{
  "id": 415670,
  "title": "Is there anyone who has implemented a preprocessor using PyTorch?",
  "url": "/competitions/asl-fingerspelling/discussion/415670",
  "author_name": "canlion",
  "post_date": "2023-06-07T13:43:26.044000",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Nice to meet you again on Kaggle today. Wishing you a pleasant competition. </p>\n<p>🤕 I am currently using the five daily submission opportunities to observe the process of [modifying a line of code, submitting, scoring failure] five times. I have been trying to identify the cause of the problem by exploring which operations are available and which ones are not. However, there are no available operations. </p>\n<p>If anyone has implemented the data normalization process using PyTorch, I would appreciate some advice.<br>\nThank you for your time.</p>",
  "messages": [
    {
      "id": 2291465,
      "postDate": "2023-06-07T15:09:33.527Z",
      "content": "<p>Preprocessing with PyTorch didn't work in previous ASL competition, unlikely to work here - too much memory requirement.  <br>\nWe can still train models with PyTorch but for submission, preprocessing should be in Tensorflow. </p>",
      "rawMarkdown": "Preprocessing with PyTorch didn't work in previous ASL competition, unlikely to work here - too much memory requirement.  \nWe can still train models with PyTorch but for submission, preprocessing should be in Tensorflow. ",
      "votes": 2,
      "replies": [
        {
          "id": 2291483,
          "postDate": "2023-06-07T15:19:49.510Z",
          "content": "<p>I used pytorch in previous ASL competition. Perhaps I didn't need a lot of memory because I performed simple preprocessing steps. The advice to perform preprocessing with tensorflow sounds like a good idea. I think I can convert the pytorch model to a tensorflow model, connect it with the preprocessing layers, and create a tflite file. I will try it out tomorrow. Thank you for the advice. 👍</p>",
          "rawMarkdown": "I used pytorch in previous ASL competition. Perhaps I didn't need a lot of memory because I performed simple preprocessing steps. The advice to perform preprocessing with tensorflow sounds like a good idea. I think I can convert the pytorch model to a tensorflow model, connect it with the preprocessing layers, and create a tflite file. I will try it out tomorrow. Thank you for the advice. 👍",
          "votes": 1
        }
      ]
    },
    {
      "id": 2291340,
      "postDate": "2023-06-07T13:43:26.043Z",
      "content": "<p>Nice to meet you again on Kaggle today. Wishing you a pleasant competition. </p>\n<p>🤕 I am currently using the five daily submission opportunities to observe the process of [modifying a line of code, submitting, scoring failure] five times. I have been trying to identify the cause of the problem by exploring which operations are available and which ones are not. However, there are no available operations. </p>\n<p>If anyone has implemented the data normalization process using PyTorch, I would appreciate some advice.<br>\nThank you for your time.</p>",
      "rawMarkdown": "Nice to meet you again on Kaggle today. Wishing you a pleasant competition. \n\n🤕 I am currently using the five daily submission opportunities to observe the process of [modifying a line of code, submitting, scoring failure] five times. I have been trying to identify the cause of the problem by exploring which operations are available and which ones are not. However, there are no available operations. \n\nIf anyone has implemented the data normalization process using PyTorch, I would appreciate some advice.\nThank you for your time.",
      "votes": 2
    }
  ],
  "comments": [
    {
      "id": 2291465,
      "author_name": "RB",
      "author_url": "",
      "post_date": "2023-06-07T15:09:33.527000",
      "content": "<p>Preprocessing with PyTorch didn't work in previous ASL competition, unlikely to work here - too much memory requirement.  <br>\nWe can still train models with PyTorch but for submission, preprocessing should be in Tensorflow. </p>",
      "votes": 2,
      "replies": [
        {
          "id": 2291483,
          "author_name": "canlion",
          "author_url": "",
          "post_date": "2023-06-07T15:19:49.510000",
          "content": "<p>I used pytorch in previous ASL competition. Perhaps I didn't need a lot of memory because I performed simple preprocessing steps. The advice to perform preprocessing with tensorflow sounds like a good idea. I think I can convert the pytorch model to a tensorflow model, connect it with the preprocessing layers, and create a tflite file. I will try it out tomorrow. Thank you for the advice. 👍</p>",
          "votes": 1,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2291465": "Preprocessing with PyTorch didn't work in previous ASL competition, unlikely to work here - too much memory requirement.  \nWe can still train models with PyTorch but for submission, preprocessing should be in Tensorflow. ",
    "2291340": "Nice to meet you again on Kaggle today. Wishing you a pleasant competition. \n\n🤕 I am currently using the five daily submission opportunities to observe the process of [modifying a line of code, submitting, scoring failure] five times. I have been trying to identify the cause of the problem by exploring which operations are available and which ones are not. However, there are no available operations. \n\nIf anyone has implemented the data normalization process using PyTorch, I would appreciate some advice.\nThank you for your time."
  }
}