{
  "id": 235213,
  "title": "PyTorch vs TensorFlow",
  "url": "/competitions/plant-pathology-2021-fgvc8/discussion/235213",
  "author_name": "",
  "post_date": "2021-04-28T11:13:05.137186300Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>While reading some solution notebooks i figure out that the most used tools is PyTorch and not TensorFlow. So i said maybe they some reason for that choice. Please tell me what is the difference? </p>",
  "messages": [
    {
      "id": "1286739",
      "postDate": "04/28/2021 11:13:05",
      "content": "<p>While reading some solution notebooks i figure out that the most used tools is PyTorch and not TensorFlow. So i said maybe they some reason for that choice. Please tell me what is the difference? </p>",
      "rawMarkdown": "While reading some solution notebooks i figure out that the most used tools is PyTorch and not TensorFlow. So i said maybe they some reason for that choice. Please tell me what is the difference?",
      "votes": null
    },
    {
      "id": "1286799",
      "postDate": "04/28/2021 12:17:54",
      "content": "<p>I´ve been thinking about it too but what I´ve seen (and read) is that Pytorch is quite easier to learn and as fast as tensorflow. Althought some people have claimed that the tensorflow community is larger I haven´t had any difficulties finding help so far. When mentioning Keras then there is that slower performance problem.</p>\n<p>Take a look at this: <a href=\"https://deepsense.ai/keras-or-pytorch/\" target=\"_blank\">https://deepsense.ai/keras-or-pytorch/</a></p>\n<p>But I´m no seasoned ML engineer so someone else might have a better explanation.</p>",
      "rawMarkdown": "I´ve been thinking about it too but what I´ve seen (and read) is that Pytorch is quite easier to learn and as fast as tensorflow. Althought some people have claimed that the tensorflow community is larger I haven´t had any difficulties finding help so far. When mentioning Keras then there is that slower performance problem.\n\nTake a look at this: https://deepsense.ai/keras-or-pytorch/\n\nBut I´m no seasoned ML engineer so someone else might have a better explanation.",
      "votes": null
    },
    {
      "id": "1286956",
      "postDate": "04/28/2021 15:02:30",
      "content": "<p>You can achieve a good result with either of them. I personally use Tensorflow here because it's more suitable for working with TPU (although you can use PyTorch XLA).</p>",
      "rawMarkdown": "You can achieve a good result with either of them. I personally use Tensorflow here because it's more suitable for working with TPU (although you can use PyTorch XLA).",
      "votes": null
    },
    {
      "id": "1300246",
      "postDate": "05/10/2021 11:00:50",
      "content": "<p>Once I tried pytorch, I cannot go back to tensorflow, since there are lot more open resources in pytorch.</p>",
      "rawMarkdown": "Once I tried pytorch, I cannot go back to tensorflow, since there are lot more open resources in pytorch.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1286799,
      "author_name": "dandierlean",
      "author_url": "",
      "post_date": "04/28/2021 12:17:54",
      "content": "<p>I´ve been thinking about it too but what I´ve seen (and read) is that Pytorch is quite easier to learn and as fast as tensorflow. Althought some people have claimed that the tensorflow community is larger I haven´t had any difficulties finding help so far. When mentioning Keras then there is that slower performance problem.</p>\n<p>Take a look at this: <a href=\"https://deepsense.ai/keras-or-pytorch/\" target=\"_blank\">https://deepsense.ai/keras-or-pytorch/</a></p>\n<p>But I´m no seasoned ML engineer so someone else might have a better explanation.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1286956,
      "author_name": "atamazian",
      "author_url": "",
      "post_date": "04/28/2021 15:02:30",
      "content": "<p>You can achieve a good result with either of them. I personally use Tensorflow here because it's more suitable for working with TPU (although you can use PyTorch XLA).</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1300246,
      "author_name": "ytepzhi",
      "author_url": "",
      "post_date": "05/10/2021 11:00:50",
      "content": "<p>Once I tried pytorch, I cannot go back to tensorflow, since there are lot more open resources in pytorch.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1286739": "While reading some solution notebooks i figure out that the most used tools is PyTorch and not TensorFlow. So i said maybe they some reason for that choice. Please tell me what is the difference?",
    "1286799": "I´ve been thinking about it too but what I´ve seen (and read) is that Pytorch is quite easier to learn and as fast as tensorflow. Althought some people have claimed that the tensorflow community is larger I haven´t had any difficulties finding help so far. When mentioning Keras then there is that slower performance problem.\n\nTake a look at this: https://deepsense.ai/keras-or-pytorch/\n\nBut I´m no seasoned ML engineer so someone else might have a better explanation.",
    "1286956": "You can achieve a good result with either of them. I personally use Tensorflow here because it's more suitable for working with TPU (although you can use PyTorch XLA).",
    "1300246": "Once I tried pytorch, I cannot go back to tensorflow, since there are lot more open resources in pytorch."
  },
  "source": "meta"
}