{
  "id": 209151,
  "title": "Need help with tta implementation with fast.ai during inference ",
  "url": "/competitions/cassava-leaf-disease-classification/discussion/209151",
  "author_name": "",
  "post_date": "2021-01-06T14:06:51.720678800Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hi dear Kaggle community!<br>\nHere is <a href=\"https://www.kaggle.com/volcanoflash/inference-test-tta\" target=\"_blank\">my notebook</a> where I'm doing inference part of my pipeline. I'm stuck with fast.ai TTA implementation that gives me very unusual error: ‘NoneType’ object is not callable (details could be found in the notebook). </p>\n<p>Google doesn't helped me, fast.ai forum - the same. Maybe Kagglers are experienced enough to find my bug/mistake. </p>\n<p>Any thoughts on topic are very welcome!</p>",
  "messages": [
    {
      "id": "1141109",
      "postDate": "01/06/2021 14:06:51",
      "content": "<p>Hi dear Kaggle community!<br>\nHere is <a href=\"https://www.kaggle.com/volcanoflash/inference-test-tta\" target=\"_blank\">my notebook</a> where I'm doing inference part of my pipeline. I'm stuck with fast.ai TTA implementation that gives me very unusual error: ‘NoneType’ object is not callable (details could be found in the notebook). </p>\n<p>Google doesn't helped me, fast.ai forum - the same. Maybe Kagglers are experienced enough to find my bug/mistake. </p>\n<p>Any thoughts on topic are very welcome!</p>",
      "rawMarkdown": "Hi dear Kaggle community!\nHere is [my notebook](https://www.kaggle.com/volcanoflash/inference-test-tta) where I'm doing inference part of my pipeline. I'm stuck with fast.ai TTA implementation that gives me very unusual error: ‘NoneType’ object is not callable (details could be found in the notebook). \n\nGoogle doesn't helped me, fast.ai forum - the same. Maybe Kagglers are experienced enough to find my bug/mistake. \n\nAny thoughts on topic are very welcome!",
      "votes": null
    },
    {
      "id": "1142108",
      "postDate": "01/07/2021 06:45:44",
      "content": "<p>I suspect you have to define the transformations for TTA in the inference notebook.</p>\n<p>Did you see Zach Mueller's <a href=\"https://www.kaggle.com/muellerzr/recreating-abhishek-s-tez-with-fastai\" target=\"_blank\">inference notebook</a> with fastai and TTA? That's a nice example to follow.</p>",
      "rawMarkdown": "I suspect you have to define the transformations for TTA in the inference notebook.\n\nDid you see Zach Mueller's [inference notebook](https://www.kaggle.com/muellerzr/recreating-abhishek-s-tez-with-fastai) with fastai and TTA? That's a nice example to follow.",
      "votes": null
    },
    {
      "id": "1143355",
      "postDate": "01/07/2021 21:20:03",
      "content": "<p>I was able to run this code properly on Paperspace machine, so the problem is somehow connected to Kaggle itself or it’s environment. Getting mad with this bug…</p>",
      "rawMarkdown": "I was able to run this code properly on Paperspace machine, so the problem is somehow connected to Kaggle itself or it’s environment. Getting mad with this bug…",
      "votes": null
    },
    {
      "id": "1143427",
      "postDate": "01/07/2021 22:42:12",
      "content": "<p>Just to be sure, you have the same package versions esp. fastai and whatever you are using for augmentations? Plus really no difference in code?</p>",
      "rawMarkdown": "Just to be sure, you have the same package versions esp. fastai and whatever you are using for augmentations? Plus really no difference in code?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1142108,
      "author_name": "bjoernholzhauer",
      "author_url": "",
      "post_date": "01/07/2021 06:45:44",
      "content": "<p>I suspect you have to define the transformations for TTA in the inference notebook.</p>\n<p>Did you see Zach Mueller's <a href=\"https://www.kaggle.com/muellerzr/recreating-abhishek-s-tez-with-fastai\" target=\"_blank\">inference notebook</a> with fastai and TTA? That's a nice example to follow.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1143355,
      "author_name": "volcanoflash",
      "author_url": "",
      "post_date": "01/07/2021 21:20:03",
      "content": "<p>I was able to run this code properly on Paperspace machine, so the problem is somehow connected to Kaggle itself or it’s environment. Getting mad with this bug…</p>",
      "votes": null,
      "replies": [
        {
          "id": 1143427,
          "author_name": "bjoernholzhauer",
          "author_url": "",
          "post_date": "01/07/2021 22:42:12",
          "content": "<p>Just to be sure, you have the same package versions esp. fastai and whatever you are using for augmentations? Plus really no difference in code?</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1141109": "Hi dear Kaggle community!\nHere is [my notebook](https://www.kaggle.com/volcanoflash/inference-test-tta) where I'm doing inference part of my pipeline. I'm stuck with fast.ai TTA implementation that gives me very unusual error: ‘NoneType’ object is not callable (details could be found in the notebook). \n\nGoogle doesn't helped me, fast.ai forum - the same. Maybe Kagglers are experienced enough to find my bug/mistake. \n\nAny thoughts on topic are very welcome!",
    "1142108": "I suspect you have to define the transformations for TTA in the inference notebook.\n\nDid you see Zach Mueller's [inference notebook](https://www.kaggle.com/muellerzr/recreating-abhishek-s-tez-with-fastai) with fastai and TTA? That's a nice example to follow.",
    "1143355": "I was able to run this code properly on Paperspace machine, so the problem is somehow connected to Kaggle itself or it’s environment. Getting mad with this bug…",
    "1143427": "Just to be sure, you have the same package versions esp. fastai and whatever you are using for augmentations? Plus really no difference in code?"
  },
  "source": "meta"
}