{
  "id": 145179,
  "title": "Notebook restarts due to memory error on gpu session but not tpu?",
  "url": "/competitions/flower-classification-with-tpus/discussion/145179",
  "author_name": "",
  "post_date": "2020-04-22T06:36:10.919140100Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>why is this the case? </p>",
  "messages": [
    {
      "id": "816166",
      "postDate": "04/22/2020 06:36:10",
      "content": "<p>why is this the case? </p>",
      "rawMarkdown": "why is this the case?",
      "votes": null
    },
    {
      "id": "816989",
      "postDate": "04/22/2020 18:50:58",
      "content": "<p>When you set a notebook to use a TPU, it runs on newer, more powerful hardware than a notebook running with a GPU is on. A TPU notebook simply has more memory available than a GPU notebook does.</p>\n\n<p>(for the processor)\nIn a GPU session, you can click on the statistics to see that the GPU has 16 GB of memory available. From the <a href=\"https://www.kaggle.com/docs/tpu\">Kaggle TPU Documentation</a>, specifically the hardware section, you can see that a TPU has 128 GB available, a significant improvement.</p>\n\n<p>(for the CPU)\nJust comparing the statistics available within the notebook, a notebook with GPU was 13 GB RAM available and one with TPU has 16 GB available.</p>",
      "rawMarkdown": "When you set a notebook to use a TPU, it runs on newer, more powerful hardware than a notebook running with a GPU is on. A TPU notebook simply has more memory available than a GPU notebook does.\n\n(for the processor)\nIn a GPU session, you can click on the statistics to see that the GPU has 16 GB of memory available. From the [Kaggle TPU Documentation](https://www.kaggle.com/docs/tpu), specifically the hardware section, you can see that a TPU has 128 GB available, a significant improvement.\n\n(for the CPU)\nJust comparing the statistics available within the notebook, a notebook with GPU was 13 GB RAM available and one with TPU has 16 GB available.",
      "votes": null
    },
    {
      "id": "817093",
      "postDate": "04/22/2020 20:37:51",
      "content": "<p>aaah that makes a lot of sense.  far simpler question than I realized.  in the stats the tpu memory isn't listed so I assumed it had the same amount as gpu.  thank you! </p>",
      "rawMarkdown": "aaah that makes a lot of sense.  far simpler question than I realized.  in the stats the tpu memory isn't listed so I assumed it had the same amount as gpu.  thank you!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 816989,
      "author_name": "awanderingsoul",
      "author_url": "",
      "post_date": "04/22/2020 18:50:58",
      "content": "<p>When you set a notebook to use a TPU, it runs on newer, more powerful hardware than a notebook running with a GPU is on. A TPU notebook simply has more memory available than a GPU notebook does.</p>\n\n<p>(for the processor)\nIn a GPU session, you can click on the statistics to see that the GPU has 16 GB of memory available. From the <a href=\"https://www.kaggle.com/docs/tpu\">Kaggle TPU Documentation</a>, specifically the hardware section, you can see that a TPU has 128 GB available, a significant improvement.</p>\n\n<p>(for the CPU)\nJust comparing the statistics available within the notebook, a notebook with GPU was 13 GB RAM available and one with TPU has 16 GB available.</p>",
      "votes": null,
      "replies": [
        {
          "id": 817093,
          "author_name": "rashanarshad",
          "author_url": "",
          "post_date": "04/22/2020 20:37:51",
          "content": "<p>aaah that makes a lot of sense.  far simpler question than I realized.  in the stats the tpu memory isn't listed so I assumed it had the same amount as gpu.  thank you! </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "816166": "why is this the case?",
    "816989": "When you set a notebook to use a TPU, it runs on newer, more powerful hardware than a notebook running with a GPU is on. A TPU notebook simply has more memory available than a GPU notebook does.\n\n(for the processor)\nIn a GPU session, you can click on the statistics to see that the GPU has 16 GB of memory available. From the [Kaggle TPU Documentation](https://www.kaggle.com/docs/tpu), specifically the hardware section, you can see that a TPU has 128 GB available, a significant improvement.\n\n(for the CPU)\nJust comparing the statistics available within the notebook, a notebook with GPU was 13 GB RAM available and one with TPU has 16 GB available.",
    "817093": "aaah that makes a lot of sense.  far simpler question than I realized.  in the stats the tpu memory isn't listed so I assumed it had the same amount as gpu.  thank you!"
  },
  "source": "meta"
}