{
  "id": 207449,
  "title": " saint inference  help",
  "url": "/competitions/riiid-test-answer-prediction/discussion/207449",
  "author_name": "",
  "post_date": "2020-12-29T18:24:54.616994900Z",
  "votes": null,
  "comment_count": 7,
  "views": 0,
  "content": "<p>I get 5 times submission error,I may have found the cause, but I don't know how to fix it</p>\n<p>Caleb's emulator reports this error: <br>\nRuntimeError: CUDA error: device-side assert triggered<br>\n But it points to my position embedding code   It mean my postion code is incorrect?<br>\nHow do I solve this problem?😫😫</p>\n<p>Thank you to all those who replied</p>",
  "messages": [
    {
      "id": "1131461",
      "postDate": "12/29/2020 18:24:54",
      "content": "<p>I get 5 times submission error,I may have found the cause, but I don't know how to fix it</p>\n<p>Caleb's emulator reports this error: <br>\nRuntimeError: CUDA error: device-side assert triggered<br>\n But it points to my position embedding code   It mean my postion code is incorrect?<br>\nHow do I solve this problem?😫😫</p>\n<p>Thank you to all those who replied</p>",
      "rawMarkdown": "I get 5 times submission error,I may have found the cause, but I don't know how to fix it\n\n Caleb's emulator reports this error: \nRuntimeError: CUDA error: device-side assert triggered\n But it points to my position embedding code   It mean my postion code is incorrect?\nHow do I solve this problem?😫😫\n\nThank you to all those who replied",
      "votes": null
    },
    {
      "id": "1131516",
      "postDate": "12/29/2020 19:27:37",
      "content": "<p>Disable GPU so you can debug your code, you'll get a more comprehensive error message.</p>",
      "rawMarkdown": "Disable GPU so you can debug your code, you'll get a more comprehensive error message.",
      "votes": null
    },
    {
      "id": "1131526",
      "postDate": "12/29/2020 19:33:24",
      "content": "<p>I got this error often locally. It can be caused by several issues. A common one is size mismatch caused by NaNs.</p>\n<p>If you add this to your preamble:</p>\n<pre><code>import os\nos.environ[\"CUDA_LAUNCH_BLOCKING\"] = \"1\"\n</code></pre>\n<p>You will get a detailed error message.</p>",
      "rawMarkdown": "I got this error often locally. It can be caused by several issues. A common one is size mismatch caused by NaNs.\n\nIf you add this to your preamble:\n```python\nimport os\nos.environ[\"CUDA_LAUNCH_BLOCKING\"] = \"1\"\n```\nYou will get a detailed error message.",
      "votes": null
    },
    {
      "id": "1132027",
      "postDate": "12/30/2020 06:01:09",
      "content": "<p>Your model has some errors in forwarding pass.<br>\nIn GPU mode:<br>\nPyTorch only shows \"CUDA error: device-side assert triggered\".<br>\nThe solution is that change to the CPU mode, and then you can get the error messages.</p>\n<pre><code># device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\ndevice = torch.device(\"cpu\")\n</code></pre>",
      "rawMarkdown": "Your model has some errors in forwarding pass.\nIn GPU mode:\nPyTorch only shows \"CUDA error: device-side assert triggered\".\nThe solution is that change to the CPU mode, and then you can get the error messages.\n\n```\n# device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\ndevice = torch.device(\"cpu\")\n```",
      "votes": null
    },
    {
      "id": "1132094",
      "postDate": "12/30/2020 06:59:01",
      "content": "<p>thank you , I will try</p>",
      "rawMarkdown": "thank you , I will try",
      "votes": null
    },
    {
      "id": "1132149",
      "postDate": "12/30/2020 07:48:01",
      "content": "<p>thank you,my dataset may have some error,I will try this code</p>",
      "rawMarkdown": "thank you,my dataset may have some error,I will try this code",
      "votes": null
    },
    {
      "id": "1132150",
      "postDate": "12/30/2020 07:48:21",
      "content": "<p>thank you👍</p>",
      "rawMarkdown": "thank you👍",
      "votes": null
    },
    {
      "id": "1132335",
      "postDate": "12/30/2020 10:18:04",
      "content": "<p>thanks eveyone ,I located the error,though I can't solve it ,So I solved the code that was giving me error😈😈</p>",
      "rawMarkdown": "thanks eveyone ,I located the error,though I can't solve it ,So I solved the code that was giving me error😈😈",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1131516,
      "author_name": "abdessalemboukil",
      "author_url": "",
      "post_date": "12/29/2020 19:27:37",
      "content": "<p>Disable GPU so you can debug your code, you'll get a more comprehensive error message.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1132094,
          "author_name": "yangxiaoshuai",
          "author_url": "",
          "post_date": "12/30/2020 06:59:01",
          "content": "<p>thank you , I will try</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1131526,
      "author_name": "scaomath",
      "author_url": "",
      "post_date": "12/29/2020 19:33:24",
      "content": "<p>I got this error often locally. It can be caused by several issues. A common one is size mismatch caused by NaNs.</p>\n<p>If you add this to your preamble:</p>\n<pre><code>import os\nos.environ[\"CUDA_LAUNCH_BLOCKING\"] = \"1\"\n</code></pre>\n<p>You will get a detailed error message.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1132150,
          "author_name": "yangxiaoshuai",
          "author_url": "",
          "post_date": "12/30/2020 07:48:21",
          "content": "<p>thank you👍</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1132027,
      "author_name": "m10515009",
      "author_url": "",
      "post_date": "12/30/2020 06:01:09",
      "content": "<p>Your model has some errors in forwarding pass.<br>\nIn GPU mode:<br>\nPyTorch only shows \"CUDA error: device-side assert triggered\".<br>\nThe solution is that change to the CPU mode, and then you can get the error messages.</p>\n<pre><code># device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\ndevice = torch.device(\"cpu\")\n</code></pre>",
      "votes": null,
      "replies": [
        {
          "id": 1132149,
          "author_name": "yangxiaoshuai",
          "author_url": "",
          "post_date": "12/30/2020 07:48:01",
          "content": "<p>thank you,my dataset may have some error,I will try this code</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1132335,
      "author_name": "yangxiaoshuai",
      "author_url": "",
      "post_date": "12/30/2020 10:18:04",
      "content": "<p>thanks eveyone ,I located the error,though I can't solve it ,So I solved the code that was giving me error😈😈</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1131461": "I get 5 times submission error,I may have found the cause, but I don't know how to fix it\n\n Caleb's emulator reports this error: \nRuntimeError: CUDA error: device-side assert triggered\n But it points to my position embedding code   It mean my postion code is incorrect?\nHow do I solve this problem?😫😫\n\nThank you to all those who replied",
    "1131516": "Disable GPU so you can debug your code, you'll get a more comprehensive error message.",
    "1131526": "I got this error often locally. It can be caused by several issues. A common one is size mismatch caused by NaNs.\n\nIf you add this to your preamble:\n```python\nimport os\nos.environ[\"CUDA_LAUNCH_BLOCKING\"] = \"1\"\n```\nYou will get a detailed error message.",
    "1132027": "Your model has some errors in forwarding pass.\nIn GPU mode:\nPyTorch only shows \"CUDA error: device-side assert triggered\".\nThe solution is that change to the CPU mode, and then you can get the error messages.\n\n```\n# device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\ndevice = torch.device(\"cpu\")\n```",
    "1132094": "thank you , I will try",
    "1132149": "thank you,my dataset may have some error,I will try this code",
    "1132150": "thank you👍",
    "1132335": "thanks eveyone ,I located the error,though I can't solve it ,So I solved the code that was giving me error😈😈"
  },
  "source": "meta"
}