{
  "id": 311666,
  "title": "Colab pro TPU  now only one session?",
  "url": "/competitions/happy-whale-and-dolphin/discussion/311666",
  "author_name": "",
  "post_date": "2022-03-08T06:31:45.338508300Z",
  "votes": 1,
  "comment_count": 4,
  "views": 0,
  "content": "<p>I used to run two sessions of TPU notebook.</p>\n<p>Now only one is allowed?</p>",
  "messages": [
    {
      "id": "1715555",
      "postDate": "03/08/2022 06:31:45",
      "content": "<p>I used to run two sessions of TPU notebook.</p>\n<p>Now only one is allowed?</p>",
      "rawMarkdown": "I used to run two sessions of TPU notebook.\n\nNow only one is allowed?",
      "votes": null
    },
    {
      "id": "1715579",
      "postDate": "03/08/2022 06:55:29",
      "content": "<p>Kaggle makes your notebook wait in the queue if there is heavy usage. TPU is popular these days.</p>",
      "rawMarkdown": "Kaggle makes your notebook wait in the queue if there is heavy usage. TPU is popular these days.",
      "votes": null
    },
    {
      "id": "1715648",
      "postDate": "03/08/2022 08:03:07",
      "content": "<p>my quota on kaggle this week almost done.  what I talk about is Colab Pro, which I have bought.</p>",
      "rawMarkdown": "my quota on kaggle this week almost done.  what I talk about is Colab Pro, which I have bought.",
      "votes": null
    },
    {
      "id": "1715657",
      "postDate": "03/08/2022 08:20:43",
      "content": "<p>Probably luck and as Colab keeps track of your usage and alters availability of services based on how much you use them. I have run 2 TPU at once but I tend not to as I don't want them terminated by Colab. </p>\n<p>I have found however that using a TPU and GPU and/or a CPU all at once doesn't seem to be an issue. While that may not help with this project, it does help if you are doing different projects or parts of projects (train/infer) that use different accelerators at once.</p>\n<p>I've been thrashing the colab TPU (as we all have) and if I do a long (15hr) run and then immediately try to run another fold, I am at serious risk of it just stopping after a few hrs. </p>\n<p>It's taking quite awhile to train 5 folds of a model. (I have Colab Pro)</p>",
      "rawMarkdown": "Probably luck and as Colab keeps track of your usage and alters availability of services based on how much you use them. I have run 2 TPU at once but I tend not to as I don't want them terminated by Colab. \n\nI have found however that using a TPU and GPU and/or a CPU all at once doesn't seem to be an issue. While that may not help with this project, it does help if you are doing different projects or parts of projects (train/infer) that use different accelerators at once.\n\nI've been thrashing the colab TPU (as we all have) and if I do a long (15hr) run and then immediately try to run another fold, I am at serious risk of it just stopping after a few hrs. \n\nIt's taking quite awhile to train 5 folds of a model. (I have Colab Pro)",
      "votes": null
    },
    {
      "id": "1715704",
      "postDate": "03/08/2022 09:24:15",
      "content": "<p>Thank you for your reply. I  bought colab Pro not long ago.</p>",
      "rawMarkdown": "Thank you for your reply. I  bought colab Pro not long ago.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1715579,
      "author_name": "vivovinco",
      "author_url": "",
      "post_date": "03/08/2022 06:55:29",
      "content": "<p>Kaggle makes your notebook wait in the queue if there is heavy usage. TPU is popular these days.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1715648,
          "author_name": "dragonzhang",
          "author_url": "",
          "post_date": "03/08/2022 08:03:07",
          "content": "<p>my quota on kaggle this week almost done.  what I talk about is Colab Pro, which I have bought.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1715657,
      "author_name": "mutantspore",
      "author_url": "",
      "post_date": "03/08/2022 08:20:43",
      "content": "<p>Probably luck and as Colab keeps track of your usage and alters availability of services based on how much you use them. I have run 2 TPU at once but I tend not to as I don't want them terminated by Colab. </p>\n<p>I have found however that using a TPU and GPU and/or a CPU all at once doesn't seem to be an issue. While that may not help with this project, it does help if you are doing different projects or parts of projects (train/infer) that use different accelerators at once.</p>\n<p>I've been thrashing the colab TPU (as we all have) and if I do a long (15hr) run and then immediately try to run another fold, I am at serious risk of it just stopping after a few hrs. </p>\n<p>It's taking quite awhile to train 5 folds of a model. (I have Colab Pro)</p>",
      "votes": null,
      "replies": [
        {
          "id": 1715704,
          "author_name": "dragonzhang",
          "author_url": "",
          "post_date": "03/08/2022 09:24:15",
          "content": "<p>Thank you for your reply. I  bought colab Pro not long ago.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1715555": "I used to run two sessions of TPU notebook.\n\nNow only one is allowed?",
    "1715579": "Kaggle makes your notebook wait in the queue if there is heavy usage. TPU is popular these days.",
    "1715648": "my quota on kaggle this week almost done.  what I talk about is Colab Pro, which I have bought.",
    "1715657": "Probably luck and as Colab keeps track of your usage and alters availability of services based on how much you use them. I have run 2 TPU at once but I tend not to as I don't want them terminated by Colab. \n\nI have found however that using a TPU and GPU and/or a CPU all at once doesn't seem to be an issue. While that may not help with this project, it does help if you are doing different projects or parts of projects (train/infer) that use different accelerators at once.\n\nI've been thrashing the colab TPU (as we all have) and if I do a long (15hr) run and then immediately try to run another fold, I am at serious risk of it just stopping after a few hrs. \n\nIt's taking quite awhile to train 5 folds of a model. (I have Colab Pro)",
    "1715704": "Thank you for your reply. I  bought colab Pro not long ago."
  },
  "source": "meta"
}