{
  "id": 207702,
  "title": "Install EfficientNet for PyTorch for offline inference",
  "url": "/competitions/cassava-leaf-disease-classification/discussion/207702",
  "author_name": "Michal Pitr",
  "post_date": "2020-12-30T22:31:42.600000",
  "votes": 3,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Since PyTorch doesn't come bundled with EfficientNet, we have to work around this inconvenience. For the submission, we are required to turn the internet off, which certainly doesn't make things easier, but there's a pretty simple solution! There are several PyTorch implementations available on GitHub, so grab whichever one you like. I'm using Luke Melas'.</p>\n<p>We can add it to our notebook using the Add Data button in the top right corner. Once we do so, it will be available even when the notebook is run offline. <br>\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4720590%2Ff55f3e27659ea5d11fdc15f0ac56e6aa%2FScreenshot%20from%202020-12-30%2023-18-34.png?generation=1609366748284203&amp;alt=media\" alt=\"\"></p>\n<p>We can search by URL or simply by name for the repo we want.<br>\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4720590%2Fc46a5ea3a7103d6cd4d3c3ac33f90263%2FScreenshot%20from%202020-12-30%2023-22-23.png?generation=1609366986389648&amp;alt=media\" alt=\"\"></p>\n<p>Once we add it, it will appear in our <code>./input/</code> directory. From here we simply install it via pip like so<br>\n<code>!pip install path</code> and import like any python module.</p>",
  "messages": [
    {
      "id": 1133062,
      "postDate": "2020-12-30T22:31:42.600Z",
      "content": "<p>Since PyTorch doesn't come bundled with EfficientNet, we have to work around this inconvenience. For the submission, we are required to turn the internet off, which certainly doesn't make things easier, but there's a pretty simple solution! There are several PyTorch implementations available on GitHub, so grab whichever one you like. I'm using Luke Melas'.</p>\n<p>We can add it to our notebook using the Add Data button in the top right corner. Once we do so, it will be available even when the notebook is run offline. <br>\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4720590%2Ff55f3e27659ea5d11fdc15f0ac56e6aa%2FScreenshot%20from%202020-12-30%2023-18-34.png?generation=1609366748284203&amp;alt=media\" alt=\"\"></p>\n<p>We can search by URL or simply by name for the repo we want.<br>\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4720590%2Fc46a5ea3a7103d6cd4d3c3ac33f90263%2FScreenshot%20from%202020-12-30%2023-22-23.png?generation=1609366986389648&amp;alt=media\" alt=\"\"></p>\n<p>Once we add it, it will appear in our <code>./input/</code> directory. From here we simply install it via pip like so<br>\n<code>!pip install path</code> and import like any python module.</p>",
      "rawMarkdown": "Since PyTorch doesn't come bundled with EfficientNet, we have to work around this inconvenience. For the submission, we are required to turn the internet off, which certainly doesn't make things easier, but there's a pretty simple solution! There are several PyTorch implementations available on GitHub, so grab whichever one you like. I'm using Luke Melas'.\n\nWe can add it to our notebook using the Add Data button in the top right corner. Once we do so, it will be available even when the notebook is run offline. \n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4720590%2Ff55f3e27659ea5d11fdc15f0ac56e6aa%2FScreenshot%20from%202020-12-30%2023-18-34.png?generation=1609366748284203&alt=media)\n\nWe can search by URL or simply by name for the repo we want.\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4720590%2Fc46a5ea3a7103d6cd4d3c3ac33f90263%2FScreenshot%20from%202020-12-30%2023-22-23.png?generation=1609366986389648&alt=media)\n\nOnce we add it, it will appear in our `./input/` directory. From here we simply install it via pip like so\n`!pip install path` and import like any python module.\n\n\n\n",
      "votes": 3
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1133062": "Since PyTorch doesn't come bundled with EfficientNet, we have to work around this inconvenience. For the submission, we are required to turn the internet off, which certainly doesn't make things easier, but there's a pretty simple solution! There are several PyTorch implementations available on GitHub, so grab whichever one you like. I'm using Luke Melas'.\n\nWe can add it to our notebook using the Add Data button in the top right corner. Once we do so, it will be available even when the notebook is run offline. \n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4720590%2Ff55f3e27659ea5d11fdc15f0ac56e6aa%2FScreenshot%20from%202020-12-30%2023-18-34.png?generation=1609366748284203&alt=media)\n\nWe can search by URL or simply by name for the repo we want.\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4720590%2Fc46a5ea3a7103d6cd4d3c3ac33f90263%2FScreenshot%20from%202020-12-30%2023-22-23.png?generation=1609366986389648&alt=media)\n\nOnce we add it, it will appear in our `./input/` directory. From here we simply install it via pip like so\n`!pip install path` and import like any python module.\n\n\n\n"
  }
}