{
  "id": 460804,
  "title": "SMP with timm backbone",
  "url": "/competitions/blood-vessel-segmentation/discussion/460804",
  "author_name": "",
  "post_date": "2023-12-11T07:25:46.586078700Z",
  "votes": 3,
  "comment_count": 8,
  "views": 0,
  "content": "<p>How do i use a timm encoder in smp unet? I could really use some advice here</p>",
  "messages": [
    {
      "id": "2557082",
      "postDate": "12/11/2023 07:25:46",
      "content": "<p>How do i use a timm encoder in smp unet? I could really use some advice here</p>",
      "rawMarkdown": "How do i use a timm encoder in smp unet? I could really use some advice here",
      "votes": null
    },
    {
      "id": "2557091",
      "postDate": "12/11/2023 07:33:28",
      "content": "<p>As stated <a href=\"https://smp.readthedocs.io/en/latest/encoders_timm.html\" target=\"_blank\">here</a>, you need to add a <code>tu-</code> prefix to the encoder you're interested in. Example - <code>tu-seresnext26d_32x4d</code>.</p>",
      "rawMarkdown": "As stated [here](https://smp.readthedocs.io/en/latest/encoders_timm.html), you need to add a `tu-` prefix to the encoder you're interested in. Example - `tu-seresnext26d_32x4d`.",
      "votes": null
    },
    {
      "id": "2557121",
      "postDate": "12/11/2023 08:07:01",
      "content": "<p>I missed this🥲🥲🥲<br>\nThanks bro</p>",
      "rawMarkdown": "I missed this🥲🥲🥲\nThanks bro",
      "votes": null
    },
    {
      "id": "2559141",
      "postDate": "12/12/2023 16:14:44",
      "content": "<p>Real question , how can you submit a notebok with a unet from smp ? if i donwload the weight i need the library to get the unet architetcture to load the weight ( so i need internet )  and if i save the entire model , i cant load it without smp install ( still need internet ) . How can i resolve this ? ( iam new with pytorch ) </p>",
      "rawMarkdown": "Real question , how can you submit a notebok with a unet from smp ? if i donwload the weight i need the library to get the unet architetcture to load the weight ( so i need internet )  and if i save the entire model , i cant load it without smp install ( still need internet ) . How can i resolve this ? ( iam new with pytorch )",
      "votes": null
    },
    {
      "id": "2559153",
      "postDate": "12/12/2023 16:19:58",
      "content": "<p>You can use the same dataset as in this notebook and pip install from wheels as done in the import section:<br>\n<a href=\"https://www.kaggle.com/code/yoyobar/2-5d-cutting-model-baseline-training\" target=\"_blank\">https://www.kaggle.com/code/yoyobar/2-5d-cutting-model-baseline-training</a></p>",
      "rawMarkdown": "You can use the same dataset as in this notebook and pip install from wheels as done in the import section:\nhttps://www.kaggle.com/code/yoyobar/2-5d-cutting-model-baseline-training",
      "votes": null
    },
    {
      "id": "2559191",
      "postDate": "12/12/2023 16:39:26",
      "content": "<p>Got the solution or you need help?</p>",
      "rawMarkdown": "Got the solution or you need help?",
      "votes": null
    },
    {
      "id": "2560126",
      "postDate": "12/13/2023 10:03:55",
      "content": "<p>Nickel ! merci beaucoup </p>",
      "rawMarkdown": "Nickel ! merci beaucoup",
      "votes": null
    },
    {
      "id": "2560138",
      "postDate": "12/13/2023 10:24:19",
      "content": "<p>got the solution , probleme resolved. Have a nice day </p>",
      "rawMarkdown": "got the solution , probleme resolved. Have a nice day",
      "votes": null
    },
    {
      "id": "2675940",
      "postDate": "03/01/2024 07:41:02",
      "content": "<p>After I read the 1st solution, I found that he add the encoder and decoder in smp, instead of tu-, anyone know how to do that?</p>",
      "rawMarkdown": "After I read the 1st solution, I found that he add the encoder and decoder in smp, instead of tu-, anyone know how to do that?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2557091,
      "author_name": "phanisrikanth",
      "author_url": "",
      "post_date": "12/11/2023 07:33:28",
      "content": "<p>As stated <a href=\"https://smp.readthedocs.io/en/latest/encoders_timm.html\" target=\"_blank\">here</a>, you need to add a <code>tu-</code> prefix to the encoder you're interested in. Example - <code>tu-seresnext26d_32x4d</code>.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2557121,
          "author_name": "bhavesjain",
          "author_url": "",
          "post_date": "12/11/2023 08:07:01",
          "content": "<p>I missed this🥲🥲🥲<br>\nThanks bro</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 2559141,
      "author_name": "dylandehili",
      "author_url": "",
      "post_date": "12/12/2023 16:14:44",
      "content": "<p>Real question , how can you submit a notebok with a unet from smp ? if i donwload the weight i need the library to get the unet architetcture to load the weight ( so i need internet )  and if i save the entire model , i cant load it without smp install ( still need internet ) . How can i resolve this ? ( iam new with pytorch ) </p>",
      "votes": null,
      "replies": [
        {
          "id": 2559153,
          "author_name": "optimo",
          "author_url": "",
          "post_date": "12/12/2023 16:19:58",
          "content": "<p>You can use the same dataset as in this notebook and pip install from wheels as done in the import section:<br>\n<a href=\"https://www.kaggle.com/code/yoyobar/2-5d-cutting-model-baseline-training\" target=\"_blank\">https://www.kaggle.com/code/yoyobar/2-5d-cutting-model-baseline-training</a></p>",
          "votes": null,
          "replies": [
            {
              "id": 2560126,
              "author_name": "dylandehili",
              "author_url": "",
              "post_date": "12/13/2023 10:03:55",
              "content": "<p>Nickel ! merci beaucoup </p>",
              "votes": null,
              "replies": []
            }
          ]
        },
        {
          "id": 2559191,
          "author_name": "bhavesjain",
          "author_url": "",
          "post_date": "12/12/2023 16:39:26",
          "content": "<p>Got the solution or you need help?</p>",
          "votes": null,
          "replies": [
            {
              "id": 2560138,
              "author_name": "dylandehili",
              "author_url": "",
              "post_date": "12/13/2023 10:24:19",
              "content": "<p>got the solution , probleme resolved. Have a nice day </p>",
              "votes": null,
              "replies": []
            }
          ]
        }
      ]
    },
    {
      "id": 2675940,
      "author_name": "athrunzala",
      "author_url": "",
      "post_date": "03/01/2024 07:41:02",
      "content": "<p>After I read the 1st solution, I found that he add the encoder and decoder in smp, instead of tu-, anyone know how to do that?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2557082": "How do i use a timm encoder in smp unet? I could really use some advice here",
    "2557091": "As stated [here](https://smp.readthedocs.io/en/latest/encoders_timm.html), you need to add a `tu-` prefix to the encoder you're interested in. Example - `tu-seresnext26d_32x4d`.",
    "2557121": "I missed this🥲🥲🥲\nThanks bro",
    "2559141": "Real question , how can you submit a notebok with a unet from smp ? if i donwload the weight i need the library to get the unet architetcture to load the weight ( so i need internet )  and if i save the entire model , i cant load it without smp install ( still need internet ) . How can i resolve this ? ( iam new with pytorch )",
    "2559153": "You can use the same dataset as in this notebook and pip install from wheels as done in the import section:\nhttps://www.kaggle.com/code/yoyobar/2-5d-cutting-model-baseline-training",
    "2559191": "Got the solution or you need help?",
    "2560126": "Nickel ! merci beaucoup",
    "2560138": "got the solution , probleme resolved. Have a nice day",
    "2675940": "After I read the 1st solution, I found that he add the encoder and decoder in smp, instead of tu-, anyone know how to do that?"
  },
  "source": "meta"
}