{
  "id": 401200,
  "title": "why only models with size smaller than 20 MB passed",
  "url": "/competitions/asl-signs/discussion/401200",
  "author_name": "",
  "post_date": "2023-04-12T07:50:40.208750Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi!</p>\n<p>I use a transformer as an encoder to recognize labels. When I train a model larger than 20 MB the submission will present 'scoring error'. I am confused as the actual limitation is 40 MB for each model. </p>\n<p>I also test the running time of models larger than 20MB: with tflite_runtime(2.11.0) in CPU environment every parquet costs about 0.09 seconds. I guess 0.09 is the time limitation for me. </p>\n<p>Could you please help me solve the problem? Thanks in advance!!!</p>",
  "messages": [
    {
      "id": "2219020",
      "postDate": "04/12/2023 07:50:40",
      "content": "<p>Hi!</p>\n<p>I use a transformer as an encoder to recognize labels. When I train a model larger than 20 MB the submission will present 'scoring error'. I am confused as the actual limitation is 40 MB for each model. </p>\n<p>I also test the running time of models larger than 20MB: with tflite_runtime(2.11.0) in CPU environment every parquet costs about 0.09 seconds. I guess 0.09 is the time limitation for me. </p>\n<p>Could you please help me solve the problem? Thanks in advance!!!</p>",
      "rawMarkdown": "Hi!\n\nI use a transformer as an encoder to recognize labels. When I train a model larger than 20 MB the submission will present 'scoring error'. I am confused as the actual limitation is 40 MB for each model. \n\nI also test the running time of models larger than 20MB: with tflite_runtime(2.11.0) in CPU environment every parquet costs about 0.09 seconds. I guess 0.09 is the time limitation for me. \n\nCould you please help me solve the problem? Thanks in advance!!!",
      "votes": null
    },
    {
      "id": "2219035",
      "postDate": "04/12/2023 08:13:59",
      "content": "<p>Did you quantize your model? You may exceed the time limit of 100ms. I've faced the same.  <a href=\"https://www.kaggle.com/competitions/asl-signs/discussion/395425\" target=\"_blank\">That</a> solved my issue</p>",
      "rawMarkdown": "Did you quantize your model? You may exceed the time limit of 100ms. I've faced the same.  [That](https://www.kaggle.com/competitions/asl-signs/discussion/395425) solved my issue",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2219035,
      "author_name": "meowmeowmeowmeowmeow",
      "author_url": "",
      "post_date": "04/12/2023 08:13:59",
      "content": "<p>Did you quantize your model? You may exceed the time limit of 100ms. I've faced the same.  <a href=\"https://www.kaggle.com/competitions/asl-signs/discussion/395425\" target=\"_blank\">That</a> solved my issue</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2219020": "Hi!\n\nI use a transformer as an encoder to recognize labels. When I train a model larger than 20 MB the submission will present 'scoring error'. I am confused as the actual limitation is 40 MB for each model. \n\nI also test the running time of models larger than 20MB: with tflite_runtime(2.11.0) in CPU environment every parquet costs about 0.09 seconds. I guess 0.09 is the time limitation for me. \n\nCould you please help me solve the problem? Thanks in advance!!!",
    "2219035": "Did you quantize your model? You may exceed the time limit of 100ms. I've faced the same.  [That](https://www.kaggle.com/competitions/asl-signs/discussion/395425) solved my issue"
  },
  "source": "meta"
}