{
  "id": 71311,
  "title": "FastAI. PyTorch.  NotImplementedError:",
  "url": "/competitions/quora-insincere-questions-classification/discussion/71311",
  "author_name": "",
  "post_date": "2018-11-12T14:15:39.156098200Z",
  "votes": 1,
  "comment_count": 5,
  "views": 0,
  "content": "<p>Anybody use FastAI for the competition? I have some problems then i try to train the model :\nLooks like fastai 0.7 inconsistent with torch 0.4</p>\n\n<p>--&gt; 176         X = self.embed._backend.Embedding.apply(words,\n    177              masked_embed_weight, padding_idx, self.embed.max_norm,\n    178              self.embed.norm_type, self.embed.scale_grad_by_freq, self.embed.sparse)</p>\n\n<p>/opt/conda/lib/python3.6/site-packages/torch/nn/backends/backend.py in <strong>getattr</strong>(self, name)\n      8         fn = self.function_classes.get(name)\n      9         if fn is None:\n---&gt; 10             raise NotImplementedError\n     11         return fn\n     12 </p>\n\n<p>NotImplementedError: </p>",
  "messages": [
    {
      "id": "419755",
      "postDate": "11/12/2018 14:15:39",
      "content": "<p>Anybody use FastAI for the competition? I have some problems then i try to train the model :\nLooks like fastai 0.7 inconsistent with torch 0.4</p>\n\n<p>--&gt; 176         X = self.embed._backend.Embedding.apply(words,\n    177              masked_embed_weight, padding_idx, self.embed.max_norm,\n    178              self.embed.norm_type, self.embed.scale_grad_by_freq, self.embed.sparse)</p>\n\n<p>/opt/conda/lib/python3.6/site-packages/torch/nn/backends/backend.py in <strong>getattr</strong>(self, name)\n      8         fn = self.function_classes.get(name)\n      9         if fn is None:\n---&gt; 10             raise NotImplementedError\n     11         return fn\n     12 </p>\n\n<p>NotImplementedError: </p>",
      "rawMarkdown": "Anybody use FastAI for the competition? I have some problems then i try to train the model :\nLooks like fastai 0.7 inconsistent with torch 0.4\n\n\n--&gt; 176         X = self.embed._backend.Embedding.apply(words,\n    177              masked_embed_weight, padding_idx, self.embed.max_norm,\n    178              self.embed.norm_type, self.embed.scale_grad_by_freq, self.embed.sparse)\n\n/opt/conda/lib/python3.6/site-packages/torch/nn/backends/backend.py in __getattr__(self, name)\n      8         fn = self.function_classes.get(name)\n      9         if fn is None:\n---&gt; 10             raise NotImplementedError\n     11         return fn\n     12 \n\nNotImplementedError:",
      "votes": null
    },
    {
      "id": "424272",
      "postDate": "11/19/2018 20:29:59",
      "content": "<p>facing the same issue. Basically downgrading to pytorch 0.3 is supposed to fix this issue. Though I'm not sure how to go about that.</p>",
      "rawMarkdown": "facing the same issue. Basically downgrading to pytorch 0.3 is supposed to fix this issue. Though I'm not sure how to go about that.",
      "votes": null
    },
    {
      "id": "424775",
      "postDate": "11/20/2018 16:36:28",
      "content": "<p>try the following in a cell. Worked for me.</p>\n\n<p>!pip install <a href=\"http://download.pytorch.org/whl/cu80/torch-0.3.1-cp36-cp36m-linux_x86_64.whl\">http://download.pytorch.org/whl/cu80/torch-0.3.1-cp36-cp36m-linux_x86_64.whl</a> \n!pip install torchtext==0.2.3</p>",
      "rawMarkdown": "try the following in a cell. Worked for me.\n\n!pip install http://download.pytorch.org/whl/cu80/torch-0.3.1-cp36-cp36m-linux_x86_64.whl \n!pip install torchtext==0.2.3",
      "votes": null
    },
    {
      "id": "425924",
      "postDate": "11/22/2018 10:13:36",
      "content": "<p>Thanks for advise, but it's kernel competition what not allow to set external packages.</p>",
      "rawMarkdown": "Thanks for advise, but it's kernel competition what not allow to set external packages.",
      "votes": null
    },
    {
      "id": "426582",
      "postDate": "11/23/2018 13:37:52",
      "content": "<p>Hey MikhailShumikhin Here is the initial version of ULMFit mention your comments <a href=\"https://www.kaggle.com/karthik7395/ulmfit-language-model-state-of-the-art\">https://www.kaggle.com/karthik7395/ulmfit-language-model-state-of-the-art</a></p>",
      "rawMarkdown": "Hey MikhailShumikhin Here is the initial version of ULMFit mention your comments https://www.kaggle.com/karthik7395/ulmfit-language-model-state-of-the-art",
      "votes": null
    },
    {
      "id": "428540",
      "postDate": "11/27/2018 12:52:05",
      "content": "<p>Thanks so much</p>",
      "rawMarkdown": "Thanks so much",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 424272,
      "author_name": "rohitgan",
      "author_url": "",
      "post_date": "11/19/2018 20:29:59",
      "content": "<p>facing the same issue. Basically downgrading to pytorch 0.3 is supposed to fix this issue. Though I'm not sure how to go about that.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 424775,
      "author_name": "rohitgan",
      "author_url": "",
      "post_date": "11/20/2018 16:36:28",
      "content": "<p>try the following in a cell. Worked for me.</p>\n\n<p>!pip install <a href=\"http://download.pytorch.org/whl/cu80/torch-0.3.1-cp36-cp36m-linux_x86_64.whl\">http://download.pytorch.org/whl/cu80/torch-0.3.1-cp36-cp36m-linux_x86_64.whl</a> \n!pip install torchtext==0.2.3</p>",
      "votes": null,
      "replies": [
        {
          "id": 425924,
          "author_name": "mikhailshumikhin",
          "author_url": "",
          "post_date": "11/22/2018 10:13:36",
          "content": "<p>Thanks for advise, but it's kernel competition what not allow to set external packages.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 426582,
      "author_name": "karthik7395",
      "author_url": "",
      "post_date": "11/23/2018 13:37:52",
      "content": "<p>Hey MikhailShumikhin Here is the initial version of ULMFit mention your comments <a href=\"https://www.kaggle.com/karthik7395/ulmfit-language-model-state-of-the-art\">https://www.kaggle.com/karthik7395/ulmfit-language-model-state-of-the-art</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 428540,
          "author_name": "mikhailshumikhin",
          "author_url": "",
          "post_date": "11/27/2018 12:52:05",
          "content": "<p>Thanks so much</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "419755": "Anybody use FastAI for the competition? I have some problems then i try to train the model :\nLooks like fastai 0.7 inconsistent with torch 0.4\n\n\n--&gt; 176         X = self.embed._backend.Embedding.apply(words,\n    177              masked_embed_weight, padding_idx, self.embed.max_norm,\n    178              self.embed.norm_type, self.embed.scale_grad_by_freq, self.embed.sparse)\n\n/opt/conda/lib/python3.6/site-packages/torch/nn/backends/backend.py in __getattr__(self, name)\n      8         fn = self.function_classes.get(name)\n      9         if fn is None:\n---&gt; 10             raise NotImplementedError\n     11         return fn\n     12 \n\nNotImplementedError:",
    "424272": "facing the same issue. Basically downgrading to pytorch 0.3 is supposed to fix this issue. Though I'm not sure how to go about that.",
    "424775": "try the following in a cell. Worked for me.\n\n!pip install http://download.pytorch.org/whl/cu80/torch-0.3.1-cp36-cp36m-linux_x86_64.whl \n!pip install torchtext==0.2.3",
    "425924": "Thanks for advise, but it's kernel competition what not allow to set external packages.",
    "426582": "Hey MikhailShumikhin Here is the initial version of ULMFit mention your comments https://www.kaggle.com/karthik7395/ulmfit-language-model-state-of-the-art",
    "428540": "Thanks so much"
  },
  "source": "meta"
}