": "My apologies. @kashnitsky you are absolutely right. I made a mistake with reading as int then converting to string, which gave a different example_id in the least significant digits.
", "rawMarkdown": "My apologies. @kashnitsky you are absolutely right. I made a mistake with reading as int then converting to string, which gave a different example_id in the least significant digits.", "votes": null } ], "comments": [ { "id": 706482, "author_name": "kashnitsky", "author_url": "", "post_date": "12/30/2019 12:58:47", "content": "Hi, Ken! Good to see you here.
\n\nThis shall not be a problem, the eval script operates with dictionaries, and the order of example ids doesn’t matter much.
", "votes": null, "replies": [ { "id": 706514, "author_name": "kenkrige", "author_url": "", "post_date": "12/30/2019 13:50:04", "content": "Hi Yury. Good to hear from you again. I hope you are well.
\n\nI am not concerned about the order of the examples. I am concerned that the examples in simplified-nq-test.jsonl are different to the examples in sample_submission.csv.
They are actually the same. Looks like you read them somewhere with Pandas, and the int type is not read correctly. Specify string type then for example ids.
", "votes": null, "replies": [] }, { "id": 706540, "author_name": "kenkrige", "author_url": "", "post_date": "12/30/2019 14:33:27", "content": "Hi Yury. I am indeed using Pandas, but even if I explicitly convert to string they are different as this very short kernel illustrates.
", "votes": null, "replies": [] }, { "id": 706625, "author_name": "kenkrige", "author_url": "", "post_date": "12/30/2019 16:19:06", "content": "My apologies. @kashnitsky you are absolutely right. I made a mistake with reading as int then converting to string, which gave a different example_id in the least significant digits.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "706411": "In the dataset for this competition, `-1011141123527297803_long` is the first `example_id` in `sample_submission.csv`. This perfectly matches the test set in the 4G download of the dataset. However, in the dataset attached to a newly created notebook kernel, `simplified-nq-test.jsonl` has different examples. The closest `example_id` to the one above is `-1011141123527297792_long`.\n\nSo the ids in `simplified-nq-test.jsonl` do not match the ids in `sample_submission.csv` for a newly created notebook. This causes confusion and submission errors.", "706482": "Hi, Ken! Good to see you here. \n\nThis shall not be a problem, the eval script operates with dictionaries, and the order of example ids doesn’t matter much.", "706514": "Hi Yury. Good to hear from you again. I hope you are well.\n\nI am not concerned about the order of the examples. I am concerned that the examples in `simplified-nq-test.jsonl` are different to the examples in `sample_submission.csv`.", "706515": "They are actually the same. Looks like you read them somewhere with Pandas, and the int type is not read correctly. Specify string type then for example ids.", "706540": "Hi Yury. I am indeed using Pandas, but even if I explicitly convert to string they are different as [this](https://www.kaggle.com/kenkrige/dataset-conflict) very short kernel illustrates.", "706625": "My apologies. @kashnitsky you are absolutely right. I made a mistake with reading as int then converting to string, which gave a different example_id in the least significant digits." }, "source": "meta" }