{
  "id": 125627,
  "title": "Interim quick submission to public LB.",
  "url": "/competitions/tensorflow2-question-answering/discussion/125627",
  "author_name": "",
  "post_date": "2020-01-12T10:03:28.260602600Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>In this competition I am training remotely using the TPU quota provided by GCP. Inference is done on GPU in a Kaggle kernel. As we reach the last 10 days of the comp, even the generous allocation of 30 hours of free GPU time from Kaggle looks like it might be a bit tight for me.</p>\n\n<p>Consequently, I have started doing inference offline as well, and submitting csv predictions directly through this <a href=\"https://www.kaggle.com/kenkrige/nq-direct-submit\">short kernel</a>. This has the added advantage of providing feedback in a couple of minutes on the public LB with no kaggle GPU time consumed.</p>",
  "messages": [
    {
      "id": "716822",
      "postDate": "01/12/2020 10:03:28",
      "content": "<p>In this competition I am training remotely using the TPU quota provided by GCP. Inference is done on GPU in a Kaggle kernel. As we reach the last 10 days of the comp, even the generous allocation of 30 hours of free GPU time from Kaggle looks like it might be a bit tight for me.</p>\n\n<p>Consequently, I have started doing inference offline as well, and submitting csv predictions directly through this <a href=\"https://www.kaggle.com/kenkrige/nq-direct-submit\">short kernel</a>. This has the added advantage of providing feedback in a couple of minutes on the public LB with no kaggle GPU time consumed.</p>",
      "rawMarkdown": "In this competition I am training remotely using the TPU quota provided by GCP. Inference is done on GPU in a Kaggle kernel. As we reach the last 10 days of the comp, even the generous allocation of 30 hours of free GPU time from Kaggle looks like it might be a bit tight for me.\n\nConsequently, I have started doing inference offline as well, and submitting csv predictions directly through this [short kernel](https://www.kaggle.com/kenkrige/nq-direct-submit). This has the added advantage of providing feedback in a couple of minutes on the public LB with no kaggle GPU time consumed.",
      "votes": null
    },
    {
      "id": "716876",
      "postDate": "01/12/2020 11:59:25",
      "content": "<p>Thank you for very useful kernel.</p>",
      "rawMarkdown": "Thank you for very useful kernel.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 716876,
      "author_name": "renxingkai",
      "author_url": "",
      "post_date": "01/12/2020 11:59:25",
      "content": "<p>Thank you for very useful kernel.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "716822": "In this competition I am training remotely using the TPU quota provided by GCP. Inference is done on GPU in a Kaggle kernel. As we reach the last 10 days of the comp, even the generous allocation of 30 hours of free GPU time from Kaggle looks like it might be a bit tight for me.\n\nConsequently, I have started doing inference offline as well, and submitting csv predictions directly through this [short kernel](https://www.kaggle.com/kenkrige/nq-direct-submit). This has the added advantage of providing feedback in a couple of minutes on the public LB with no kaggle GPU time consumed.",
    "716876": "Thank you for very useful kernel."
  },
  "source": "meta"
}