{
  "id": 205938,
  "title": "Can a single batch of real api consist of only `lectures`(`content_type`==1) ?",
  "url": "/competitions/riiid-test-answer-prediction/discussion/205938",
  "author_name": "",
  "post_date": "2020-12-22T14:49:17.268221900Z",
  "votes": 2,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I wonder whether a single batch of <code>test_df</code> from test api consists only of <code>lectures</code>, whose <code>content_id</code> == 1</p>",
  "messages": [
    {
      "id": "1122592",
      "postDate": "12/22/2020 14:49:17",
      "content": "<p>I wonder whether a single batch of <code>test_df</code> from test api consists only of <code>lectures</code>, whose <code>content_id</code> == 1</p>",
      "rawMarkdown": "I wonder whether a single batch of `test_df` from test api consists only of `lectures`, whose `content_id` == 1",
      "votes": null
    },
    {
      "id": "1122639",
      "postDate": "12/22/2020 15:29:19",
      "content": "<p>Test it out. You can create a dummy sub based off of the starter notebook that intentionally fails if it encounters the condition you specified above, or submits 0.5 predictions otherwise.</p>\n<p>If it gets to the end successfully, you have the answer.</p>",
      "rawMarkdown": "Test it out. You can create a dummy sub based off of the starter notebook that intentionally fails if it encounters the condition you specified above, or submits 0.5 predictions otherwise.\n\nIf it gets to the end successfully, you have the answer.",
      "votes": null
    },
    {
      "id": "1122794",
      "postDate": "12/22/2020 17:54:59",
      "content": "<p>Possible. Towards the end of the test batches you will be down to a single user - the one with the most records in the test set. Since lectures are never included in task containers with questions and there can only be one task container per user in each batch, if that user has a lecture in its sequence after the records from all the other users have already been served, you would end up with a batch consisting of a single lecture.</p>",
      "rawMarkdown": "Possible. Towards the end of the test batches you will be down to a single user - the one with the most records in the test set. Since lectures are never included in task containers with questions and there can only be one task container per user in each batch, if that user has a lecture in its sequence after the records from all the other users have already been served, you would end up with a batch consisting of a single lecture.",
      "votes": null
    },
    {
      "id": "1122901",
      "postDate": "12/22/2020 19:19:17",
      "content": "<p>Because it's possible, I use <br>\n<code>score = model.predict(inputs) if X_tes.shape[0] &gt; 0 else np.array([], dtype = np.float32)</code></p>",
      "rawMarkdown": "Because it's possible, I use \n``score = model.predict(inputs) if X_tes.shape[0] > 0 else np.array([], dtype = np.float32)``",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1122639,
      "author_name": "authman",
      "author_url": "",
      "post_date": "12/22/2020 15:29:19",
      "content": "<p>Test it out. You can create a dummy sub based off of the starter notebook that intentionally fails if it encounters the condition you specified above, or submits 0.5 predictions otherwise.</p>\n<p>If it gets to the end successfully, you have the answer.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1122794,
      "author_name": "calebeverett",
      "author_url": "",
      "post_date": "12/22/2020 17:54:59",
      "content": "<p>Possible. Towards the end of the test batches you will be down to a single user - the one with the most records in the test set. Since lectures are never included in task containers with questions and there can only be one task container per user in each batch, if that user has a lecture in its sequence after the records from all the other users have already been served, you would end up with a batch consisting of a single lecture.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1122901,
      "author_name": "mamasinkgs",
      "author_url": "",
      "post_date": "12/22/2020 19:19:17",
      "content": "<p>Because it's possible, I use <br>\n<code>score = model.predict(inputs) if X_tes.shape[0] &gt; 0 else np.array([], dtype = np.float32)</code></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1122592": "I wonder whether a single batch of `test_df` from test api consists only of `lectures`, whose `content_id` == 1",
    "1122639": "Test it out. You can create a dummy sub based off of the starter notebook that intentionally fails if it encounters the condition you specified above, or submits 0.5 predictions otherwise.\n\nIf it gets to the end successfully, you have the answer.",
    "1122794": "Possible. Towards the end of the test batches you will be down to a single user - the one with the most records in the test set. Since lectures are never included in task containers with questions and there can only be one task container per user in each batch, if that user has a lecture in its sequence after the records from all the other users have already been served, you would end up with a batch consisting of a single lecture.",
    "1122901": "Because it's possible, I use \n``score = model.predict(inputs) if X_tes.shape[0] > 0 else np.array([], dtype = np.float32)``"
  },
  "source": "meta"
}