{
  "id": 230479,
  "title": "Bring HuggingFace models to Kaggle easily",
  "url": "/competitions/coleridgeinitiative-show-us-the-data/discussion/230479",
  "author_name": "",
  "post_date": "2021-04-04T00:13:08.919863800Z",
  "votes": 16,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi everyone,</p>\n<p>Since using HuggingFace's pretrained models in kernel competitions has become pretty common for NLP competitions, I wrote a very small package to make that a bit easier since I didn't know of any simple way to do that.</p>\n<p>The API is pretty simple: It is just one function that takes the model page URL as input :)</p>\n<p>The advantage is that the models don't have to be loaded into memory and the upload to Kaggle is just a flag.</p>\n<p>The package essentially downloads the files to disk and can directly upload them to your Kaggle account. These are added as a dataset.</p>\n<p>This is the link to the repo:<br>\n<a href=\"https://github.com/shabie/mytransformers\" target=\"_blank\"><strong>https://github.com/shabie/mytransformers</strong></a></p>\n<p>On how to use the files, kindly have a look at this tutorial that actually gave me the idea to create this:<br>\n<a href=\"https://www.kaggle.com/jamesmcguigan/coleridge-huggingface-question-answering\" target=\"_blank\">https://www.kaggle.com/jamesmcguigan/coleridge-huggingface-question-answering</a>.</p>\n<p>I'd appreciate any and all feedback. Please know I wrote this in a day so it will surely have several rough edges.</p>",
  "messages": [
    {
      "id": "1262210",
      "postDate": "04/04/2021 00:13:08",
      "content": "<p>Hi everyone,</p>\n<p>Since using HuggingFace's pretrained models in kernel competitions has become pretty common for NLP competitions, I wrote a very small package to make that a bit easier since I didn't know of any simple way to do that.</p>\n<p>The API is pretty simple: It is just one function that takes the model page URL as input :)</p>\n<p>The advantage is that the models don't have to be loaded into memory and the upload to Kaggle is just a flag.</p>\n<p>The package essentially downloads the files to disk and can directly upload them to your Kaggle account. These are added as a dataset.</p>\n<p>This is the link to the repo:<br>\n<a href=\"https://github.com/shabie/mytransformers\" target=\"_blank\"><strong>https://github.com/shabie/mytransformers</strong></a></p>\n<p>On how to use the files, kindly have a look at this tutorial that actually gave me the idea to create this:<br>\n<a href=\"https://www.kaggle.com/jamesmcguigan/coleridge-huggingface-question-answering\" target=\"_blank\">https://www.kaggle.com/jamesmcguigan/coleridge-huggingface-question-answering</a>.</p>\n<p>I'd appreciate any and all feedback. Please know I wrote this in a day so it will surely have several rough edges.</p>",
      "rawMarkdown": "Hi everyone,\n\nSince using HuggingFace's pretrained models in kernel competitions has become pretty common for NLP competitions, I wrote a very small package to make that a bit easier since I didn't know of any simple way to do that.\n\nThe API is pretty simple: It is just one function that takes the model page URL as input :)\n\nThe advantage is that the models don't have to be loaded into memory and the upload to Kaggle is just a flag.\n\nThe package essentially downloads the files to disk and can directly upload them to your Kaggle account. These are added as a dataset.\n\nThis is the link to the repo:\n[**https://github.com/shabie/mytransformers**](https://github.com/shabie/mytransformers)\n\nOn how to use the files, kindly have a look at this tutorial that actually gave me the idea to create this:\nhttps://www.kaggle.com/jamesmcguigan/coleridge-huggingface-question-answering.\n\nI'd appreciate any and all feedback. Please know I wrote this in a day so it will surely have several rough edges.",
      "votes": null
    },
    {
      "id": "1289819",
      "postDate": "05/01/2021 12:02:21",
      "content": "<p>This is super helpful. Thanks for sharing!!!</p>",
      "rawMarkdown": "This is super helpful. Thanks for sharing!!!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1289819,
      "author_name": "kartikbhargav",
      "author_url": "",
      "post_date": "05/01/2021 12:02:21",
      "content": "<p>This is super helpful. Thanks for sharing!!!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1262210": "Hi everyone,\n\nSince using HuggingFace's pretrained models in kernel competitions has become pretty common for NLP competitions, I wrote a very small package to make that a bit easier since I didn't know of any simple way to do that.\n\nThe API is pretty simple: It is just one function that takes the model page URL as input :)\n\nThe advantage is that the models don't have to be loaded into memory and the upload to Kaggle is just a flag.\n\nThe package essentially downloads the files to disk and can directly upload them to your Kaggle account. These are added as a dataset.\n\nThis is the link to the repo:\n[**https://github.com/shabie/mytransformers**](https://github.com/shabie/mytransformers)\n\nOn how to use the files, kindly have a look at this tutorial that actually gave me the idea to create this:\nhttps://www.kaggle.com/jamesmcguigan/coleridge-huggingface-question-answering.\n\nI'd appreciate any and all feedback. Please know I wrote this in a day so it will surely have several rough edges.",
    "1289819": "This is super helpful. Thanks for sharing!!!"
  },
  "source": "meta"
}