{
  "id": 552026,
  "title": "How to create sequence data on each symbol when submitting?",
  "url": "/competitions/jane-street-real-time-market-data-forecasting/discussion/552026",
  "author_name": "",
  "post_date": "2024-12-17T07:38:31.309012200Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I want to submit RNNs. <br>\nBut it seems that test data is given timestamp by timestamp,  which means <code>JSInferenceServer</code> will repeatly call <code>predict</code>, and every time pass a test dataset with new <code>time_id</code> to the <code>predict</code>.<br>\nMay I ask if my understanding of the test data api is correct? If so, how can I create sequence data on each symbol with multiple step(i.e. <code>time_step=64</code>) and batch_size(i.e. <code>batch_size=64</code>)? </p>",
  "messages": [
    {
      "id": "3073996",
      "postDate": "12/17/2024 07:38:31",
      "content": "<p>I want to submit RNNs. <br>\nBut it seems that test data is given timestamp by timestamp,  which means <code>JSInferenceServer</code> will repeatly call <code>predict</code>, and every time pass a test dataset with new <code>time_id</code> to the <code>predict</code>.<br>\nMay I ask if my understanding of the test data api is correct? If so, how can I create sequence data on each symbol with multiple step(i.e. <code>time_step=64</code>) and batch_size(i.e. <code>batch_size=64</code>)? </p>",
      "rawMarkdown": "I want to submit RNNs. \nBut it seems that test data is given timestamp by timestamp,  which means `JSInferenceServer` will repeatly call `predict`, and every time pass a test dataset with new `time_id` to the `predict`.\nMay I ask if my understanding of the test data api is correct? If so, how can I create sequence data on each symbol with multiple step(i.e. `time_step=64`) and batch_size(i.e. `batch_size=64`)?",
      "votes": null
    },
    {
      "id": "3074005",
      "postDate": "12/17/2024 07:56:56",
      "content": "<p>create your own cache. </p>\n<p>see here: <a href=\"https://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/543081\" target=\"_blank\">https://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/543081</a></p>",
      "rawMarkdown": "create your own cache. \n\nsee here: https://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/543081",
      "votes": null
    },
    {
      "id": "3074058",
      "postDate": "12/17/2024 09:19:00",
      "content": "<p>Thanks! Does <code>row_id</code> increase continuously on different test data sets? If so, we can store the current test data and combine them with the next test data.</p>",
      "rawMarkdown": "Thanks! Does `row_id` increase continuously on different test data sets? If so, we can store the current test data and combine them with the next test data.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3074005,
      "author_name": "shiyili",
      "author_url": "",
      "post_date": "12/17/2024 07:56:56",
      "content": "<p>create your own cache. </p>\n<p>see here: <a href=\"https://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/543081\" target=\"_blank\">https://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/543081</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 3074058,
          "author_name": "zzhstaboy",
          "author_url": "",
          "post_date": "12/17/2024 09:19:00",
          "content": "<p>Thanks! Does <code>row_id</code> increase continuously on different test data sets? If so, we can store the current test data and combine them with the next test data.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3073996": "I want to submit RNNs. \nBut it seems that test data is given timestamp by timestamp,  which means `JSInferenceServer` will repeatly call `predict`, and every time pass a test dataset with new `time_id` to the `predict`.\nMay I ask if my understanding of the test data api is correct? If so, how can I create sequence data on each symbol with multiple step(i.e. `time_step=64`) and batch_size(i.e. `batch_size=64`)?",
    "3074005": "create your own cache. \n\nsee here: https://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/543081",
    "3074058": "Thanks! Does `row_id` increase continuously on different test data sets? If so, we can store the current test data and combine them with the next test data."
  },
  "source": "meta"
}