{
  "id": 167646,
  "title": "Additional GPU",
  "url": "/competitions/prostate-cancer-grade-assessment/discussion/167646",
  "author_name": "Andreas Horlbeck",
  "post_date": "2020-07-17T11:45:03.161000",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I know that this topic is stressed a lot but I am frustrated and want to discussion again. </p>\n\n<p>The Limitation of 30 hours GPU is a huge problem but 100% clear that there budget limitations. </p>\n\n<p>But couldnt there be an agreeement, like you can choose TPU or additional GPU, so that you could have 60 hours in a week. Because here on a kaggle a lot of competitions have a scientific background there should not be such a technical limitation to get the best result. Speaking for me I would also pay for additional gpu. It should be affordable to not be a disadvantage for people with not much money like students but yeah I would pay. \nWhat do you mean?</p>",
  "messages": [
    {
      "id": 932934,
      "postDate": "2020-07-17T11:45:03.163Z",
      "content": "<p>I know that this topic is stressed a lot but I am frustrated and want to discussion again. </p>\n\n<p>The Limitation of 30 hours GPU is a huge problem but 100% clear that there budget limitations. </p>\n\n<p>But couldnt there be an agreeement, like you can choose TPU or additional GPU, so that you could have 60 hours in a week. Because here on a kaggle a lot of competitions have a scientific background there should not be such a technical limitation to get the best result. Speaking for me I would also pay for additional gpu. It should be affordable to not be a disadvantage for people with not much money like students but yeah I would pay. \nWhat do you mean?</p>",
      "rawMarkdown": "I know that this topic is stressed a lot but I am frustrated and want to discussion again. \n\nThe Limitation of 30 hours GPU is a huge problem but 100% clear that there budget limitations. \n\nBut couldnt there be an agreeement, like you can choose TPU or additional GPU, so that you could have 60 hours in a week. Because here on a kaggle a lot of competitions have a scientific background there should not be such a technical limitation to get the best result. Speaking for me I would also pay for additional gpu. It should be affordable to not be a disadvantage for people with not much money like students but yeah I would pay. \nWhat do you mean?",
      "votes": 2
    },
    {
      "id": 938003,
      "postDate": "2020-07-21T09:44:25.703Z",
      "content": "<p>Any limitation would always be annoying to some users but it makes sense for a number of reasons 🙂</p>\n\n<p>Still, I think it might be a reasonable idea to keep the same total weekly quota (30h+30h) or even reduce it to 50h but allow a user to distribute the hours between TPU and GPU depending on their needs, or simply have something like \"accelerator quota\". Of course, I don't really know if the costs from the two are comparable.</p>",
      "rawMarkdown": "Any limitation would always be annoying to some users but it makes sense for a number of reasons 🙂\n\nStill, I think it might be a reasonable idea to keep the same total weekly quota (30h+30h) or even reduce it to 50h but allow a user to distribute the hours between TPU and GPU depending on their needs, or simply have something like \"accelerator quota\". Of course, I don't really know if the costs from the two are comparable."
    }
  ],
  "comments": [
    {
      "id": 938003,
      "author_name": "Nikita Kozodoi",
      "author_url": "",
      "post_date": "2020-07-21T09:44:25.703000",
      "content": "<p>Any limitation would always be annoying to some users but it makes sense for a number of reasons 🙂</p>\n\n<p>Still, I think it might be a reasonable idea to keep the same total weekly quota (30h+30h) or even reduce it to 50h but allow a user to distribute the hours between TPU and GPU depending on their needs, or simply have something like \"accelerator quota\". Of course, I don't really know if the costs from the two are comparable.</p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "932934": "I know that this topic is stressed a lot but I am frustrated and want to discussion again. \n\nThe Limitation of 30 hours GPU is a huge problem but 100% clear that there budget limitations. \n\nBut couldnt there be an agreeement, like you can choose TPU or additional GPU, so that you could have 60 hours in a week. Because here on a kaggle a lot of competitions have a scientific background there should not be such a technical limitation to get the best result. Speaking for me I would also pay for additional gpu. It should be affordable to not be a disadvantage for people with not much money like students but yeah I would pay. \nWhat do you mean?",
    "938003": "Any limitation would always be annoying to some users but it makes sense for a number of reasons 🙂\n\nStill, I think it might be a reasonable idea to keep the same total weekly quota (30h+30h) or even reduce it to 50h but allow a user to distribute the hours between TPU and GPU depending on their needs, or simply have something like \"accelerator quota\". Of course, I don't really know if the costs from the two are comparable."
  }
}