{
  "id": 206973,
  "title": "How to contact kaggle team to fix a problem in my account?",
  "url": "/competitions/riiid-test-answer-prediction/discussion/206973",
  "author_name": "",
  "post_date": "2020-12-27T13:26:46.973341Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>There is a bug in my account that disables GPU during submission, and since I am using transformers, this limits the model size I can use or else my notebook will timeout. </p>\n<p>I tried mailing kaggle staff and no one answered me! How else to contact them, or if someone from kaggle staff sees this, please, help! I am sure I can have a higher score for my model without this limitation.</p>\n<p>To verify this, I submitted this solution which failed:</p>\n<pre><code>if device == torch.device('cuda'):\n    for (test_data,sample_prediction_df) in iter_test:\n\n        test_data['answered_correctly'] = 0.5\n        env.predict(test_data.loc[test_data['content_type_id'] == 0, ['row_id', 'answered_correctly']])\n</code></pre>",
  "messages": [
    {
      "id": "1128480",
      "postDate": "12/27/2020 13:26:46",
      "content": "<p>There is a bug in my account that disables GPU during submission, and since I am using transformers, this limits the model size I can use or else my notebook will timeout. </p>\n<p>I tried mailing kaggle staff and no one answered me! How else to contact them, or if someone from kaggle staff sees this, please, help! I am sure I can have a higher score for my model without this limitation.</p>\n<p>To verify this, I submitted this solution which failed:</p>\n<pre><code>if device == torch.device('cuda'):\n    for (test_data,sample_prediction_df) in iter_test:\n\n        test_data['answered_correctly'] = 0.5\n        env.predict(test_data.loc[test_data['content_type_id'] == 0, ['row_id', 'answered_correctly']])\n</code></pre>",
      "rawMarkdown": "There is a bug in my account that disables GPU during submission, and since I am using transformers, this limits the model size I can use or else my notebook will timeout. \n\nI tried mailing kaggle staff and no one answered me! How else to contact them, or if someone from kaggle staff sees this, please, help! I am sure I can have a higher score for my model without this limitation.\n\nTo verify this, I submitted this solution which failed:\n\n\t\t\t\n\tif device == torch.device('cuda'):\n\t\tfor (test_data,sample_prediction_df) in iter_test:\n\n\t\t\ttest_data['answered_correctly'] = 0.5\n\t\t\tenv.predict(test_data.loc[test_data['content_type_id'] == 0, ['row_id', 'answered_correctly']])",
      "votes": null
    },
    {
      "id": "1129747",
      "postDate": "12/28/2020 14:23:25",
      "content": "<p>Hey Abdessalem,</p>\n<p>The best way is to contact Kaggle <a href=\"https://www.kaggle.com/contact\" target=\"_blank\">support</a>. However, please note that Kaggle support cannot debug code for you. We show that your account has used GPUs this week (reminder, you have a weekly limit to use). If you have any error messages, please take a screenshot and include them with your request.</p>\n<p>Thanks!</p>",
      "rawMarkdown": "Hey Abdessalem,\n\nThe best way is to contact Kaggle [support](https://www.kaggle.com/contact). However, please note that Kaggle support cannot debug code for you. We show that your account has used GPUs this week (reminder, you have a weekly limit to use). If you have any error messages, please take a screenshot and include them with your request.\n\nThanks!",
      "votes": null
    },
    {
      "id": "1129767",
      "postDate": "12/28/2020 14:36:56",
      "content": "<p>Hello Addison, </p>\n<p>Thanks for taking time to respond!</p>\n<p>This is only but a dummy code to verify my hypothesis: GPU does not work when I submit my model to be graded. This is not my actual code.</p>\n<p>The code above submits a dummy prediction (0.5) only if GPU is enabled, and make the submission fail otherwise; and indeed it failed, meaning that GPU is not active while the notebook is being graded in your servers.</p>\n<p>And yes, I have used GPU this week in my account, and it works perfectly, my problem only arises during grading. I believe my notebook timeout (or take significantly more time) because of this problem given the slower nature matrix operations on CPU.</p>\n<p>Best regards, <br>\nAbdessalem Boukil.</p>",
      "rawMarkdown": "Hello Addison, \n\nThanks for taking time to respond!\n\nThis is only but a dummy code to verify my hypothesis: GPU does not work when I submit my model to be graded. This is not my actual code.\n\nThe code above submits a dummy prediction (0.5) only if GPU is enabled, and make the submission fail otherwise; and indeed it failed, meaning that GPU is not active while the notebook is being graded in your servers.\n\nAnd yes, I have used GPU this week in my account, and it works perfectly, my problem only arises during grading. I believe my notebook timeout (or take significantly more time) because of this problem given the slower nature matrix operations on CPU.\n\nBest regards, \nAbdessalem Boukil.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1129747,
      "author_name": "addisonhoward",
      "author_url": "",
      "post_date": "12/28/2020 14:23:25",
      "content": "<p>Hey Abdessalem,</p>\n<p>The best way is to contact Kaggle <a href=\"https://www.kaggle.com/contact\" target=\"_blank\">support</a>. However, please note that Kaggle support cannot debug code for you. We show that your account has used GPUs this week (reminder, you have a weekly limit to use). If you have any error messages, please take a screenshot and include them with your request.</p>\n<p>Thanks!</p>",
      "votes": null,
      "replies": [
        {
          "id": 1129767,
          "author_name": "abdessalemboukil",
          "author_url": "",
          "post_date": "12/28/2020 14:36:56",
          "content": "<p>Hello Addison, </p>\n<p>Thanks for taking time to respond!</p>\n<p>This is only but a dummy code to verify my hypothesis: GPU does not work when I submit my model to be graded. This is not my actual code.</p>\n<p>The code above submits a dummy prediction (0.5) only if GPU is enabled, and make the submission fail otherwise; and indeed it failed, meaning that GPU is not active while the notebook is being graded in your servers.</p>\n<p>And yes, I have used GPU this week in my account, and it works perfectly, my problem only arises during grading. I believe my notebook timeout (or take significantly more time) because of this problem given the slower nature matrix operations on CPU.</p>\n<p>Best regards, <br>\nAbdessalem Boukil.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1128480": "There is a bug in my account that disables GPU during submission, and since I am using transformers, this limits the model size I can use or else my notebook will timeout. \n\nI tried mailing kaggle staff and no one answered me! How else to contact them, or if someone from kaggle staff sees this, please, help! I am sure I can have a higher score for my model without this limitation.\n\nTo verify this, I submitted this solution which failed:\n\n\t\t\t\n\tif device == torch.device('cuda'):\n\t\tfor (test_data,sample_prediction_df) in iter_test:\n\n\t\t\ttest_data['answered_correctly'] = 0.5\n\t\t\tenv.predict(test_data.loc[test_data['content_type_id'] == 0, ['row_id', 'answered_correctly']])",
    "1129747": "Hey Abdessalem,\n\nThe best way is to contact Kaggle [support](https://www.kaggle.com/contact). However, please note that Kaggle support cannot debug code for you. We show that your account has used GPUs this week (reminder, you have a weekly limit to use). If you have any error messages, please take a screenshot and include them with your request.\n\nThanks!",
    "1129767": "Hello Addison, \n\nThanks for taking time to respond!\n\nThis is only but a dummy code to verify my hypothesis: GPU does not work when I submit my model to be graded. This is not my actual code.\n\nThe code above submits a dummy prediction (0.5) only if GPU is enabled, and make the submission fail otherwise; and indeed it failed, meaning that GPU is not active while the notebook is being graded in your servers.\n\nAnd yes, I have used GPU this week in my account, and it works perfectly, my problem only arises during grading. I believe my notebook timeout (or take significantly more time) because of this problem given the slower nature matrix operations on CPU.\n\nBest regards, \nAbdessalem Boukil."
  },
  "source": "meta"
}