{
  "id": 71811,
  "title": "Preinstalled packages and mismatched versions",
  "url": "/competitions/quora-insincere-questions-classification/discussion/71811",
  "author_name": "",
  "post_date": "2018-11-16T19:42:16.595382400Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>From my understanding, for this Kernel only competition we have to use the preinstalled packages that come with the kernel. What if there is a mismatch between packages causing them to have issues? In this case, Fast.ai 0.7 is known to have issues with PyTorch 0.4.1. I ran into a few of them working on this competition. Do we have to just bite the bullet and ignore those packages and try something else? Are we allowed to upgrade/downgrade \"existing\" packages?</p>",
  "messages": [
    {
      "id": "422797",
      "postDate": "11/16/2018 19:42:16",
      "content": "<p>From my understanding, for this Kernel only competition we have to use the preinstalled packages that come with the kernel. What if there is a mismatch between packages causing them to have issues? In this case, Fast.ai 0.7 is known to have issues with PyTorch 0.4.1. I ran into a few of them working on this competition. Do we have to just bite the bullet and ignore those packages and try something else? Are we allowed to upgrade/downgrade \"existing\" packages?</p>",
      "rawMarkdown": "From my understanding, for this Kernel only competition we have to use the preinstalled packages that come with the kernel. What if there is a mismatch between packages causing them to have issues? In this case, Fast.ai 0.7 is known to have issues with PyTorch 0.4.1. I ran into a few of them working on this competition. Do we have to just bite the bullet and ignore those packages and try something else? Are we allowed to upgrade/downgrade \"existing\" packages?",
      "votes": null
    },
    {
      "id": "422820",
      "postDate": "11/16/2018 20:29:32",
      "content": "<p>I wonder whether you can just do a PR to the docker file: <a href=\"https://github.com/Kaggle/docker-python/blob/master/Dockerfile#L439\">https://github.com/Kaggle/docker-python/blob/master/Dockerfile#L439</a></p>",
      "rawMarkdown": "I wonder whether you can just do a PR to the docker file: https://github.com/Kaggle/docker-python/blob/master/Dockerfile#L439",
      "votes": null
    },
    {
      "id": "422876",
      "postDate": "11/16/2018 23:20:18",
      "content": "<p>Thanks for the link, they already know Pytorch .4 does not work and just install Fast.ai with no deps, they have some comments in the file. Ideally we would have an option to use v1 for both libraries. I was more curious if we can do anything given the contest rules. I guess it is just a waiting game for these \"Kernel\" only competitions.</p>",
      "rawMarkdown": "Thanks for the link, they already know Pytorch .4 does not work and just install Fast.ai with no deps, they have some comments in the file. Ideally we would have an option to use v1 for both libraries. I was more curious if we can do anything given the contest rules. I guess it is just a waiting game for these \"Kernel\" only competitions.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 422820,
      "author_name": "shujian",
      "author_url": "",
      "post_date": "11/16/2018 20:29:32",
      "content": "<p>I wonder whether you can just do a PR to the docker file: <a href=\"https://github.com/Kaggle/docker-python/blob/master/Dockerfile#L439\">https://github.com/Kaggle/docker-python/blob/master/Dockerfile#L439</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 422876,
          "author_name": "sariabod",
          "author_url": "",
          "post_date": "11/16/2018 23:20:18",
          "content": "<p>Thanks for the link, they already know Pytorch .4 does not work and just install Fast.ai with no deps, they have some comments in the file. Ideally we would have an option to use v1 for both libraries. I was more curious if we can do anything given the contest rules. I guess it is just a waiting game for these \"Kernel\" only competitions.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "422797": "From my understanding, for this Kernel only competition we have to use the preinstalled packages that come with the kernel. What if there is a mismatch between packages causing them to have issues? In this case, Fast.ai 0.7 is known to have issues with PyTorch 0.4.1. I ran into a few of them working on this competition. Do we have to just bite the bullet and ignore those packages and try something else? Are we allowed to upgrade/downgrade \"existing\" packages?",
    "422820": "I wonder whether you can just do a PR to the docker file: https://github.com/Kaggle/docker-python/blob/master/Dockerfile#L439",
    "422876": "Thanks for the link, they already know Pytorch .4 does not work and just install Fast.ai with no deps, they have some comments in the file. Ideally we would have an option to use v1 for both libraries. I was more curious if we can do anything given the contest rules. I guess it is just a waiting game for these \"Kernel\" only competitions."
  },
  "source": "meta"
}