{
  "id": 268689,
  "title": "Error when using lr_find() of fastai",
  "url": "/competitions/rsna-miccai-brain-tumor-radiogenomic-classification/discussion/268689",
  "author_name": "Prakash Pandey",
  "post_date": "2021-08-28T11:35:51.034000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hello everyone, <br>\nI am using fastai for this competition. Whenever I try to find a suitable learning rate using lr_find() method, the following error shows up :<br>\n<code>RuntimeError: Caught RuntimeError in DataLoader worker process 0. \n  Original Traceback (most recent call last):\n  File \"/opt/conda/lib/python3.7/site-packages/torch/utils/data/_utils/worker.py\", line 198, in _worker_loop\n    data = fetcher.fetch(index)\n  File \"/opt/conda/lib/python3.7/site-packages/torch/utils/data/_utils/fetch.py\", line 34, in fetch\n    data = next(self.dataset_iter)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 118, in create_batches\n    yield from map(self.do_batch, self.chunkify(res))\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 144, in do_batch\n    def do_batch(self, b): return self.retain(self.create_batch(self.before_batch(b)), b)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 143, in create_batch\n    def create_batch(self, b): return (fa_collate,fa_convert)[self.prebatched](b)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 50, in fa_collate\n    else type(t[0])([fa_collate(s) for s in zip(*t)]) if isinstance(b, Sequence)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 50, in &lt;listcomp&gt;\n    else type(t[0])([fa_collate(s) for s in zip(*t)]) if isinstance(b, Sequence)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 49, in fa_collate\n    return (default_collate(t) if isinstance(b, _collate_types)\n  File \"/opt/conda/lib/python3.7/site-packages/torch/utils/data/_utils/collate.py\", line 55, in default_collate\n    return torch.stack(batch, 0, out=out)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/torch_core.py\", line 329, in __torch_function__\n    res = super().__torch_function__(func, types, args=args, kwargs=kwargs)\n  File \"/opt/conda/lib/python3.7/site-packages/torch/tensor.py\", line 995, in __torch_function__\n    ret = func(*args, **kwargs)\nRuntimeError: result type Float can't be cast to the desired output type Int</code></p>\n<p>I am unable to understand where does the dataloader cast a Float result to an Int. <br>\nPlease help.</p>",
  "messages": [
    {
      "id": 1494071,
      "postDate": "2021-08-28T11:35:51.033Z",
      "content": "<p>Hello everyone, <br>\nI am using fastai for this competition. Whenever I try to find a suitable learning rate using lr_find() method, the following error shows up :<br>\n<code>RuntimeError: Caught RuntimeError in DataLoader worker process 0. \n  Original Traceback (most recent call last):\n  File \"/opt/conda/lib/python3.7/site-packages/torch/utils/data/_utils/worker.py\", line 198, in _worker_loop\n    data = fetcher.fetch(index)\n  File \"/opt/conda/lib/python3.7/site-packages/torch/utils/data/_utils/fetch.py\", line 34, in fetch\n    data = next(self.dataset_iter)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 118, in create_batches\n    yield from map(self.do_batch, self.chunkify(res))\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 144, in do_batch\n    def do_batch(self, b): return self.retain(self.create_batch(self.before_batch(b)), b)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 143, in create_batch\n    def create_batch(self, b): return (fa_collate,fa_convert)[self.prebatched](b)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 50, in fa_collate\n    else type(t[0])([fa_collate(s) for s in zip(*t)]) if isinstance(b, Sequence)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 50, in &lt;listcomp&gt;\n    else type(t[0])([fa_collate(s) for s in zip(*t)]) if isinstance(b, Sequence)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 49, in fa_collate\n    return (default_collate(t) if isinstance(b, _collate_types)\n  File \"/opt/conda/lib/python3.7/site-packages/torch/utils/data/_utils/collate.py\", line 55, in default_collate\n    return torch.stack(batch, 0, out=out)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/torch_core.py\", line 329, in __torch_function__\n    res = super().__torch_function__(func, types, args=args, kwargs=kwargs)\n  File \"/opt/conda/lib/python3.7/site-packages/torch/tensor.py\", line 995, in __torch_function__\n    ret = func(*args, **kwargs)\nRuntimeError: result type Float can't be cast to the desired output type Int</code></p>\n<p>I am unable to understand where does the dataloader cast a Float result to an Int. <br>\nPlease help.</p>",
      "rawMarkdown": "Hello everyone, \nI am using fastai for this competition. Whenever I try to find a suitable learning rate using lr_find() method, the following error shows up :\n`RuntimeError: Caught RuntimeError in DataLoader worker process 0. \n  Original Traceback (most recent call last):\n  File \"/opt/conda/lib/python3.7/site-packages/torch/utils/data/_utils/worker.py\", line 198, in _worker_loop\n    data = fetcher.fetch(index)\n  File \"/opt/conda/lib/python3.7/site-packages/torch/utils/data/_utils/fetch.py\", line 34, in fetch\n    data = next(self.dataset_iter)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 118, in create_batches\n    yield from map(self.do_batch, self.chunkify(res))\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 144, in do_batch\n    def do_batch(self, b): return self.retain(self.create_batch(self.before_batch(b)), b)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 143, in create_batch\n    def create_batch(self, b): return (fa_collate,fa_convert)[self.prebatched](b)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 50, in fa_collate\n    else type(t[0])([fa_collate(s) for s in zip(*t)]) if isinstance(b, Sequence)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 50, in <listcomp>\n    else type(t[0])([fa_collate(s) for s in zip(*t)]) if isinstance(b, Sequence)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 49, in fa_collate\n    return (default_collate(t) if isinstance(b, _collate_types)\n  File \"/opt/conda/lib/python3.7/site-packages/torch/utils/data/_utils/collate.py\", line 55, in default_collate\n    return torch.stack(batch, 0, out=out)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/torch_core.py\", line 329, in __torch_function__\n    res = super().__torch_function__(func, types, args=args, kwargs=kwargs)\n  File \"/opt/conda/lib/python3.7/site-packages/torch/tensor.py\", line 995, in __torch_function__\n    ret = func(*args, **kwargs)\nRuntimeError: result type Float can't be cast to the desired output type Int`\n\nI am unable to understand where does the dataloader cast a Float result to an Int. \nPlease help.\n"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1494071": "Hello everyone, \nI am using fastai for this competition. Whenever I try to find a suitable learning rate using lr_find() method, the following error shows up :\n`RuntimeError: Caught RuntimeError in DataLoader worker process 0. \n  Original Traceback (most recent call last):\n  File \"/opt/conda/lib/python3.7/site-packages/torch/utils/data/_utils/worker.py\", line 198, in _worker_loop\n    data = fetcher.fetch(index)\n  File \"/opt/conda/lib/python3.7/site-packages/torch/utils/data/_utils/fetch.py\", line 34, in fetch\n    data = next(self.dataset_iter)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 118, in create_batches\n    yield from map(self.do_batch, self.chunkify(res))\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 144, in do_batch\n    def do_batch(self, b): return self.retain(self.create_batch(self.before_batch(b)), b)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 143, in create_batch\n    def create_batch(self, b): return (fa_collate,fa_convert)[self.prebatched](b)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 50, in fa_collate\n    else type(t[0])([fa_collate(s) for s in zip(*t)]) if isinstance(b, Sequence)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 50, in <listcomp>\n    else type(t[0])([fa_collate(s) for s in zip(*t)]) if isinstance(b, Sequence)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/data/load.py\", line 49, in fa_collate\n    return (default_collate(t) if isinstance(b, _collate_types)\n  File \"/opt/conda/lib/python3.7/site-packages/torch/utils/data/_utils/collate.py\", line 55, in default_collate\n    return torch.stack(batch, 0, out=out)\n  File \"/opt/conda/lib/python3.7/site-packages/fastai/torch_core.py\", line 329, in __torch_function__\n    res = super().__torch_function__(func, types, args=args, kwargs=kwargs)\n  File \"/opt/conda/lib/python3.7/site-packages/torch/tensor.py\", line 995, in __torch_function__\n    ret = func(*args, **kwargs)\nRuntimeError: result type Float can't be cast to the desired output type Int`\n\nI am unable to understand where does the dataloader cast a Float result to an Int. \nPlease help.\n"
  }
}