{
  "id": 390515,
  "title": "does anyone try tensor-rt with tensorflow?",
  "url": "/competitions/rsna-breast-cancer-detection/discussion/390515",
  "author_name": "Simon Alerdic",
  "post_date": "2023-02-26T00:28:26.787000",
  "votes": 0,
  "comment_count": 3,
  "views": 0,
  "content": "<p>as per the title goes.</p>",
  "messages": [
    {
      "id": 2163200,
      "postDate": "2023-02-28T16:52:24.907Z",
      "content": "<p>I tried, but no luck.<br>\nI had to turn off XLA optimizations, when I used tensor-rt. (Maybe there is any other way to do though.)<br>\nI don't remember the accurate benchmark, but XLA optimized version is slightly faster than tensor-rt version.</p>\n<p>I published my <a href=\"https://www.kaggle.com/code/kena0ki/trt-optimized-rsna-convnextv2-inference/notebook\" target=\"_blank\">notebook</a>.</p>",
      "rawMarkdown": "I tried, but no luck.\nI had to turn off XLA optimizations, when I used tensor-rt. (Maybe there is any other way to do though.)\nI don't remember the accurate benchmark, but XLA optimized version is slightly faster than tensor-rt version.\n \nI published my [notebook](https://www.kaggle.com/code/kena0ki/trt-optimized-rsna-convnextv2-inference/notebook).\n"
    },
    {
      "id": 2159864,
      "postDate": "2023-02-26T05:21:28.597Z",
      "content": "<p>we had tried few days back and ran into some issues. You can refer to the following discussion:</p>\n<p><a href=\"https://www.kaggle.com/competitions/rsna-breast-cancer-detection/discussion/386739\" target=\"_blank\">https://www.kaggle.com/competitions/rsna-breast-cancer-detection/discussion/386739</a></p>",
      "rawMarkdown": "we had tried few days back and ran into some issues. You can refer to the following discussion:\n\nhttps://www.kaggle.com/competitions/rsna-breast-cancer-detection/discussion/386739",
      "replies": [
        {
          "id": 2160399,
          "postDate": "2023-02-26T16:31:13.990Z",
          "content": "<p>Thanks but I'm asking for keras / tensorflow based model. But in that discussion, all is about pytorch. Did I miss something?</p>",
          "rawMarkdown": "Thanks but I'm asking for keras / tensorflow based model. But in that discussion, all is about pytorch. Did I miss something?"
        }
      ]
    },
    {
      "id": 2159698,
      "postDate": "2023-02-26T00:28:26.787Z",
      "content": "<p>as per the title goes.</p>",
      "rawMarkdown": "as per the title goes."
    }
  ],
  "comments": [
    {
      "id": 2163200,
      "author_name": "kena0ki",
      "author_url": "",
      "post_date": "2023-02-28T16:52:24.907000",
      "content": "<p>I tried, but no luck.<br>\nI had to turn off XLA optimizations, when I used tensor-rt. (Maybe there is any other way to do though.)<br>\nI don't remember the accurate benchmark, but XLA optimized version is slightly faster than tensor-rt version.</p>\n<p>I published my <a href=\"https://www.kaggle.com/code/kena0ki/trt-optimized-rsna-convnextv2-inference/notebook\" target=\"_blank\">notebook</a>.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 2159864,
      "author_name": "sandy1112",
      "author_url": "",
      "post_date": "2023-02-26T05:21:28.597000",
      "content": "<p>we had tried few days back and ran into some issues. You can refer to the following discussion:</p>\n<p><a href=\"https://www.kaggle.com/competitions/rsna-breast-cancer-detection/discussion/386739\" target=\"_blank\">https://www.kaggle.com/competitions/rsna-breast-cancer-detection/discussion/386739</a></p>",
      "votes": 0,
      "replies": [
        {
          "id": 2160399,
          "author_name": "Simon Alerdic",
          "author_url": "",
          "post_date": "2023-02-26T16:31:13.990000",
          "content": "<p>Thanks but I'm asking for keras / tensorflow based model. But in that discussion, all is about pytorch. Did I miss something?</p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2163200": "I tried, but no luck.\nI had to turn off XLA optimizations, when I used tensor-rt. (Maybe there is any other way to do though.)\nI don't remember the accurate benchmark, but XLA optimized version is slightly faster than tensor-rt version.\n \nI published my [notebook](https://www.kaggle.com/code/kena0ki/trt-optimized-rsna-convnextv2-inference/notebook).\n",
    "2159864": "we had tried few days back and ran into some issues. You can refer to the following discussion:\n\nhttps://www.kaggle.com/competitions/rsna-breast-cancer-detection/discussion/386739",
    "2159698": "as per the title goes."
  }
}