{
  "id": 116745,
  "title": "GPU limit before submisson",
  "url": "/competitions/rsna-intracranial-hemorrhage-detection/discussion/116745",
  "author_name": "Claudio Fanconi",
  "post_date": "2019-11-11T07:51:28.813000",
  "votes": 0,
  "comment_count": 7,
  "views": 0,
  "content": "<p>Hey guys I have a problem... I have already used up all my 30 gpu hours this week. However, unfortunately, my last notebook which performs inference and ensembling had a minor issue in the last cell and thus does not produce a submission file.</p>\n\n<p>How would you guys get around this problem?\nI am trying to run it on a CPU, as it´s only predicting and not training. However, I am not confident that it´ll work, because of the very large test data set.</p>\n\n<p>At this point, I am kinda very frustrated with the GPU limits that have been imposed... I am really annoyed if I need to use a lower score, because of my own stupidity and the unnability to correct it..</p>",
  "messages": [
    {
      "id": 670392,
      "postDate": "2019-11-11T12:17:51.310Z",
      "content": "<p>Try to use google-colab</p>",
      "rawMarkdown": "Try to use google-colab",
      "votes": 1,
      "replies": [
        {
          "id": 670407,
          "postDate": "2019-11-11T12:36:07.470Z",
          "content": "<p>Can you commit from google collab? And can you download the whole dataset on the notebook on colab? <a href=\"/kio9999\">@kio9999</a> </p>",
          "rawMarkdown": "Can you commit from google collab? And can you download the whole dataset on the notebook on colab? @kio9999 "
        },
        {
          "id": 670633,
          "postDate": "2019-11-11T17:08:59.773Z",
          "content": "<p>If you only need to perform inference, you can try downloading the test set from <a href=\"https://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/discussion/115855\">here</a>.</p>",
          "rawMarkdown": "If you only need to perform inference, you can try downloading the test set from [here](https://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/discussion/115855)."
        },
        {
          "id": 670732,
          "postDate": "2019-11-11T19:33:51.893Z",
          "content": "<p>Thank you <a href=\"/cparrarojas\">@cparrarojas</a> ! However, how can I commit it to kaggle? Because if the committing fails, then I don't believe it can be submitted...</p>",
          "rawMarkdown": "Thank you @cparrarojas ! However, how can I commit it to kaggle? Because if the committing fails, then I don't believe it can be submitted...\n"
        },
        {
          "id": 670736,
          "postDate": "2019-11-11T19:36:55.563Z",
          "content": "<p>You can upload submission file without committing </p>",
          "rawMarkdown": "You can upload submission file without committing ",
          "votes": 1
        },
        {
          "id": 670747,
          "postDate": "2019-11-11T19:46:00.470Z",
          "content": "<p><a href=\"/fanconic\">@fanconic</a> This is not a kernels-only competition. You can simply upload the CSV as <a href=\"/kio9999\">@kio9999</a> says.</p>",
          "rawMarkdown": "@fanconic This is not a kernels-only competition. You can simply upload the CSV as @kio9999 says.",
          "votes": 1
        },
        {
          "id": 670749,
          "postDate": "2019-11-11T19:46:15.473Z",
          "content": "<p>Alright thanks! I was already panicking! </p>",
          "rawMarkdown": "Alright thanks! I was already panicking! "
        }
      ]
    },
    {
      "id": 670238,
      "postDate": "2019-11-11T07:51:28.813Z",
      "content": "<p>Hey guys I have a problem... I have already used up all my 30 gpu hours this week. However, unfortunately, my last notebook which performs inference and ensembling had a minor issue in the last cell and thus does not produce a submission file.</p>\n\n<p>How would you guys get around this problem?\nI am trying to run it on a CPU, as it´s only predicting and not training. However, I am not confident that it´ll work, because of the very large test data set.</p>\n\n<p>At this point, I am kinda very frustrated with the GPU limits that have been imposed... I am really annoyed if I need to use a lower score, because of my own stupidity and the unnability to correct it..</p>",
      "rawMarkdown": "Hey guys I have a problem... I have already used up all my 30 gpu hours this week. However, unfortunately, my last notebook which performs inference and ensembling had a minor issue in the last cell and thus does not produce a submission file.\n\nHow would you guys get around this problem?\nI am trying to run it on a CPU, as it´s only predicting and not training. However, I am not confident that it´ll work, because of the very large test data set.\n\nAt this point, I am kinda very frustrated with the GPU limits that have been imposed... I am really annoyed if I need to use a lower score, because of my own stupidity and the unnability to correct it..\n"
    }
  ],
  "comments": [
    {
      "id": 670392,
      "author_name": "Suraj Soni",
      "author_url": "",
      "post_date": "2019-11-11T12:17:51.310000",
      "content": "<p>Try to use google-colab</p>",
      "votes": 1,
      "replies": [
        {
          "id": 670407,
          "author_name": "Claudio Fanconi",
          "author_url": "",
          "post_date": "2019-11-11T12:36:07.470000",
          "content": "<p>Can you commit from google collab? And can you download the whole dataset on the notebook on colab? <a href=\"/kio9999\">@kio9999</a> </p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 670633,
          "author_name": "César Parra Rojas",
          "author_url": "",
          "post_date": "2019-11-11T17:08:59.773000",
          "content": "<p>If you only need to perform inference, you can try downloading the test set from <a href=\"https://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/discussion/115855\">here</a>.</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 670732,
          "author_name": "Claudio Fanconi",
          "author_url": "",
          "post_date": "2019-11-11T19:33:51.893000",
          "content": "<p>Thank you <a href=\"/cparrarojas\">@cparrarojas</a> ! However, how can I commit it to kaggle? Because if the committing fails, then I don't believe it can be submitted...</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 670736,
          "author_name": "Suraj Soni",
          "author_url": "",
          "post_date": "2019-11-11T19:36:55.563000",
          "content": "<p>You can upload submission file without committing </p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 670747,
          "author_name": "César Parra Rojas",
          "author_url": "",
          "post_date": "2019-11-11T19:46:00.470000",
          "content": "<p><a href=\"/fanconic\">@fanconic</a> This is not a kernels-only competition. You can simply upload the CSV as <a href=\"/kio9999\">@kio9999</a> says.</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 670749,
          "author_name": "Claudio Fanconi",
          "author_url": "",
          "post_date": "2019-11-11T19:46:15.473000",
          "content": "<p>Alright thanks! I was already panicking! </p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "670392": "Try to use google-colab",
    "670238": "Hey guys I have a problem... I have already used up all my 30 gpu hours this week. However, unfortunately, my last notebook which performs inference and ensembling had a minor issue in the last cell and thus does not produce a submission file.\n\nHow would you guys get around this problem?\nI am trying to run it on a CPU, as it´s only predicting and not training. However, I am not confident that it´ll work, because of the very large test data set.\n\nAt this point, I am kinda very frustrated with the GPU limits that have been imposed... I am really annoyed if I need to use a lower score, because of my own stupidity and the unnability to correct it..\n"
  }
}