{
  "id": 127832,
  "title": "AWS EC2 and SageMaker; which should I choose?",
  "url": "/competitions/deepfake-detection-challenge/discussion/127832",
  "author_name": "",
  "post_date": "2020-01-27T04:29:38.933792Z",
  "votes": 2,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Since the whole training dataset of this competition is freaking huge, I've considered using cloud-based services for training models. There are so many options provided by Google, Amazon, etc., but I will try AWS first. Now, I have to make another decision, using EC2 or SageMaker. I know SageMaker provides us more interesting things, proper configurations, etc. with a higher price, yet do I really need these advantages to just training my models? Is EC2 the better choice for a beginner like me?</p>\n\n<p>Thanks in advance!</p>",
  "messages": [
    {
      "id": "730078",
      "postDate": "01/27/2020 04:29:38",
      "content": "<p>Since the whole training dataset of this competition is freaking huge, I've considered using cloud-based services for training models. There are so many options provided by Google, Amazon, etc., but I will try AWS first. Now, I have to make another decision, using EC2 or SageMaker. I know SageMaker provides us more interesting things, proper configurations, etc. with a higher price, yet do I really need these advantages to just training my models? Is EC2 the better choice for a beginner like me?</p>\n\n<p>Thanks in advance!</p>",
      "rawMarkdown": "Since the whole training dataset of this competition is freaking huge, I've considered using cloud-based services for training models. There are so many options provided by Google, Amazon, etc., but I will try AWS first. Now, I have to make another decision, using EC2 or SageMaker. I know SageMaker provides us more interesting things, proper configurations, etc. with a higher price, yet do I really need these advantages to just training my models? Is EC2 the better choice for a beginner like me?\n\nThanks in advance!",
      "votes": null
    },
    {
      "id": "730154",
      "postDate": "01/27/2020 06:49:55",
      "content": "<p>I guess it's more a question of preference rather than anything else. Honestly, the whole AWS/google ecosystem is a bit complicated so it might take a while to get used to. Without prior experience, I would advise SageMaker over plain EC2 instances. Why? Because you have everything pre-installed and notebooks at your fingertips ...which makes it similar to kaggle in that respect.</p>",
      "rawMarkdown": "I guess it's more a question of preference rather than anything else. Honestly, the whole AWS/google ecosystem is a bit complicated so it might take a while to get used to. Without prior experience, I would advise SageMaker over plain EC2 instances. Why? Because you have everything pre-installed and notebooks at your fingertips ...which makes it similar to kaggle in that respect.",
      "votes": null
    },
    {
      "id": "730163",
      "postDate": "01/27/2020 07:02:07",
      "content": "<p>Thanks for your suggestion!</p>",
      "rawMarkdown": "Thanks for your suggestion!",
      "votes": null
    },
    {
      "id": "731494",
      "postDate": "01/28/2020 17:41:33",
      "content": "<p><a href=\"/phunghieu\">@phunghieu</a> for the preprocessing you need more CPU and higher IOPS on the disk. And for training you need GPU. SageMaker is helpful because it comes preinstalled with different python virtual environments with most of the libraries needed for deep learning. </p>",
      "rawMarkdown": "phunghieu for the preprocessing you need more CPU and higher IOPS on the disk. And for training you need GPU. SageMaker is helpful because it comes preinstalled with different python virtual environments with most of the libraries needed for deep learning.",
      "votes": null
    },
    {
      "id": "731614",
      "postDate": "01/28/2020 20:21:34",
      "content": "<p>Thank <a href=\"/sciarrilli\">@sciarrilli</a>, I will try SageMaker first.</p>",
      "rawMarkdown": "Thank @sciarrilli, I will try SageMaker first.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 730154,
      "author_name": "dagnelies",
      "author_url": "",
      "post_date": "01/27/2020 06:49:55",
      "content": "<p>I guess it's more a question of preference rather than anything else. Honestly, the whole AWS/google ecosystem is a bit complicated so it might take a while to get used to. Without prior experience, I would advise SageMaker over plain EC2 instances. Why? Because you have everything pre-installed and notebooks at your fingertips ...which makes it similar to kaggle in that respect.</p>",
      "votes": null,
      "replies": [
        {
          "id": 730163,
          "author_name": "phunghieu",
          "author_url": "",
          "post_date": "01/27/2020 07:02:07",
          "content": "<p>Thanks for your suggestion!</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 731494,
      "author_name": "sciarrilli",
      "author_url": "",
      "post_date": "01/28/2020 17:41:33",
      "content": "<p><a href=\"/phunghieu\">@phunghieu</a> for the preprocessing you need more CPU and higher IOPS on the disk. And for training you need GPU. SageMaker is helpful because it comes preinstalled with different python virtual environments with most of the libraries needed for deep learning. </p>",
      "votes": null,
      "replies": [
        {
          "id": 731614,
          "author_name": "phunghieu",
          "author_url": "",
          "post_date": "01/28/2020 20:21:34",
          "content": "<p>Thank <a href=\"/sciarrilli\">@sciarrilli</a>, I will try SageMaker first.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "730078": "Since the whole training dataset of this competition is freaking huge, I've considered using cloud-based services for training models. There are so many options provided by Google, Amazon, etc., but I will try AWS first. Now, I have to make another decision, using EC2 or SageMaker. I know SageMaker provides us more interesting things, proper configurations, etc. with a higher price, yet do I really need these advantages to just training my models? Is EC2 the better choice for a beginner like me?\n\nThanks in advance!",
    "730154": "I guess it's more a question of preference rather than anything else. Honestly, the whole AWS/google ecosystem is a bit complicated so it might take a while to get used to. Without prior experience, I would advise SageMaker over plain EC2 instances. Why? Because you have everything pre-installed and notebooks at your fingertips ...which makes it similar to kaggle in that respect.",
    "730163": "Thanks for your suggestion!",
    "731494": "phunghieu for the preprocessing you need more CPU and higher IOPS on the disk. And for training you need GPU. SageMaker is helpful because it comes preinstalled with different python virtual environments with most of the libraries needed for deep learning.",
    "731614": "Thank @sciarrilli, I will try SageMaker first."
  },
  "source": "meta"
}