{
  "id": 163623,
  "title": "hello Colab TPU User's",
  "url": "/competitions/alaska2-image-steganalysis/discussion/163623",
  "author_name": "Prashant Arora",
  "post_date": "2020-07-02T19:17:55.737000",
  "votes": 3,
  "comment_count": 8,
  "views": 0,
  "content": "<p>All those colab users who work on TPUs , i have created something for you guys.\nNotebook : <a href=\"https://www.kaggle.com/prashantarorat/alaska-image-stegnalysis\">https://www.kaggle.com/prashantarorat/alaska-image-stegnalysis</a>\nI have created a dataset of google cloud resource files of all images in the dataset and along with it a train.csv file for all labels for these images.\nYou can find the dataset here : <a href=\"https://www.kaggle.com/prashantarorat/gcs-path-files-alaska-image-steganalysis\">https://www.kaggle.com/prashantarorat/gcs-path-files-alaska-image-steganalysis</a>.\nNote : Kindly read the dataset description once.\nI hope it will be helpful to someone.</p>",
  "messages": [
    {
      "id": 922368,
      "postDate": "2020-07-10T04:04:24.350Z",
      "content": "<p>Hi,\nThe easy way is to start the CPU kernel on Kaggle. Create GCS dataset path, copy these paths in your colab notebook and use the TPU to train the model there. That is what I am doing to train it on colab. I can go for 15 epochs in one session.</p>",
      "rawMarkdown": "Hi,\nThe easy way is to start the CPU kernel on Kaggle. Create GCS dataset path, copy these paths in your colab notebook and use the TPU to train the model there. That is what I am doing to train it on colab. I can go for 15 epochs in one session.",
      "votes": 1,
      "replies": [
        {
          "id": 922380,
          "postDate": "2020-07-10T04:27:42.300Z",
          "content": "<p>I am trying this method, but got the error: UnavailableError: Socket closed</p>",
          "rawMarkdown": "I am trying this method, but got the error: UnavailableError: Socket closed"
        },
        {
          "id": 922425,
          "postDate": "2020-07-10T05:15:39.893Z",
          "content": "<p>Not sure why you get the error. It is running fine for me. The error is I think related to TPU socket being closed. But, if you are connected properly then you should be able to use it properly. </p>",
          "rawMarkdown": "Not sure why you get the error. It is running fine for me. The error is I think related to TPU socket being closed. But, if you are connected properly then you should be able to use it properly. "
        }
      ]
    },
    {
      "id": 912864,
      "postDate": "2020-07-02T19:17:55.737Z",
      "content": "<p>All those colab users who work on TPUs , i have created something for you guys.\nNotebook : <a href=\"https://www.kaggle.com/prashantarorat/alaska-image-stegnalysis\">https://www.kaggle.com/prashantarorat/alaska-image-stegnalysis</a>\nI have created a dataset of google cloud resource files of all images in the dataset and along with it a train.csv file for all labels for these images.\nYou can find the dataset here : <a href=\"https://www.kaggle.com/prashantarorat/gcs-path-files-alaska-image-steganalysis\">https://www.kaggle.com/prashantarorat/gcs-path-files-alaska-image-steganalysis</a>.\nNote : Kindly read the dataset description once.\nI hope it will be helpful to someone.</p>",
      "rawMarkdown": "All those colab users who work on TPUs , i have created something for you guys.\nNotebook : https://www.kaggle.com/prashantarorat/alaska-image-stegnalysis\nI have created a dataset of google cloud resource files of all images in the dataset and along with it a train.csv file for all labels for these images.\nYou can find the dataset here : https://www.kaggle.com/prashantarorat/gcs-path-files-alaska-image-steganalysis.\nNote : Kindly read the dataset description once.\nI hope it will be helpful to someone.",
      "votes": 2
    },
    {
      "id": 929090,
      "postDate": "2020-07-14T13:01:06.253Z",
      "content": "<p>Here the dataset , GCS Paths i am using for training ,  might be helpful to some one.</p>\n\n<p>train_filenames = ['gs://kds-1cff6e52fc1e4f94dbf6063e728f4e6910594b7755191f1e86211c65/train00-10000.tfrec',# 0-10k records dataset\n                   'gs://kds-1cff6e52fc1e4f94dbf6063e728f4e6910594b7755191f1e86211c65/train01-10000.tfrec',\n                   'gs://kds-1cff6e52fc1e4f94dbf6063e728f4e6910594b7755191f1e86211c65/train02-10000.tfrec',\n                   'gs://kds-1cff6e52fc1e4f94dbf6063e728f4e6910594b7755191f1e86211c65/train03-10000.tfrec',\n                   'gs://kds-bb4a6e7977bd5f2855fee24a530ac9c59b2de7802c2252c97f6f7c46/train04-10000.tfrec',# 10-20k records dataset\n                   'gs://kds-bb4a6e7977bd5f2855fee24a530ac9c59b2de7802c2252c97f6f7c46/train05-10000.tfrec',\n                   'gs://kds-bb4a6e7977bd5f2855fee24a530ac9c59b2de7802c2252c97f6f7c46/train06-10000.tfrec',\n                   'gs://kds-bb4a6e7977bd5f2855fee24a530ac9c59b2de7802c2252c97f6f7c46/train07-10000.tfrec',\n                   'gs://kds-2aec829a2d0e6b19c0e72eaef7af9790171b107f7cad500a2e350503/train00-10000.tfrec',# 20-30k records dataset\n                   'gs://kds-2aec829a2d0e6b19c0e72eaef7af9790171b107f7cad500a2e350503/train01-10000.tfrec', \n                   'gs://kds-2aec829a2d0e6b19c0e72eaef7af9790171b107f7cad500a2e350503/train02-10000.tfrec',\n                   'gs://kds-2aec829a2d0e6b19c0e72eaef7af9790171b107f7cad500a2e350503/train03-10000.tfrec',\n                   'gs://kds-2faf8abf60467c7fd1b16c2e8b01d9dd9caf8917478391bc4abc251d/train00-10000.tfrec',# 30-40k records dataset\n                   'gs://kds-2faf8abf60467c7fd1b16c2e8b01d9dd9caf8917478391bc4abc251d/train01-10000.tfrec', \n                   'gs://kds-2faf8abf60467c7fd1b16c2e8b01d9dd9caf8917478391bc4abc251d/train02-10000.tfrec', \n                   'gs://kds-2faf8abf60467c7fd1b16c2e8b01d9dd9caf8917478391bc4abc251d/train03-10000.tfrec',\n                   'gs://kds-6dd4c0f662ca6b031664dff0b79e3095deb60e4a17e2374de321157e/train00-10000.tfrec',# 40-50k records dataset\n                   'gs://kds-6dd4c0f662ca6b031664dff0b79e3095deb60e4a17e2374de321157e/train01-10000.tfrec',\n                   'gs://kds-6dd4c0f662ca6b031664dff0b79e3095deb60e4a17e2374de321157e/train02-10000.tfrec',\n                   'gs://kds-6dd4c0f662ca6b031664dff0b79e3095deb60e4a17e2374de321157e/train03-10000.tfrec',\n                   'gs://kds-58a3c2b02fed339d2fcf18261258911a0decf6a7168daa5f7b70d86e/train00-10000.tfrec',# 50-60k records dataset\n                   'gs://kds-58a3c2b02fed339d2fcf18261258911a0decf6a7168daa5f7b70d86e/train01-10000.tfrec',\n                   'gs://kds-58a3c2b02fed339d2fcf18261258911a0decf6a7168daa5f7b70d86e/train02-10000.tfrec',\n                   'gs://kds-58a3c2b02fed339d2fcf18261258911a0decf6a7168daa5f7b70d86e/train03-10000.tfrec',\n                   'gs://kds-009d573b8fee6b4015937f0f66798b8f637397621568372d329704e0/train00-10000.tfrec',# 60-70k records dataset\n                   'gs://kds-009d573b8fee6b4015937f0f66798b8f637397621568372d329704e0/train01-10000.tfrec',\n                   'gs://kds-009d573b8fee6b4015937f0f66798b8f637397621568372d329704e0/train02-10000.tfrec',\n                   'gs://kds-009d573b8fee6b4015937f0f66798b8f637397621568372d329704e0/train03-10000.tfrec',\n                   'gs://kds-ebb46d1d90d28db6d1547555e4b0c3ff63407bc5240f785d9eddea9e/train00-10000.tfrec',# 70-75k records dataset\n                   'gs://kds-ebb46d1d90d28db6d1547555e4b0c3ff63407bc5240f785d9eddea9e/train01-10000.tfrec']</p>",
      "rawMarkdown": "Here the dataset , GCS Paths i am using for training ,  might be helpful to some one.\n\ntrain_filenames = ['gs://kds-1cff6e52fc1e4f94dbf6063e728f4e6910594b7755191f1e86211c65/train00-10000.tfrec',# 0-10k records dataset\n                   'gs://kds-1cff6e52fc1e4f94dbf6063e728f4e6910594b7755191f1e86211c65/train01-10000.tfrec',\n                   'gs://kds-1cff6e52fc1e4f94dbf6063e728f4e6910594b7755191f1e86211c65/train02-10000.tfrec',\n                   'gs://kds-1cff6e52fc1e4f94dbf6063e728f4e6910594b7755191f1e86211c65/train03-10000.tfrec',\n                   'gs://kds-bb4a6e7977bd5f2855fee24a530ac9c59b2de7802c2252c97f6f7c46/train04-10000.tfrec',# 10-20k records dataset\n                   'gs://kds-bb4a6e7977bd5f2855fee24a530ac9c59b2de7802c2252c97f6f7c46/train05-10000.tfrec',\n                   'gs://kds-bb4a6e7977bd5f2855fee24a530ac9c59b2de7802c2252c97f6f7c46/train06-10000.tfrec',\n                   'gs://kds-bb4a6e7977bd5f2855fee24a530ac9c59b2de7802c2252c97f6f7c46/train07-10000.tfrec',\n                   'gs://kds-2aec829a2d0e6b19c0e72eaef7af9790171b107f7cad500a2e350503/train00-10000.tfrec',# 20-30k records dataset\n                   'gs://kds-2aec829a2d0e6b19c0e72eaef7af9790171b107f7cad500a2e350503/train01-10000.tfrec', \n                   'gs://kds-2aec829a2d0e6b19c0e72eaef7af9790171b107f7cad500a2e350503/train02-10000.tfrec',\n                   'gs://kds-2aec829a2d0e6b19c0e72eaef7af9790171b107f7cad500a2e350503/train03-10000.tfrec',\n                   'gs://kds-2faf8abf60467c7fd1b16c2e8b01d9dd9caf8917478391bc4abc251d/train00-10000.tfrec',# 30-40k records dataset\n                   'gs://kds-2faf8abf60467c7fd1b16c2e8b01d9dd9caf8917478391bc4abc251d/train01-10000.tfrec', \n                   'gs://kds-2faf8abf60467c7fd1b16c2e8b01d9dd9caf8917478391bc4abc251d/train02-10000.tfrec', \n                   'gs://kds-2faf8abf60467c7fd1b16c2e8b01d9dd9caf8917478391bc4abc251d/train03-10000.tfrec',\n                   'gs://kds-6dd4c0f662ca6b031664dff0b79e3095deb60e4a17e2374de321157e/train00-10000.tfrec',# 40-50k records dataset\n                   'gs://kds-6dd4c0f662ca6b031664dff0b79e3095deb60e4a17e2374de321157e/train01-10000.tfrec',\n                   'gs://kds-6dd4c0f662ca6b031664dff0b79e3095deb60e4a17e2374de321157e/train02-10000.tfrec',\n                   'gs://kds-6dd4c0f662ca6b031664dff0b79e3095deb60e4a17e2374de321157e/train03-10000.tfrec',\n                   'gs://kds-58a3c2b02fed339d2fcf18261258911a0decf6a7168daa5f7b70d86e/train00-10000.tfrec',# 50-60k records dataset\n                   'gs://kds-58a3c2b02fed339d2fcf18261258911a0decf6a7168daa5f7b70d86e/train01-10000.tfrec',\n                   'gs://kds-58a3c2b02fed339d2fcf18261258911a0decf6a7168daa5f7b70d86e/train02-10000.tfrec',\n                   'gs://kds-58a3c2b02fed339d2fcf18261258911a0decf6a7168daa5f7b70d86e/train03-10000.tfrec',\n                   'gs://kds-009d573b8fee6b4015937f0f66798b8f637397621568372d329704e0/train00-10000.tfrec',# 60-70k records dataset\n                   'gs://kds-009d573b8fee6b4015937f0f66798b8f637397621568372d329704e0/train01-10000.tfrec',\n                   'gs://kds-009d573b8fee6b4015937f0f66798b8f637397621568372d329704e0/train02-10000.tfrec',\n                   'gs://kds-009d573b8fee6b4015937f0f66798b8f637397621568372d329704e0/train03-10000.tfrec',\n                   'gs://kds-ebb46d1d90d28db6d1547555e4b0c3ff63407bc5240f785d9eddea9e/train00-10000.tfrec',# 70-75k records dataset\n                   'gs://kds-ebb46d1d90d28db6d1547555e4b0c3ff63407bc5240f785d9eddea9e/train01-10000.tfrec']"
    },
    {
      "id": 922336,
      "postDate": "2020-07-10T03:18:09.423Z",
      "content": "<p><a href=\"/prashantarorat\">@prashantarorat</a> The dataset link is 404 now...</p>",
      "rawMarkdown": "@prashantarorat The dataset link is 404 now..."
    },
    {
      "id": 913470,
      "postDate": "2020-07-03T08:27:27.107Z",
      "content": "<p>Have you encountered TPU out of memory issue?</p>",
      "rawMarkdown": "Have you encountered TPU out of memory issue?\n",
      "isDeleted": true,
      "replies": [
        {
          "id": 913639,
          "postDate": "2020-07-03T10:26:55.750Z",
          "content": "<p>Nope.</p>",
          "rawMarkdown": "Nope."
        },
        {
          "id": 922370,
          "postDate": "2020-07-10T04:06:15.367Z",
          "content": "<p>You need a smaller batch size if you encounter such an issue. On colab, we have TPU V2 while on Kaggle we have TPU V3 which has more memory. This will resolve your issue. </p>",
          "rawMarkdown": "You need a smaller batch size if you encounter such an issue. On colab, we have TPU V2 while on Kaggle we have TPU V3 which has more memory. This will resolve your issue. "
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 922368,
      "author_name": "Urvish",
      "author_url": "",
      "post_date": "2020-07-10T04:04:24.350000",
      "content": "<p>Hi,\nThe easy way is to start the CPU kernel on Kaggle. Create GCS dataset path, copy these paths in your colab notebook and use the TPU to train the model there. That is what I am doing to train it on colab. I can go for 15 epochs in one session.</p>",
      "votes": 1,
      "replies": [
        {
          "id": 922380,
          "author_name": "ld",
          "author_url": "",
          "post_date": "2020-07-10T04:27:42.300000",
          "content": "<p>I am trying this method, but got the error: UnavailableError: Socket closed</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 922425,
          "author_name": "Urvish",
          "author_url": "",
          "post_date": "2020-07-10T05:15:39.893000",
          "content": "<p>Not sure why you get the error. It is running fine for me. The error is I think related to TPU socket being closed. But, if you are connected properly then you should be able to use it properly. </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 929090,
      "author_name": "Prashant Arora",
      "author_url": "",
      "post_date": "2020-07-14T13:01:06.253000",
      "content": "<p>Here the dataset , GCS Paths i am using for training ,  might be helpful to some one.</p>\n\n<p>train_filenames = ['gs://kds-1cff6e52fc1e4f94dbf6063e728f4e6910594b7755191f1e86211c65/train00-10000.tfrec',# 0-10k records dataset\n                   'gs://kds-1cff6e52fc1e4f94dbf6063e728f4e6910594b7755191f1e86211c65/train01-10000.tfrec',\n                   'gs://kds-1cff6e52fc1e4f94dbf6063e728f4e6910594b7755191f1e86211c65/train02-10000.tfrec',\n                   'gs://kds-1cff6e52fc1e4f94dbf6063e728f4e6910594b7755191f1e86211c65/train03-10000.tfrec',\n                   'gs://kds-bb4a6e7977bd5f2855fee24a530ac9c59b2de7802c2252c97f6f7c46/train04-10000.tfrec',# 10-20k records dataset\n                   'gs://kds-bb4a6e7977bd5f2855fee24a530ac9c59b2de7802c2252c97f6f7c46/train05-10000.tfrec',\n                   'gs://kds-bb4a6e7977bd5f2855fee24a530ac9c59b2de7802c2252c97f6f7c46/train06-10000.tfrec',\n                   'gs://kds-bb4a6e7977bd5f2855fee24a530ac9c59b2de7802c2252c97f6f7c46/train07-10000.tfrec',\n                   'gs://kds-2aec829a2d0e6b19c0e72eaef7af9790171b107f7cad500a2e350503/train00-10000.tfrec',# 20-30k records dataset\n                   'gs://kds-2aec829a2d0e6b19c0e72eaef7af9790171b107f7cad500a2e350503/train01-10000.tfrec', \n                   'gs://kds-2aec829a2d0e6b19c0e72eaef7af9790171b107f7cad500a2e350503/train02-10000.tfrec',\n                   'gs://kds-2aec829a2d0e6b19c0e72eaef7af9790171b107f7cad500a2e350503/train03-10000.tfrec',\n                   'gs://kds-2faf8abf60467c7fd1b16c2e8b01d9dd9caf8917478391bc4abc251d/train00-10000.tfrec',# 30-40k records dataset\n                   'gs://kds-2faf8abf60467c7fd1b16c2e8b01d9dd9caf8917478391bc4abc251d/train01-10000.tfrec', \n                   'gs://kds-2faf8abf60467c7fd1b16c2e8b01d9dd9caf8917478391bc4abc251d/train02-10000.tfrec', \n                   'gs://kds-2faf8abf60467c7fd1b16c2e8b01d9dd9caf8917478391bc4abc251d/train03-10000.tfrec',\n                   'gs://kds-6dd4c0f662ca6b031664dff0b79e3095deb60e4a17e2374de321157e/train00-10000.tfrec',# 40-50k records dataset\n                   'gs://kds-6dd4c0f662ca6b031664dff0b79e3095deb60e4a17e2374de321157e/train01-10000.tfrec',\n                   'gs://kds-6dd4c0f662ca6b031664dff0b79e3095deb60e4a17e2374de321157e/train02-10000.tfrec',\n                   'gs://kds-6dd4c0f662ca6b031664dff0b79e3095deb60e4a17e2374de321157e/train03-10000.tfrec',\n                   'gs://kds-58a3c2b02fed339d2fcf18261258911a0decf6a7168daa5f7b70d86e/train00-10000.tfrec',# 50-60k records dataset\n                   'gs://kds-58a3c2b02fed339d2fcf18261258911a0decf6a7168daa5f7b70d86e/train01-10000.tfrec',\n                   'gs://kds-58a3c2b02fed339d2fcf18261258911a0decf6a7168daa5f7b70d86e/train02-10000.tfrec',\n                   'gs://kds-58a3c2b02fed339d2fcf18261258911a0decf6a7168daa5f7b70d86e/train03-10000.tfrec',\n                   'gs://kds-009d573b8fee6b4015937f0f66798b8f637397621568372d329704e0/train00-10000.tfrec',# 60-70k records dataset\n                   'gs://kds-009d573b8fee6b4015937f0f66798b8f637397621568372d329704e0/train01-10000.tfrec',\n                   'gs://kds-009d573b8fee6b4015937f0f66798b8f637397621568372d329704e0/train02-10000.tfrec',\n                   'gs://kds-009d573b8fee6b4015937f0f66798b8f637397621568372d329704e0/train03-10000.tfrec',\n                   'gs://kds-ebb46d1d90d28db6d1547555e4b0c3ff63407bc5240f785d9eddea9e/train00-10000.tfrec',# 70-75k records dataset\n                   'gs://kds-ebb46d1d90d28db6d1547555e4b0c3ff63407bc5240f785d9eddea9e/train01-10000.tfrec']</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 922336,
      "author_name": "ld",
      "author_url": "",
      "post_date": "2020-07-10T03:18:09.423000",
      "content": "<p><a href=\"/prashantarorat\">@prashantarorat</a> The dataset link is 404 now...</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 913470,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-07-03T08:27:27.107000",
      "content": "<p>Have you encountered TPU out of memory issue?</p>",
      "votes": 0,
      "replies": [
        {
          "id": 913639,
          "author_name": "Prashant Arora",
          "author_url": "",
          "post_date": "2020-07-03T10:26:55.750000",
          "content": "<p>Nope.</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 922370,
          "author_name": "Urvish",
          "author_url": "",
          "post_date": "2020-07-10T04:06:15.367000",
          "content": "<p>You need a smaller batch size if you encounter such an issue. On colab, we have TPU V2 while on Kaggle we have TPU V3 which has more memory. This will resolve your issue. </p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "922368": "Hi,\nThe easy way is to start the CPU kernel on Kaggle. Create GCS dataset path, copy these paths in your colab notebook and use the TPU to train the model there. That is what I am doing to train it on colab. I can go for 15 epochs in one session.",
    "912864": "All those colab users who work on TPUs , i have created something for you guys.\nNotebook : https://www.kaggle.com/prashantarorat/alaska-image-stegnalysis\nI have created a dataset of google cloud resource files of all images in the dataset and along with it a train.csv file for all labels for these images.\nYou can find the dataset here : https://www.kaggle.com/prashantarorat/gcs-path-files-alaska-image-steganalysis.\nNote : Kindly read the dataset description once.\nI hope it will be helpful to someone.",
    "929090": "Here the dataset , GCS Paths i am using for training ,  might be helpful to some one.\n\ntrain_filenames = ['gs://kds-1cff6e52fc1e4f94dbf6063e728f4e6910594b7755191f1e86211c65/train00-10000.tfrec',# 0-10k records dataset\n                   'gs://kds-1cff6e52fc1e4f94dbf6063e728f4e6910594b7755191f1e86211c65/train01-10000.tfrec',\n                   'gs://kds-1cff6e52fc1e4f94dbf6063e728f4e6910594b7755191f1e86211c65/train02-10000.tfrec',\n                   'gs://kds-1cff6e52fc1e4f94dbf6063e728f4e6910594b7755191f1e86211c65/train03-10000.tfrec',\n                   'gs://kds-bb4a6e7977bd5f2855fee24a530ac9c59b2de7802c2252c97f6f7c46/train04-10000.tfrec',# 10-20k records dataset\n                   'gs://kds-bb4a6e7977bd5f2855fee24a530ac9c59b2de7802c2252c97f6f7c46/train05-10000.tfrec',\n                   'gs://kds-bb4a6e7977bd5f2855fee24a530ac9c59b2de7802c2252c97f6f7c46/train06-10000.tfrec',\n                   'gs://kds-bb4a6e7977bd5f2855fee24a530ac9c59b2de7802c2252c97f6f7c46/train07-10000.tfrec',\n                   'gs://kds-2aec829a2d0e6b19c0e72eaef7af9790171b107f7cad500a2e350503/train00-10000.tfrec',# 20-30k records dataset\n                   'gs://kds-2aec829a2d0e6b19c0e72eaef7af9790171b107f7cad500a2e350503/train01-10000.tfrec', \n                   'gs://kds-2aec829a2d0e6b19c0e72eaef7af9790171b107f7cad500a2e350503/train02-10000.tfrec',\n                   'gs://kds-2aec829a2d0e6b19c0e72eaef7af9790171b107f7cad500a2e350503/train03-10000.tfrec',\n                   'gs://kds-2faf8abf60467c7fd1b16c2e8b01d9dd9caf8917478391bc4abc251d/train00-10000.tfrec',# 30-40k records dataset\n                   'gs://kds-2faf8abf60467c7fd1b16c2e8b01d9dd9caf8917478391bc4abc251d/train01-10000.tfrec', \n                   'gs://kds-2faf8abf60467c7fd1b16c2e8b01d9dd9caf8917478391bc4abc251d/train02-10000.tfrec', \n                   'gs://kds-2faf8abf60467c7fd1b16c2e8b01d9dd9caf8917478391bc4abc251d/train03-10000.tfrec',\n                   'gs://kds-6dd4c0f662ca6b031664dff0b79e3095deb60e4a17e2374de321157e/train00-10000.tfrec',# 40-50k records dataset\n                   'gs://kds-6dd4c0f662ca6b031664dff0b79e3095deb60e4a17e2374de321157e/train01-10000.tfrec',\n                   'gs://kds-6dd4c0f662ca6b031664dff0b79e3095deb60e4a17e2374de321157e/train02-10000.tfrec',\n                   'gs://kds-6dd4c0f662ca6b031664dff0b79e3095deb60e4a17e2374de321157e/train03-10000.tfrec',\n                   'gs://kds-58a3c2b02fed339d2fcf18261258911a0decf6a7168daa5f7b70d86e/train00-10000.tfrec',# 50-60k records dataset\n                   'gs://kds-58a3c2b02fed339d2fcf18261258911a0decf6a7168daa5f7b70d86e/train01-10000.tfrec',\n                   'gs://kds-58a3c2b02fed339d2fcf18261258911a0decf6a7168daa5f7b70d86e/train02-10000.tfrec',\n                   'gs://kds-58a3c2b02fed339d2fcf18261258911a0decf6a7168daa5f7b70d86e/train03-10000.tfrec',\n                   'gs://kds-009d573b8fee6b4015937f0f66798b8f637397621568372d329704e0/train00-10000.tfrec',# 60-70k records dataset\n                   'gs://kds-009d573b8fee6b4015937f0f66798b8f637397621568372d329704e0/train01-10000.tfrec',\n                   'gs://kds-009d573b8fee6b4015937f0f66798b8f637397621568372d329704e0/train02-10000.tfrec',\n                   'gs://kds-009d573b8fee6b4015937f0f66798b8f637397621568372d329704e0/train03-10000.tfrec',\n                   'gs://kds-ebb46d1d90d28db6d1547555e4b0c3ff63407bc5240f785d9eddea9e/train00-10000.tfrec',# 70-75k records dataset\n                   'gs://kds-ebb46d1d90d28db6d1547555e4b0c3ff63407bc5240f785d9eddea9e/train01-10000.tfrec']",
    "922336": "@prashantarorat The dataset link is 404 now...",
    "913470": "Have you encountered TPU out of memory issue?\n"
  }
}