{
  "id": 273768,
  "title": "How can I speed up my upload speed! It is slower than my training speed!",
  "url": "/competitions/landmark-retrieval-2021/discussion/273768",
  "author_name": "",
  "post_date": "2021-09-22T11:44:09.065319600Z",
  "votes": 1,
  "comment_count": 7,
  "views": 0,
  "content": "<p>Submitting a model is a nightmare for me. It takes a long time for each submission.<br>\nI cannot use KAGGLE API directly. Because my server is in China.<br>\nBesides VPN, do I have other options… (I had already spent too much money…)</p>",
  "messages": [
    {
      "id": "1520551",
      "postDate": "09/22/2021 11:44:09",
      "content": "<p>Submitting a model is a nightmare for me. It takes a long time for each submission.<br>\nI cannot use KAGGLE API directly. Because my server is in China.<br>\nBesides VPN, do I have other options… (I had already spent too much money…)</p>",
      "rawMarkdown": "Submitting a model is a nightmare for me. It takes a long time for each submission.\nI cannot use KAGGLE API directly. Because my server is in China.\nBesides VPN, do I have other options... (I had already spent too much money...)",
      "votes": null
    },
    {
      "id": "1520874",
      "postDate": "09/22/2021 16:51:14",
      "content": "<p>You may upload your model on any faster working filehoster, join a team and ask your teammates to submit the model.</p>",
      "rawMarkdown": "You may upload your model on any faster working filehoster, join a team and ask your teammates to submit the model.",
      "votes": null
    },
    {
      "id": "1522176",
      "postDate": "09/23/2021 23:37:37",
      "content": "<p>You can upload the weights to COS (Cloud Object Storage), get the direct url of your weights and then create Kaggle dataset by url.</p>",
      "rawMarkdown": "You can upload the weights to COS (Cloud Object Storage), get the direct url of your weights and then create Kaggle dataset by url.",
      "votes": null
    },
    {
      "id": "1522410",
      "postDate": "09/24/2021 06:38:34",
      "content": "<p>Without internet allowed ?</p>",
      "rawMarkdown": "Without internet allowed ?",
      "votes": null
    },
    {
      "id": "1522425",
      "postDate": "09/24/2021 07:02:14",
      "content": "<p>I think the problem he encountered might be that was too slow to upload the model weights. Creating <br>\na dataset via Remote Files URL or Github Repository can bypass such network problems. <br>\nOnce the dataset has been created, it can be used in notebooks without internet access.</p>",
      "rawMarkdown": "I think the problem he encountered might be that was too slow to upload the model weights. Creating \na dataset via Remote Files URL or Github Repository can bypass such network problems. \nOnce the dataset has been created, it can be used in notebooks without internet access.",
      "votes": null
    },
    {
      "id": "1522446",
      "postDate": "09/24/2021 07:26:03",
      "content": "<p>Ok, if that is the case, it works (only that github is limited to 100mb). I first thought there was a problem with the submission, not sure if that's equivalent to uploading problems.We'll see if its a solution for <a href=\"https://www.kaggle.com/xiaowangiiiii\" target=\"_blank\">@xiaowangiiiii</a> .</p>",
      "rawMarkdown": "Ok, if that is the case, it works (only that github is limited to 100mb). I first thought there was a problem with the submission, not sure if that's equivalent to uploading problems.We'll see if its a solution for @xiaowangiiiii .",
      "votes": null
    },
    {
      "id": "1522465",
      "postDate": "09/24/2021 07:46:23",
      "content": "<p>Thats good idea. <br>\nBtw, Do I need to buy COS?</p>",
      "rawMarkdown": "Thats good idea. \nBtw, Do I need to buy COS?",
      "votes": null
    },
    {
      "id": "1522468",
      "postDate": "09/24/2021 08:00:42",
      "content": "<p>Using COS usually needs to pay, but I mentioned it only because it's the only way I know that can generate file direct links. <br>\nYou can look for other free ways to get file URLs, or create datasets via Github Repository is also a good choice.</p>",
      "rawMarkdown": "Using COS usually needs to pay, but I mentioned it only because it's the only way I know that can generate file direct links. \nYou can look for other free ways to get file URLs, or create datasets via Github Repository is also a good choice.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1520874,
      "author_name": "alexanderbader",
      "author_url": "",
      "post_date": "09/22/2021 16:51:14",
      "content": "<p>You may upload your model on any faster working filehoster, join a team and ask your teammates to submit the model.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1522176,
      "author_name": "apretrue",
      "author_url": "",
      "post_date": "09/23/2021 23:37:37",
      "content": "<p>You can upload the weights to COS (Cloud Object Storage), get the direct url of your weights and then create Kaggle dataset by url.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1522410,
          "author_name": "alexanderbader",
          "author_url": "",
          "post_date": "09/24/2021 06:38:34",
          "content": "<p>Without internet allowed ?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1522425,
          "author_name": "apretrue",
          "author_url": "",
          "post_date": "09/24/2021 07:02:14",
          "content": "<p>I think the problem he encountered might be that was too slow to upload the model weights. Creating <br>\na dataset via Remote Files URL or Github Repository can bypass such network problems. <br>\nOnce the dataset has been created, it can be used in notebooks without internet access.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1522446,
          "author_name": "alexanderbader",
          "author_url": "",
          "post_date": "09/24/2021 07:26:03",
          "content": "<p>Ok, if that is the case, it works (only that github is limited to 100mb). I first thought there was a problem with the submission, not sure if that's equivalent to uploading problems.We'll see if its a solution for <a href=\"https://www.kaggle.com/xiaowangiiiii\" target=\"_blank\">@xiaowangiiiii</a> .</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1522465,
          "author_name": "xiaowangiiiii",
          "author_url": "",
          "post_date": "09/24/2021 07:46:23",
          "content": "<p>Thats good idea. <br>\nBtw, Do I need to buy COS?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1522468,
          "author_name": "apretrue",
          "author_url": "",
          "post_date": "09/24/2021 08:00:42",
          "content": "<p>Using COS usually needs to pay, but I mentioned it only because it's the only way I know that can generate file direct links. <br>\nYou can look for other free ways to get file URLs, or create datasets via Github Repository is also a good choice.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1520551": "Submitting a model is a nightmare for me. It takes a long time for each submission.\nI cannot use KAGGLE API directly. Because my server is in China.\nBesides VPN, do I have other options... (I had already spent too much money...)",
    "1520874": "You may upload your model on any faster working filehoster, join a team and ask your teammates to submit the model.",
    "1522176": "You can upload the weights to COS (Cloud Object Storage), get the direct url of your weights and then create Kaggle dataset by url.",
    "1522410": "Without internet allowed ?",
    "1522425": "I think the problem he encountered might be that was too slow to upload the model weights. Creating \na dataset via Remote Files URL or Github Repository can bypass such network problems. \nOnce the dataset has been created, it can be used in notebooks without internet access.",
    "1522446": "Ok, if that is the case, it works (only that github is limited to 100mb). I first thought there was a problem with the submission, not sure if that's equivalent to uploading problems.We'll see if its a solution for @xiaowangiiiii .",
    "1522465": "Thats good idea. \nBtw, Do I need to buy COS?",
    "1522468": "Using COS usually needs to pay, but I mentioned it only because it's the only way I know that can generate file direct links. \nYou can look for other free ways to get file URLs, or create datasets via Github Repository is also a good choice."
  },
  "source": "meta"
}