├── .devcontainer.json ├── .github └── workflows │ └── main.yml ├── .gitignore ├── 00_torch.ipynb ├── 01_fastai.dataloader.ipynb ├── 02_fastai.learner.ipynb ├── 03_fastai.transform.ipynb ├── 04_fastai.datasets.ipynb ├── CHANGELOG.md ├── CONTRIBUTING.md ├── LICENSE ├── MANIFEST.in ├── Makefile ├── README.md ├── docker-compose.yml ├── docs ├── .gitignore ├── _config.yml ├── _data │ ├── sidebars │ │ └── home_sidebar.yml │ └── topnav.yml ├── error.fastai.html ├── error.torch.html ├── fastai.data.html ├── fastai.dataloader.html ├── fastai.datasets.html ├── fastai.learner.html ├── fastai.transform.html ├── feed.xml ├── index.html ├── sidebar.json ├── sitemap.xml └── torch.html ├── fastdebug ├── __init__.py ├── _nbdev.py ├── fastai │ ├── __init__.py │ ├── dataloader.py │ ├── datasets.py │ ├── learner.py │ └── transform.py └── torch.py ├── index.ipynb ├── settings.ini └── setup.py /.devcontainer.json: -------------------------------------------------------------------------------- 1 | { 2 | "name": "nbdev_template-codespaces", 3 | "dockerComposeFile": "docker-compose.yml", 4 | "service": "watcher", 5 | "settings": {"terminal.integrated.shell.linux": "/bin/bash"}, 6 | "mounts": [ "source=/var/run/docker.sock,target=/var/run/docker.sock,type=bind" ], 7 | "forwardPorts": [4000, 8080], 8 | "appPort": [4000, 8080], 9 | "extensions": ["ms-python.python", 10 | "ms-azuretools.vscode-docker"], 11 | "runServices": ["notebook", "jekyll", "watcher"], 12 | "postStartCommand": "pip install -e ." 13 | } 14 | -------------------------------------------------------------------------------- /.github/workflows/main.yml: -------------------------------------------------------------------------------- 1 | name: CI 2 | on: [push, pull_request] 3 | jobs: 4 | build: 5 | runs-on: ubuntu-latest 6 | steps: 7 | - uses: actions/checkout@v1 8 | - uses: actions/setup-python@v1 9 | with: 10 | python-version: '3.6' 11 | architecture: 'x64' 12 | - name: Install the library 13 | run: | 14 | pip install nbdev jupyter 15 | pip install -e . 16 | - name: Read all notebooks 17 | run: | 18 | nbdev_read_nbs 19 | - name: Check if all notebooks are cleaned 20 | run: | 21 | echo "Check we are starting with clean git checkout" 22 | if [ -n "$(git status -uno -s)" ]; then echo "git status is not clean"; false; fi 23 | echo "Trying to strip out notebooks" 24 | nbdev_clean_nbs 25 | echo "Check that strip out was unnecessary" 26 | git status -s # display the status to see which nbs need cleaning up 27 | if [ -n "$(git status -uno -s)" ]; then echo -e "!!! Detected unstripped out notebooks\n!!!Remember to run nbdev_install_git_hooks"; false; fi 28 | - name: Check if there is no diff library/notebooks 29 | run: | 30 | if [ -n "$(nbdev_diff_nbs)" ]; then echo -e "!!! Detected difference between the notebooks and the library"; false; fi 31 | - name: Run tests 32 | run: | 33 | nbdev_test_nbs 34 | -------------------------------------------------------------------------------- /.gitignore: -------------------------------------------------------------------------------- 1 | *.bak 2 | .gitattributes 3 | .last_checked 4 | .gitconfig 5 | *.bak 6 | *.log 7 | *~ 8 | ~* 9 | _tmp* 10 | tmp* 11 | tags 12 | 13 | # Byte-compiled / optimized / DLL files 14 | __pycache__/ 15 | *.py[cod] 16 | *$py.class 17 | 18 | # C extensions 19 | *.so 20 | 21 | # Distribution / packaging 22 | .Python 23 | env/ 24 | build/ 25 | develop-eggs/ 26 | dist/ 27 | downloads/ 28 | eggs/ 29 | .eggs/ 30 | lib/ 31 | lib64/ 32 | parts/ 33 | sdist/ 34 | var/ 35 | wheels/ 36 | *.egg-info/ 37 | .installed.cfg 38 | *.egg 39 | 40 | # PyInstaller 41 | # Usually these files are written by a python script from a template 42 | # before PyInstaller builds the exe, so as to inject date/other infos into it. 43 | *.manifest 44 | *.spec 45 | 46 | # Installer logs 47 | pip-log.txt 48 | pip-delete-this-directory.txt 49 | 50 | # Unit test / coverage reports 51 | htmlcov/ 52 | .tox/ 53 | .coverage 54 | .coverage.* 55 | .cache 56 | nosetests.xml 57 | coverage.xml 58 | *.cover 59 | .hypothesis/ 60 | 61 | # Translations 62 | *.mo 63 | *.pot 64 | 65 | # Django stuff: 66 | *.log 67 | local_settings.py 68 | 69 | # Flask stuff: 70 | instance/ 71 | .webassets-cache 72 | 73 | # Scrapy stuff: 74 | .scrapy 75 | 76 | # Sphinx documentation 77 | docs/_build/ 78 | 79 | # PyBuilder 80 | target/ 81 | 82 | # Jupyter Notebook 83 | .ipynb_checkpoints 84 | 85 | # pyenv 86 | .python-version 87 | 88 | # celery beat schedule file 89 | celerybeat-schedule 90 | 91 | # SageMath parsed files 92 | *.sage.py 93 | 94 | # dotenv 95 | .env 96 | 97 | # virtualenv 98 | .venv 99 | venv/ 100 | ENV/ 101 | 102 | # Spyder project settings 103 | .spyderproject 104 | .spyproject 105 | 106 | # Rope project settings 107 | .ropeproject 108 | 109 | # mkdocs documentation 110 | /site 111 | 112 | # mypy 113 | .mypy_cache/ 114 | 115 | .vscode 116 | *.swp 117 | 118 | # osx generated files 119 | .DS_Store 120 | .DS_Store? 121 | .Trashes 122 | ehthumbs.db 123 | Thumbs.db 124 | .idea 125 | 126 | # pytest 127 | .pytest_cache 128 | 129 | # tools/trust-doc-nbs 130 | docs_src/.last_checked 131 | 132 | # symlinks to fastai 133 | docs_src/fastai 134 | tools/fastai 135 | 136 | # link checker 137 | checklink/cookies.txt 138 | 139 | # .gitconfig is now autogenerated 140 | .gitconfig 141 | 142 | -------------------------------------------------------------------------------- /00_torch.ipynb: -------------------------------------------------------------------------------- 1 | { 2 | "cells": [ 3 | { 4 | "cell_type": "code", 5 | "execution_count": null, 6 | "metadata": {}, 7 | "outputs": [], 8 | "source": [ 9 | "# default_exp torch" 10 | ] 11 | }, 12 | { 13 | "cell_type": "markdown", 14 | "metadata": {}, 15 | "source": [ 16 | "# Pytorch Errors\n", 17 | "\n", 18 | "> All the possible errors that fastdebug can support and verbosify involving Pytorch" 19 | ] 20 | }, 21 | { 22 | "cell_type": "code", 23 | "execution_count": null, 24 | "metadata": {}, 25 | "outputs": [], 26 | "source": [ 27 | "#hide\n", 28 | "from nbdev.showdoc import *\n", 29 | "from fastcore.test import test_eq" 30 | ] 31 | }, 32 | { 33 | "cell_type": "code", 34 | "execution_count": null, 35 | "metadata": {}, 36 | "outputs": [ 37 | { 38 | "name": "stderr", 39 | "output_type": "stream", 40 | "text": [ 41 | "/mnt/d/lib/python3.7/site-packages/torch/cuda/__init__.py:52: UserWarning: CUDA initialization: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx (Triggered internally at /pytorch/c10/cuda/CUDAFunctions.cpp:100.)\n", 42 | " return torch._C._cuda_getDeviceCount() > 0\n" 43 | ] 44 | } 45 | ], 46 | "source": [ 47 | "#export\n", 48 | "import torch\n", 49 | "import re\n", 50 | "from fastai.callback.hook import Hook\n", 51 | "from fastai.torch_core import to_detach\n", 52 | "from fastai.layers import flatten_model\n", 53 | "\n", 54 | "from fastcore.basics import store_attr" 55 | ] 56 | }, 57 | { 58 | "cell_type": "markdown", 59 | "metadata": {}, 60 | "source": [ 61 | "## Errors\n", 62 | "\n", 63 | "While some errrors are specifically designed for the [fastai](https://docs.fast.ai) library, the general idea still holds true in raw `Pytorch` as well. " 64 | ] 65 | }, 66 | { 67 | "cell_type": "code", 68 | "execution_count": null, 69 | "metadata": {}, 70 | "outputs": [], 71 | "source": [ 72 | "#export\n", 73 | "def device_error(e:Exception, a:str, b:str) -> Exception:\n", 74 | " \"\"\"\n", 75 | " Verbose error for if `a` and `b` are on different devices\n", 76 | " Should be used when checking if a model is on the same device, or two tensors\n", 77 | " \"\"\"\n", 78 | " inp, weight, _ = e.args[0].replace('( ', '').split(')')\n", 79 | " inp = inp.replace('Input type', f'{a} has type: \\t\\t')\n", 80 | " weight = weight.replace(' and weight type', f'{b} have type: \\t')\n", 81 | " err = f'Mismatch between weight types\\n\\n{inp})\\n{weight})\\n\\nBoth should be the same.'\n", 82 | " e.args = [err]\n", 83 | " raise e" 84 | ] 85 | }, 86 | { 87 | "cell_type": "markdown", 88 | "metadata": {}, 89 | "source": [ 90 | "The device error provides a much more readable error when `a` and `b` were on two different devices. An situation is below:\n", 91 | "```python\n", 92 | "inp = torch.rand().cuda()\n", 93 | "model = model.cpu()\n", 94 | "try:\n", 95 | " _ = model(inp)\n", 96 | "except Exception as e:\n", 97 | " device_error(e, 'Input type', 'Model weights')\n", 98 | "```\n", 99 | "And our new log:\n", 100 | "```bash\n", 101 | "---------------------------------------------------------------------------\n", 102 | "RuntimeError Traceback (most recent call last)\n", 103 | " in ()\n", 104 | " 2 model(x)\n", 105 | " 3 except Exception as e:\n", 106 | "----> 4 device_error(e, 'Input type', 'Model weights')\n", 107 | "\n", 108 | "10 frames\n", 109 | "/usr/local/lib/python3.7/dist-packages/torch/tensor.py in __torch_function__(cls, func, types, args, kwargs)\n", 110 | " 993 \n", 111 | " 994 with _C.DisableTorchFunction():\n", 112 | "--> 995 ret = func(*args, **kwargs)\n", 113 | " 996 return _convert(ret, cls)\n", 114 | " 997 \n", 115 | "\n", 116 | "RuntimeError: Mismatch between weight types\n", 117 | "\n", 118 | "Input type has type: \t\t (torch.cuda.FloatTensor)\n", 119 | "Model weights have type: \t (torch.FloatTensor)\n", 120 | "\n", 121 | "Both should be the same.\n", 122 | "```" 123 | ] 124 | }, 125 | { 126 | "cell_type": "code", 127 | "execution_count": null, 128 | "metadata": {}, 129 | "outputs": [], 130 | "source": [ 131 | "#export\n", 132 | "def hook_fn(m, i):\n", 133 | " \"Simple hook fn to return the layer\"\n", 134 | " return m" 135 | ] 136 | }, 137 | { 138 | "cell_type": "code", 139 | "execution_count": null, 140 | "metadata": {}, 141 | "outputs": [], 142 | "source": [ 143 | "#export\n", 144 | "class PreHook(Hook):\n", 145 | " \"Creates and registers a hook on `m` with `hook_func` as a forward pre_hook\"\n", 146 | " def __init__(self, m, hook_func, is_forward=True, detach=True, cpu=False, gather=False):\n", 147 | " store_attr('hook_func,detach,cpu,gather')\n", 148 | " f = m.register_forward_pre_hook if is_forward else m.register_backward_pre_hook\n", 149 | " self.hook = f(self.hook_fn)\n", 150 | " self.stored,self.removed = None, False\n", 151 | "\n", 152 | " def hook_fn(self, module, inp):\n", 153 | " \"Applies `hook_fn` to `module` and `inp`\"\n", 154 | " if self.detach:\n", 155 | " inp = to_detach(inp, cpu=self.cpu, gather=self.gather)\n", 156 | " self.stored = self.hook_func(module, inp)" 157 | ] 158 | }, 159 | { 160 | "cell_type": "code", 161 | "execution_count": null, 162 | "metadata": {}, 163 | "outputs": [], 164 | "source": [ 165 | "#export\n", 166 | "class ForwardHooks():\n", 167 | " \"Create several forward-hooks on the modules in `ms` with `hook_func`\"\n", 168 | " def __init__(self, ms, hook_func, is_forward=True, detach=True, cpu=False):\n", 169 | " self.hooks = []\n", 170 | " for i, m in enumerate(flatten_model(ms)):\n", 171 | " self.hooks.append(PreHook(m, hook_func, is_forward, detach, cpu))" 172 | ] 173 | }, 174 | { 175 | "cell_type": "code", 176 | "execution_count": null, 177 | "metadata": {}, 178 | "outputs": [], 179 | "source": [ 180 | "#export\n", 181 | "def hook_outputs(modules, detach=True, cpu=False, grad=False):\n", 182 | " \"Return `Hooks` that store activations of all `modules` in `self.stored`\"\n", 183 | " return ForwardHooks(modules, hook_fn, detach=detach, cpu=cpu, is_forward=not grad)" 184 | ] 185 | }, 186 | { 187 | "cell_type": "markdown", 188 | "metadata": {}, 189 | "source": [ 190 | "By using forward hooks, we can locate our problem layers when they arrive rather than trying to figure out which one it is through a list of confusing errors.\n", 191 | "\n", 192 | "For this tutorial and testing we'll purposefully write a broken model:" 193 | ] 194 | }, 195 | { 196 | "cell_type": "code", 197 | "execution_count": null, 198 | "metadata": {}, 199 | "outputs": [], 200 | "source": [ 201 | "from torch import nn\n", 202 | "m = nn.Sequential(\n", 203 | " nn.Conv2d(3,3,1),\n", 204 | " nn.ReLU(),\n", 205 | " nn.Linear(3,2)\n", 206 | ")" 207 | ] 208 | }, 209 | { 210 | "cell_type": "code", 211 | "execution_count": null, 212 | "metadata": {}, 213 | "outputs": [], 214 | "source": [ 215 | "#export\n", 216 | "def layer_error(e:Exception, model, *inp) -> Exception:\n", 217 | " \"\"\"\n", 218 | " Verbose error for when there is a size mismatch between some input and the model. \n", 219 | " `model` should be any torch model\n", 220 | " `inp` is the input that went to the model\n", 221 | " \"\"\"\n", 222 | " args = e.args[0].replace(\"Expected\", \"Model expected\")\n", 223 | " hooks = hook_outputs(model)\n", 224 | " try:\n", 225 | " _ = model(*inp)\n", 226 | " except:\n", 227 | " pass\n", 228 | " finally:\n", 229 | " layers,num = [], 0\n", 230 | " for i, layer in enumerate(hooks.hooks):\n", 231 | " if layer.stored is not None: \n", 232 | " layers.append(layer.stored)\n", 233 | " num += 1\n", 234 | " layer = layers[-1]\n", 235 | " [h.remove() for h in hooks.hooks]\n", 236 | " e.args = [f'Size mismatch between input tensors and what the model expects\\n{\"-\"*76}\\nLayer: {i}, {layer}\\nError: {args}']\n", 237 | " raise e" 238 | ] 239 | }, 240 | { 241 | "cell_type": "markdown", 242 | "metadata": {}, 243 | "source": [ 244 | "`layer_error` can be used anywhere that you want to check that the inputs are right for some model.\n", 245 | "\n", 246 | "Let's use our `m` model from earlier to show an example:" 247 | ] 248 | }, 249 | { 250 | "cell_type": "code", 251 | "execution_count": null, 252 | "metadata": {}, 253 | "outputs": [ 254 | { 255 | "ename": "RuntimeError", 256 | "evalue": "Size mismatch between input tensors and what the model expects\n----------------------------------------------------------------------------\nLayer: 2, Conv2d(3, 3, kernel_size=(1, 1), stride=(1, 1))\nError: Model expected 4-dimensional input for 4-dimensional weight [3, 3, 1, 1], but got 3-dimensional input of size [5, 2, 3] instead", 257 | "output_type": "error", 258 | "traceback": [ 259 | "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m", 260 | "\u001b[0;31mRuntimeError\u001b[0m Traceback (most recent call last)", 261 | "\u001b[0;32m\u001b[0m in \u001b[0;36m\u001b[0;34m()\u001b[0m\n\u001b[1;32m 4\u001b[0m \u001b[0mm\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minp\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 5\u001b[0m \u001b[0;32mexcept\u001b[0m \u001b[0mException\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 6\u001b[0;31m \u001b[0mlayer_error\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0me\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mm\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minp\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m", 262 | "\u001b[0;32m\u001b[0m in \u001b[0;36mlayer_error\u001b[0;34m(e, model, *inp)\u001b[0m\n\u001b[1;32m 21\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mh\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mremove\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mh\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mhooks\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mhooks\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 22\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0margs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34mf'Size mismatch between input tensors and what the model expects\\n{\"-\"*76}\\nLayer: {i}, {layer}\\nError: {args}'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 23\u001b[0;31m \u001b[0;32mraise\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m", 263 | "\u001b[0;32m\u001b[0m in \u001b[0;36m\u001b[0;34m()\u001b[0m\n\u001b[1;32m 2\u001b[0m \u001b[0minp\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrand\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m5\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;36m2\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m3\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 3\u001b[0m \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 4\u001b[0;31m \u001b[0mm\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minp\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 5\u001b[0m \u001b[0;32mexcept\u001b[0m \u001b[0mException\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 6\u001b[0m \u001b[0mlayer_error\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0me\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mm\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minp\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 264 | "\u001b[0;32m/mnt/d/lib/python3.7/site-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m_call_impl\u001b[0;34m(self, *input, **kwargs)\u001b[0m\n\u001b[1;32m 725\u001b[0m \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_slow_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 726\u001b[0m \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 727\u001b[0;31m \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 728\u001b[0m for hook in itertools.chain(\n\u001b[1;32m 729\u001b[0m \u001b[0m_global_forward_hooks\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 265 | "\u001b[0;32m/mnt/d/lib/python3.7/site-packages/torch/nn/modules/container.py\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, input)\u001b[0m\n\u001b[1;32m 115\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 116\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mmodule\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 117\u001b[0;31m \u001b[0minput\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodule\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 118\u001b[0m \u001b[0;32mreturn\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 119\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n", 266 | "\u001b[0;32m/mnt/d/lib/python3.7/site-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m_call_impl\u001b[0;34m(self, *input, **kwargs)\u001b[0m\n\u001b[1;32m 725\u001b[0m \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_slow_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 726\u001b[0m \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 727\u001b[0;31m \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 728\u001b[0m for hook in itertools.chain(\n\u001b[1;32m 729\u001b[0m \u001b[0m_global_forward_hooks\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 267 | "\u001b[0;32m/mnt/d/lib/python3.7/site-packages/torch/nn/modules/conv.py\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, input)\u001b[0m\n\u001b[1;32m 421\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 422\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mTensor\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m->\u001b[0m \u001b[0mTensor\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 423\u001b[0;31m \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_conv_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mweight\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 424\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 425\u001b[0m \u001b[0;32mclass\u001b[0m \u001b[0mConv3d\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0m_ConvNd\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 268 | "\u001b[0;32m/mnt/d/lib/python3.7/site-packages/torch/nn/modules/conv.py\u001b[0m in \u001b[0;36m_conv_forward\u001b[0;34m(self, input, weight)\u001b[0m\n\u001b[1;32m 418\u001b[0m _pair(0), self.dilation, self.groups)\n\u001b[1;32m 419\u001b[0m return F.conv2d(input, weight, self.bias, self.stride,\n\u001b[0;32m--> 420\u001b[0;31m self.padding, self.dilation, self.groups)\n\u001b[0m\u001b[1;32m 421\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 422\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mTensor\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m->\u001b[0m \u001b[0mTensor\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 269 | "\u001b[0;31mRuntimeError\u001b[0m: Size mismatch between input tensors and what the model expects\n----------------------------------------------------------------------------\nLayer: 2, Conv2d(3, 3, kernel_size=(1, 1), stride=(1, 1))\nError: Model expected 4-dimensional input for 4-dimensional weight [3, 3, 1, 1], but got 3-dimensional input of size [5, 2, 3] instead" 270 | ] 271 | } 272 | ], 273 | "source": [ 274 | "#failing\n", 275 | "inp = torch.rand(5,2, 3)\n", 276 | "try:\n", 277 | " m(inp)\n", 278 | "except Exception as e:\n", 279 | " layer_error(e, m, inp)" 280 | ] 281 | }, 282 | { 283 | "cell_type": "markdown", 284 | "metadata": {}, 285 | "source": [ 286 | "This will also work with multi-input and multi-output models:" 287 | ] 288 | }, 289 | { 290 | "cell_type": "code", 291 | "execution_count": null, 292 | "metadata": {}, 293 | "outputs": [], 294 | "source": [ 295 | "class DoubleInputModel(nn.Sequential):\n", 296 | " def __init__(self):\n", 297 | " super().__init__()\n", 298 | " self.layers = nn.Sequential(nn.Conv2d(3,3,1),\n", 299 | " nn.ReLU(),\n", 300 | " nn.Linear(3,2))\n", 301 | " def forward(self, a, b):\n", 302 | " return self.layers(a), self.layers(b)" 303 | ] 304 | }, 305 | { 306 | "cell_type": "code", 307 | "execution_count": null, 308 | "metadata": {}, 309 | "outputs": [], 310 | "source": [ 311 | "model = DoubleInputModel()" 312 | ] 313 | }, 314 | { 315 | "cell_type": "code", 316 | "execution_count": null, 317 | "metadata": {}, 318 | "outputs": [ 319 | { 320 | "ename": "RuntimeError", 321 | "evalue": "Size mismatch between input tensors and what the model expects\n----------------------------------------------------------------------------\nLayer: 2, Conv2d(3, 3, kernel_size=(1, 1), stride=(1, 1))\nError: Model expected 4-dimensional input for 4-dimensional weight [3, 3, 1, 1], but got 3-dimensional input of size [5, 2, 3] instead", 322 | "output_type": "error", 323 | "traceback": [ 324 | "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m", 325 | "\u001b[0;31mRuntimeError\u001b[0m Traceback (most recent call last)", 326 | "\u001b[0;32m\u001b[0m in \u001b[0;36m\u001b[0;34m()\u001b[0m\n\u001b[1;32m 4\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minp\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minp\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 5\u001b[0m \u001b[0;32mexcept\u001b[0m \u001b[0mException\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 6\u001b[0;31m \u001b[0mlayer_error\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0me\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minp\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minp\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m", 327 | "\u001b[0;32m\u001b[0m in \u001b[0;36mlayer_error\u001b[0;34m(e, model, *inp)\u001b[0m\n\u001b[1;32m 21\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mh\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mremove\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mh\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mhooks\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mhooks\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 22\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0margs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34mf'Size mismatch between input tensors and what the model expects\\n{\"-\"*76}\\nLayer: {i}, {layer}\\nError: {args}'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 23\u001b[0;31m \u001b[0;32mraise\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m", 328 | "\u001b[0;32m\u001b[0m in \u001b[0;36m\u001b[0;34m()\u001b[0m\n\u001b[1;32m 2\u001b[0m \u001b[0minp\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrand\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m5\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;36m2\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m3\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 3\u001b[0m \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 4\u001b[0;31m \u001b[0mmodel\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minp\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minp\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 5\u001b[0m \u001b[0;32mexcept\u001b[0m \u001b[0mException\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 6\u001b[0m \u001b[0mlayer_error\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0me\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minp\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minp\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 329 | "\u001b[0;32m/mnt/d/lib/python3.7/site-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m_call_impl\u001b[0;34m(self, *input, **kwargs)\u001b[0m\n\u001b[1;32m 725\u001b[0m \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_slow_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 726\u001b[0m \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 727\u001b[0;31m \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 728\u001b[0m for hook in itertools.chain(\n\u001b[1;32m 729\u001b[0m \u001b[0m_global_forward_hooks\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 330 | "\u001b[0;32m\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, a, b)\u001b[0m\n\u001b[1;32m 6\u001b[0m nn.Linear(3,2))\n\u001b[1;32m 7\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0ma\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 8\u001b[0;31m \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlayers\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0ma\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlayers\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m", 331 | "\u001b[0;32m/mnt/d/lib/python3.7/site-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m_call_impl\u001b[0;34m(self, *input, **kwargs)\u001b[0m\n\u001b[1;32m 725\u001b[0m \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_slow_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 726\u001b[0m \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 727\u001b[0;31m \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 728\u001b[0m for hook in itertools.chain(\n\u001b[1;32m 729\u001b[0m \u001b[0m_global_forward_hooks\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 332 | "\u001b[0;32m/mnt/d/lib/python3.7/site-packages/torch/nn/modules/container.py\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, input)\u001b[0m\n\u001b[1;32m 115\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 116\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mmodule\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 117\u001b[0;31m \u001b[0minput\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodule\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 118\u001b[0m \u001b[0;32mreturn\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 119\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n", 333 | "\u001b[0;32m/mnt/d/lib/python3.7/site-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m_call_impl\u001b[0;34m(self, *input, **kwargs)\u001b[0m\n\u001b[1;32m 725\u001b[0m \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_slow_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 726\u001b[0m \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 727\u001b[0;31m \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 728\u001b[0m for hook in itertools.chain(\n\u001b[1;32m 729\u001b[0m \u001b[0m_global_forward_hooks\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 334 | "\u001b[0;32m/mnt/d/lib/python3.7/site-packages/torch/nn/modules/conv.py\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, input)\u001b[0m\n\u001b[1;32m 421\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 422\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mTensor\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m->\u001b[0m \u001b[0mTensor\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 423\u001b[0;31m \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_conv_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mweight\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 424\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 425\u001b[0m \u001b[0;32mclass\u001b[0m \u001b[0mConv3d\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0m_ConvNd\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 335 | "\u001b[0;32m/mnt/d/lib/python3.7/site-packages/torch/nn/modules/conv.py\u001b[0m in \u001b[0;36m_conv_forward\u001b[0;34m(self, input, weight)\u001b[0m\n\u001b[1;32m 418\u001b[0m _pair(0), self.dilation, self.groups)\n\u001b[1;32m 419\u001b[0m return F.conv2d(input, weight, self.bias, self.stride,\n\u001b[0;32m--> 420\u001b[0;31m self.padding, self.dilation, self.groups)\n\u001b[0m\u001b[1;32m 421\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 422\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mTensor\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m->\u001b[0m \u001b[0mTensor\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 336 | "\u001b[0;31mRuntimeError\u001b[0m: Size mismatch between input tensors and what the model expects\n----------------------------------------------------------------------------\nLayer: 2, Conv2d(3, 3, kernel_size=(1, 1), stride=(1, 1))\nError: Model expected 4-dimensional input for 4-dimensional weight [3, 3, 1, 1], but got 3-dimensional input of size [5, 2, 3] instead" 337 | ] 338 | } 339 | ], 340 | "source": [ 341 | "#failing\n", 342 | "inp = torch.rand(5,2, 3)\n", 343 | "try:\n", 344 | " model(inp, inp)\n", 345 | "except Exception as e:\n", 346 | " layer_error(e, model, inp, inp)" 347 | ] 348 | }, 349 | { 350 | "cell_type": "markdown", 351 | "metadata": {}, 352 | "source": [ 353 | "Much more readable!" 354 | ] 355 | } 356 | ], 357 | "metadata": { 358 | "kernelspec": { 359 | "display_name": "Python 3", 360 | "language": "python", 361 | "name": "python3" 362 | } 363 | }, 364 | "nbformat": 4, 365 | "nbformat_minor": 2 366 | } 367 | -------------------------------------------------------------------------------- /01_fastai.dataloader.ipynb: -------------------------------------------------------------------------------- 1 | { 2 | "cells": [ 3 | { 4 | "cell_type": "code", 5 | "execution_count": null, 6 | "metadata": {}, 7 | "outputs": [], 8 | "source": [ 9 | "#default_exp fastai.dataloader" 10 | ] 11 | }, 12 | { 13 | "cell_type": "markdown", 14 | "metadata": {}, 15 | "source": [ 16 | "# DataLoader Errors\n", 17 | "> Errors and exceptions for any step of the `DataLoader` process" 18 | ] 19 | }, 20 | { 21 | "cell_type": "markdown", 22 | "metadata": {}, 23 | "source": [ 24 | "This includes `after_item`, `after_batch`, and collating. Anything in relation to the `Datasets` or anything before the `DataLoader` process can be found in `fastdebug.fastai.dataset`" 25 | ] 26 | }, 27 | { 28 | "cell_type": "code", 29 | "execution_count": null, 30 | "metadata": {}, 31 | "outputs": [], 32 | "source": [ 33 | "#export\n", 34 | "import inflect\n", 35 | "from fastcore.basics import patch\n", 36 | "from fastai.data.core import TfmdDL\n", 37 | "from fastai.data.load import DataLoader, fa_collate, fa_convert" 38 | ] 39 | }, 40 | { 41 | "cell_type": "code", 42 | "execution_count": null, 43 | "metadata": {}, 44 | "outputs": [], 45 | "source": [ 46 | "#export\n", 47 | "def collate_error(e:Exception, batch):\n", 48 | " \"\"\"\n", 49 | " Raises an explicit error when the batch could not collate, stating\n", 50 | " what items in the batch are different sizes and their types\n", 51 | " \"\"\"\n", 52 | " p = inflect.engine()\n", 53 | " err = f'Error when trying to collate the data into batches with fa_collate, '\n", 54 | " err += 'at least two tensors in the batch are not the same size.\\n\\n'\n", 55 | " # we need to iterate through the entire batch and find a mismatch\n", 56 | " length = len(batch[0])\n", 57 | " for idx in range(length): # for each type in the batch\n", 58 | " for i, item in enumerate(batch):\n", 59 | " if i == 0:\n", 60 | " shape_a = item[idx].shape\n", 61 | " type_a = item[idx].__class__.__name__\n", 62 | " elif item[idx].shape != shape_a:\n", 63 | " shape_b = item[idx].shape\n", 64 | " if shape_a != shape_b:\n", 65 | " err += f'Mismatch found within the {p.ordinal(idx)} axis of the batch and is of type {type_a}:\\n'\n", 66 | " err += f'The first item has shape: {shape_a}\\n'\n", 67 | " err += f'The {p.number_to_words(p.ordinal(i+1))} item has shape: {shape_b}\\n\\n'\n", 68 | " err += f'Please include a transform in `after_item` that ensures all data of type {type_a} is the same size'\n", 69 | " e.args = [err]\n", 70 | " raise e" 71 | ] 72 | }, 73 | { 74 | "cell_type": "code", 75 | "execution_count": null, 76 | "metadata": {}, 77 | "outputs": [], 78 | "source": [ 79 | "#export\n", 80 | "@patch\n", 81 | "def create_batch(self:DataLoader, b):\n", 82 | " \"Collate a list of items into a batch.\"\n", 83 | " func = (fa_collate,fa_convert)[self.prebatched]\n", 84 | " try:\n", 85 | " return func(b)\n", 86 | " except Exception as e:\n", 87 | " if not self.prebatched:\n", 88 | " collate_error(e, b) \n", 89 | " else: raise e" 90 | ] 91 | }, 92 | { 93 | "cell_type": "markdown", 94 | "metadata": {}, 95 | "source": [ 96 | "`collate_error` is `@patch`'d into `DataLoader`'s `create_batch` function through importing this module, so if there is any possible reason why the data cannot be collated into the batch, it is presented to the user.\n", 97 | "\n", 98 | "An example is below, where we forgot to include an item transform that resizes all our images to the same size:" 99 | ] 100 | }, 101 | { 102 | "cell_type": "code", 103 | "execution_count": null, 104 | "metadata": {}, 105 | "outputs": [ 106 | { 107 | "ename": "RuntimeError", 108 | "evalue": "ignored", 109 | "output_type": "error", 110 | "traceback": [ 111 | "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m", 112 | "\u001b[0;31mRuntimeError\u001b[0m Traceback (most recent call last)", 113 | "\u001b[0;32m\u001b[0m in \u001b[0;36m\u001b[0;34m()\u001b[0m\n\u001b[1;32m 6\u001b[0m label_func=lambda x: x[0].isupper())\n\u001b[1;32m 7\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 8\u001b[0;31m \u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mdls\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrain\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mone_batch\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m", 114 | "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/fastai/data/load.py\u001b[0m in \u001b[0;36mone_batch\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m 148\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mone_batch\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 149\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mn\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32mNone\u001b[0m \u001b[0;32mand\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m==\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;32mraise\u001b[0m \u001b[0mValueError\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34mf'This DataLoader does not contain any batches'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 150\u001b[0;31m \u001b[0;32mwith\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfake_l\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mno_multiproc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mres\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfirst\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 151\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mhasattr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'it'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mdelattr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'it'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 152\u001b[0m \u001b[0;32mreturn\u001b[0m \u001b[0mres\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 115 | "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/fastcore/basics.py\u001b[0m in \u001b[0;36mfirst\u001b[0;34m(x, f, negate, **kwargs)\u001b[0m\n\u001b[1;32m 545\u001b[0m \u001b[0mx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0miter\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 546\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mf\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfilter_ex\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mf\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mf\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnegate\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mnegate\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgen\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 547\u001b[0;31m \u001b[0;32mreturn\u001b[0m \u001b[0mnext\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 548\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 549\u001b[0m \u001b[0;31m# Cell\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 116 | "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/fastai/data/load.py\u001b[0m in \u001b[0;36m__iter__\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m 107\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbefore_iter\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 108\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__idxs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_idxs\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;31m# called in context of main process (not workers/subprocesses)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 109\u001b[0;31m \u001b[0;32mfor\u001b[0m \u001b[0mb\u001b[0m \u001b[0;32min\u001b[0m \u001b[0m_loaders\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfake_l\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnum_workers\u001b[0m\u001b[0;34m==\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfake_l\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 110\u001b[0m \u001b[0;31m# fix issue 2899. If the process start method isn't fork, the data will be copied to cuda in learner one_batch.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 111\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdevice\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32mNone\u001b[0m \u001b[0;32mand\u001b[0m \u001b[0mmultiprocessing\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_start_method\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlower\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;34m\"fork\"\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 117 | "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/torch/utils/data/dataloader.py\u001b[0m in \u001b[0;36m__next__\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m 433\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_sampler_iter\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 434\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_reset\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 435\u001b[0;31m \u001b[0mdata\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_next_data\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 436\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_num_yielded\u001b[0m \u001b[0;34m+=\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 437\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_dataset_kind\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0m_DatasetKind\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mIterable\u001b[0m \u001b[0;32mand\u001b[0m\u001b[0;31m \u001b[0m\u001b[0;31m\\\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 118 | "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/torch/utils/data/dataloader.py\u001b[0m in \u001b[0;36m_next_data\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m 473\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0m_next_data\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 474\u001b[0m \u001b[0mindex\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_next_index\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;31m# may raise StopIteration\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 475\u001b[0;31m \u001b[0mdata\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_dataset_fetcher\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfetch\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mindex\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;31m# may raise StopIteration\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 476\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_pin_memory\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 477\u001b[0m \u001b[0mdata\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_utils\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpin_memory\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpin_memory\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdata\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 119 | "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/torch/utils/data/_utils/fetch.py\u001b[0m in \u001b[0;36mfetch\u001b[0;34m(self, possibly_batched_index)\u001b[0m\n\u001b[1;32m 32\u001b[0m \u001b[0;32mraise\u001b[0m \u001b[0mStopIteration\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 33\u001b[0m \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 34\u001b[0;31m \u001b[0mdata\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnext\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdataset_iter\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 35\u001b[0m \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcollate_fn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdata\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 36\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n", 120 | "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/fastai/data/load.py\u001b[0m in \u001b[0;36mcreate_batches\u001b[0;34m(self, samps)\u001b[0m\n\u001b[1;32m 118\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdataset\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mit\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0miter\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdataset\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 119\u001b[0m \u001b[0mres\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfilter\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;32mlambda\u001b[0m \u001b[0mo\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mo\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmap\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdo_item\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msamps\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 120\u001b[0;31m \u001b[0;32myield\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mmap\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdo_batch\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mchunkify\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mres\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 121\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 122\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mnew\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdataset\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcls\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 121 | "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/fastai/data/load.py\u001b[0m in \u001b[0;36mdo_batch\u001b[0;34m(self, b)\u001b[0m\n\u001b[1;32m 144\u001b[0m \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;32mraise\u001b[0m \u001b[0mIndexError\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"Cannot index an iterable dataset numerically - must use `None`.\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 145\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mcreate_batch\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;32mreturn\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mfa_collate\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mfa_convert\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mprebatched\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 146\u001b[0;31m \u001b[0;32mdef\u001b[0m \u001b[0mdo_batch\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mretain\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcreate_batch\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbefore_batch\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 147\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mto\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdevice\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdevice\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mdevice\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 148\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mone_batch\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 122 | "\u001b[0;32m\u001b[0m in \u001b[0;36mcreate_batch\u001b[0;34m(self, b)\u001b[0m\n\u001b[1;32m 8\u001b[0m \u001b[0;32mexcept\u001b[0m \u001b[0mException\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 9\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mprebatched\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 10\u001b[0;31m \u001b[0mcollate_error\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0me\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 11\u001b[0m \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;32mraise\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 123 | "\u001b[0;32m\u001b[0m in \u001b[0;36mcollate_error\u001b[0;34m(e, batch)\u001b[0m\n\u001b[1;32m 23\u001b[0m \u001b[0merr\u001b[0m \u001b[0;34m+=\u001b[0m \u001b[0;34mf'Please include a transform in `after_item` that ensures all data of type {type_a} is the same size'\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 24\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0margs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0merr\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 25\u001b[0;31m \u001b[0;32mraise\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m", 124 | "\u001b[0;32m\u001b[0m in \u001b[0;36mcreate_batch\u001b[0;34m(self, b)\u001b[0m\n\u001b[1;32m 5\u001b[0m \u001b[0mfunc\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mfa_collate\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mfa_convert\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mprebatched\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 6\u001b[0m \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 7\u001b[0;31m \u001b[0;32mreturn\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 8\u001b[0m \u001b[0;32mexcept\u001b[0m \u001b[0mException\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 9\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mprebatched\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 125 | "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/fastai/data/load.py\u001b[0m in \u001b[0;36mfa_collate\u001b[0;34m(t)\u001b[0m\n\u001b[1;32m 48\u001b[0m \u001b[0mb\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mt\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 49\u001b[0m return (default_collate(t) if isinstance(b, _collate_types)\n\u001b[0;32m---> 50\u001b[0;31m \u001b[0;32melse\u001b[0m \u001b[0mtype\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mt\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mfa_collate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0ms\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0ms\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mzip\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0mt\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mb\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mSequence\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 51\u001b[0m else default_collate(t))\n\u001b[1;32m 52\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n", 126 | "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/fastai/data/load.py\u001b[0m in \u001b[0;36m\u001b[0;34m(.0)\u001b[0m\n\u001b[1;32m 48\u001b[0m \u001b[0mb\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mt\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 49\u001b[0m return (default_collate(t) if isinstance(b, _collate_types)\n\u001b[0;32m---> 50\u001b[0;31m \u001b[0;32melse\u001b[0m \u001b[0mtype\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mt\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mfa_collate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0ms\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0ms\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mzip\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0mt\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mb\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mSequence\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 51\u001b[0m else default_collate(t))\n\u001b[1;32m 52\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n", 127 | "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/fastai/data/load.py\u001b[0m in \u001b[0;36mfa_collate\u001b[0;34m(t)\u001b[0m\n\u001b[1;32m 47\u001b[0m \u001b[0;34m\"A replacement for PyTorch `default_collate` which maintains types and handles `Sequence`s\"\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 48\u001b[0m \u001b[0mb\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mt\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 49\u001b[0;31m return (default_collate(t) if isinstance(b, _collate_types)\n\u001b[0m\u001b[1;32m 50\u001b[0m \u001b[0;32melse\u001b[0m \u001b[0mtype\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mt\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mfa_collate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0ms\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0ms\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mzip\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0mt\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mb\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mSequence\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 51\u001b[0m else default_collate(t))\n", 128 | "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/torch/utils/data/_utils/collate.py\u001b[0m in \u001b[0;36mdefault_collate\u001b[0;34m(batch)\u001b[0m\n\u001b[1;32m 53\u001b[0m \u001b[0mstorage\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0melem\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mstorage\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_new_shared\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnumel\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 54\u001b[0m \u001b[0mout\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0melem\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnew\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mstorage\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 55\u001b[0;31m \u001b[0;32mreturn\u001b[0m \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mstack\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mbatch\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mout\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mout\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 56\u001b[0m \u001b[0;32melif\u001b[0m \u001b[0melem_type\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__module__\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;34m'numpy'\u001b[0m \u001b[0;32mand\u001b[0m \u001b[0melem_type\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__name__\u001b[0m \u001b[0;34m!=\u001b[0m \u001b[0;34m'str_'\u001b[0m\u001b[0;31m \u001b[0m\u001b[0;31m\\\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 57\u001b[0m \u001b[0;32mand\u001b[0m \u001b[0melem_type\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__name__\u001b[0m \u001b[0;34m!=\u001b[0m \u001b[0;34m'string_'\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 129 | "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/fastai/torch_core.py\u001b[0m in \u001b[0;36m__torch_function__\u001b[0;34m(self, func, types, args, kwargs)\u001b[0m\n\u001b[1;32m 327\u001b[0m \u001b[0mconvert\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 328\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0m_torch_handled\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_opt\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mconvert\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mtypes\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtype\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mTensor\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 329\u001b[0;31m \u001b[0mres\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0msuper\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__torch_function__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfunc\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtypes\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0margs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mkwargs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 330\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mconvert\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mres\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mconvert\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mres\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 331\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mres\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mTensorBase\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mres\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mset_meta\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mas_copy\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", 130 | "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/torch/tensor.py\u001b[0m in \u001b[0;36m__torch_function__\u001b[0;34m(cls, func, types, args, kwargs)\u001b[0m\n\u001b[1;32m 993\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 994\u001b[0m \u001b[0;32mwith\u001b[0m \u001b[0m_C\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mDisableTorchFunction\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 995\u001b[0;31m \u001b[0mret\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 996\u001b[0m \u001b[0;32mreturn\u001b[0m \u001b[0m_convert\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mret\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcls\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 997\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n", 131 | "\u001b[0;31mRuntimeError\u001b[0m: Error when trying to collate the data into batches with fa_collate, at least two tensors in the batch are not the same size.\n\nMismatch found within the 0th axis of the batch and is of type TensorImage:\nThe first item has shape: torch.Size([3, 500, 333])\nThe second item has shape: torch.Size([3, 333, 500])\n\nPlease include a transform in `after_item` that ensures all data of type TensorImage is the same size" 132 | ] 133 | } 134 | ], 135 | "source": [ 136 | "#failing\n", 137 | "from fastai.vision.all import *\n", 138 | "path = untar_data(URLs.PETS)/'images'\n", 139 | "dls = ImageDataLoaders.from_name_func(\n", 140 | " path, get_image_files(path), valid_pct=0.2,\n", 141 | " label_func=lambda x: x[0].isupper())\n", 142 | "\n", 143 | "x,y = dls.train.one_batch()" 144 | ] 145 | }, 146 | { 147 | "cell_type": "code", 148 | "execution_count": null, 149 | "metadata": {}, 150 | "outputs": [], 151 | "source": [ 152 | "#export\n", 153 | "@patch\n", 154 | "def new(self:TfmdDL, dataset=None, cls=None, **kwargs):\n", 155 | " res = super(TfmdDL, self).new(dataset, cls, do_setup=False, **kwargs)\n", 156 | " if not hasattr(self, '_n_inp') or not hasattr(self, '_types'):\n", 157 | " try:\n", 158 | " self._one_pass()\n", 159 | " res._n_inp,res._types = self._n_inp,self._types\n", 160 | " except Exception as e: \n", 161 | " print(\"Could not do one pass in your dataloader, there is something wrong in it\")\n", 162 | " raise e\n", 163 | " else: res._n_inp,res._types = self._n_inp,self._types\n", 164 | " return res" 165 | ] 166 | } 167 | ], 168 | "metadata": { 169 | "kernelspec": { 170 | "display_name": "Python 3", 171 | "language": "python", 172 | "name": "python3" 173 | } 174 | }, 175 | "nbformat": 4, 176 | "nbformat_minor": 1 177 | } 178 | -------------------------------------------------------------------------------- /02_fastai.learner.ipynb: -------------------------------------------------------------------------------- 1 | { 2 | "cells": [ 3 | { 4 | "cell_type": "code", 5 | "execution_count": null, 6 | "metadata": {}, 7 | "outputs": [], 8 | "source": [ 9 | "#default_exp fastai.learner" 10 | ] 11 | }, 12 | { 13 | "cell_type": "code", 14 | "execution_count": null, 15 | "metadata": {}, 16 | "outputs": [], 17 | "source": [ 18 | "#hide\n", 19 | "from nbdev.showdoc import *" 20 | ] 21 | }, 22 | { 23 | "cell_type": "markdown", 24 | "metadata": {}, 25 | "source": [ 26 | "# Learner Errors\n", 27 | "> In-place fastai specific errors to ease debugging" 28 | ] 29 | }, 30 | { 31 | "cell_type": "code", 32 | "execution_count": null, 33 | "metadata": {}, 34 | "outputs": [], 35 | "source": [ 36 | "#export\n", 37 | "from fastdebug.torch import layer_error, device_error\n", 38 | "\n", 39 | "from fastai.data.all import *\n", 40 | "from fastai.optimizer import *\n", 41 | "from fastai.learner import *\n", 42 | "from fastai.callback.core import event\n", 43 | "from fastai.callback.training import ShortEpochCallback\n", 44 | "from fastai.torch_core import default_device\n", 45 | "\n", 46 | "\n", 47 | "from fastcore.basics import patch\n", 48 | "from fastcore.meta import delegates" 49 | ] 50 | }, 51 | { 52 | "cell_type": "markdown", 53 | "metadata": {}, 54 | "source": [ 55 | "This notebook contains a series of various errors that can be used when running with `fastai`. It should be noted here that there is no other imports or magic you need to do to use this section of the library other then: `from fastdebug import *`. It will automatically load in what's needed.\n", 56 | "\n", 57 | "As a style choice, we are choosing to do the `.*` notation as this loads in not only all of our errors, but also replaces sections of `fastai`'s code to inject some error handling (as we'll see later)" 58 | ] 59 | }, 60 | { 61 | "cell_type": "markdown", 62 | "metadata": {}, 63 | "source": [ 64 | "## Error Types" 65 | ] 66 | }, 67 | { 68 | "cell_type": "code", 69 | "execution_count": null, 70 | "metadata": {}, 71 | "outputs": [], 72 | "source": [ 73 | "#export\n", 74 | "def loss_func_error(e:Exception, learn) -> Exception:\n", 75 | " \"\"\"\n", 76 | " Error that should be run when there is an issue when working with the loss function\n", 77 | " \n", 78 | " Raises with a message stating the shapes of the inputs and targs, and the error\n", 79 | " \"\"\"\n", 80 | " err = f'There was an issue with calculating the loss with `{getattr(learn.loss_func, \"__name__\", learn.loss_func)}`'\n", 81 | " err += f'\\n\\nPrediction shape(s): {[p.shape for p in listify(learn.pred)]}'\n", 82 | " err += f'\\nLabel Shape(s): {[y.shape for y in learn.yb]}'\n", 83 | " err += f'\\nError: {e.args[0]}'\n", 84 | " e.args = [err]\n", 85 | " raise e" 86 | ] 87 | }, 88 | { 89 | "cell_type": "code", 90 | "execution_count": null, 91 | "metadata": {}, 92 | "outputs": [], 93 | "source": [ 94 | "#export\n", 95 | "def callback_error(e:Exception, cb:str, event_name:str) -> Exception:\n", 96 | " \"\"\"\n", 97 | " Raises an error from when a Callback event failed, showing what event, the name of the Callback and the trace\n", 98 | " \"\"\"\n", 99 | " e.args = [f\"Exception raised in the {cb} Callback during {event_name}:\\n\\n{e.args[0]}\"]\n", 100 | " raise e" 101 | ] 102 | }, 103 | { 104 | "cell_type": "code", 105 | "execution_count": null, 106 | "metadata": {}, 107 | "outputs": [], 108 | "source": [ 109 | "#export\n", 110 | "def catch_pred_errors(e:Exception, model) -> Exception:\n", 111 | " \"Catches any errors relating to prediction that are either related to the device or model layers. Else raise `e`\"\n", 112 | " if \"Input type\" in e.args[0]: device_error(e, 'Input', 'Model weights')\n", 113 | " elif \"Expected\" in e.args[0]: layer_error(e, model)\n", 114 | " else: raise e # anything else " 115 | ] 116 | }, 117 | { 118 | "cell_type": "code", 119 | "execution_count": null, 120 | "metadata": {}, 121 | "outputs": [], 122 | "source": [ 123 | "#export\n", 124 | "def catch_loss_errors(e:Exception, learn):\n", 125 | " \"Catches any errors that occur with the loss function and its calculation\"\n", 126 | " if \"Input type\" in e.args[0]: device_error(e, 'Model prediction', 'Truths')\n", 127 | " else: loss_func_error(e, learn)" 128 | ] 129 | }, 130 | { 131 | "cell_type": "markdown", 132 | "metadata": {}, 133 | "source": [ 134 | "## Modifications and Enhancements to the fastai Source Code and `Learner`:" 135 | ] 136 | }, 137 | { 138 | "cell_type": "code", 139 | "execution_count": null, 140 | "metadata": {}, 141 | "outputs": [], 142 | "source": [ 143 | "#export\n", 144 | "@patch\n", 145 | "def sanity_check(self:Learner, show_table=False):\n", 146 | " \"Performs a short epoch and uses all the callbacks in `self.cbs` on the CPU to ensure nothing is broken\"\n", 147 | " device = getattr(self.dls, 'device', default_device())\n", 148 | " if hasattr(self.dls, 'device'):\n", 149 | " self.dls.device = 'cpu'\n", 150 | " else:\n", 151 | " # Using raw torch\n", 152 | " self.model.to('cpu')\n", 153 | " self.save('tmp')\n", 154 | " cbs = [ShortEpochCallback(short_valid=False)]\n", 155 | " if show_table:\n", 156 | " with self.no_bar(), self.no_logging():\n", 157 | " self.fit(1, cbs=cbs)\n", 158 | " else:\n", 159 | " self.fit(1, cbs=cbs)\n", 160 | " if hasattr(self.dls, 'device'):\n", 161 | " self.dls.device = device\n", 162 | " else:\n", 163 | " self.model.to(device)\n", 164 | " self.load('tmp')" 165 | ] 166 | }, 167 | { 168 | "cell_type": "code", 169 | "execution_count": null, 170 | "metadata": {}, 171 | "outputs": [], 172 | "source": [ 173 | "#export\n", 174 | "@patch\n", 175 | "@delegates(Learner.sanity_check)\n", 176 | "def __init__(self:Learner, dls, model, loss_func=None, opt_func=Adam, lr=defaults.lr, splitter=trainable_params, cbs=None,\n", 177 | " metrics=None, path=None, model_dir='models', wd=None, wd_bn_bias=False, train_bn=True,\n", 178 | " moms=(0.95,0.85,0.95), sanity_check=False, **kwargs):\n", 179 | " \"Group together a `model`, some `dls` and a `loss_func` to handle training, potentially run a sanity check\"\n", 180 | " path = Path(path) if path is not None else getattr(dls, 'path', Path('.'))\n", 181 | " if loss_func is None:\n", 182 | " loss_func = getattr(dls.train_ds, 'loss_func', None)\n", 183 | " assert loss_func is not None, \"Could not infer loss function from the data, please pass a loss function.\"\n", 184 | " self.dls,self.model = dls,model\n", 185 | " store_attr(but='dls,model,cbs')\n", 186 | " self.training,self.create_mbar,self.logger,self.opt,self.cbs = False,True,print,None,L()\n", 187 | " self.add_cbs(L(defaults.callbacks)+L(cbs))\n", 188 | " self(\"after_create\")\n", 189 | " if sanity_check: self.sanity_check(**kwargs)" 190 | ] 191 | }, 192 | { 193 | "cell_type": "code", 194 | "execution_count": null, 195 | "metadata": {}, 196 | "outputs": [ 197 | { 198 | "data": { 199 | "text/markdown": [ 200 | "

Learner.__init__[source]

\n", 201 | "\n", 202 | "> Learner.__init__(**`dls`**, **`model`**, **`loss_func`**=*`None`*, **`opt_func`**=*`Adam`*, **`lr`**=*`0.001`*, **`splitter`**=*`trainable_params`*, **`cbs`**=*`None`*, **`metrics`**=*`None`*, **`path`**=*`None`*, **`model_dir`**=*`'models'`*, **`wd`**=*`None`*, **`wd_bn_bias`**=*`False`*, **`train_bn`**=*`True`*, **`moms`**=*`(0.95, 0.85, 0.95)`*, **`sanity_check`**=*`False`*, **`show_table`**=*`False`*)\n", 203 | "\n", 204 | "Group together a `model`, some `dls` and a `loss_func` to handle training, potentially run a sanity check" 205 | ], 206 | "text/plain": [ 207 | "" 208 | ] 209 | }, 210 | "metadata": {}, 211 | "output_type": "display_data" 212 | } 213 | ], 214 | "source": [ 215 | "show_doc(Learner.__init__)" 216 | ] 217 | }, 218 | { 219 | "cell_type": "code", 220 | "execution_count": null, 221 | "metadata": {}, 222 | "outputs": [ 223 | { 224 | "data": { 225 | "text/markdown": [ 226 | "

Learner.sanity_check[source]

\n", 227 | "\n", 228 | "> Learner.sanity_check(**`show_table`**=*`False`*)\n", 229 | "\n", 230 | "Performs a short epoch and uses all the callbacks in `self.cbs` on the CPU to ensure nothing is broken" 231 | ], 232 | "text/plain": [ 233 | "" 234 | ] 235 | }, 236 | "metadata": {}, 237 | "output_type": "display_data" 238 | } 239 | ], 240 | "source": [ 241 | "show_doc(Learner.sanity_check)" 242 | ] 243 | }, 244 | { 245 | "cell_type": "markdown", 246 | "metadata": {}, 247 | "source": [ 248 | "With `sanity_check`, you can make sure that you've set everything up properly and you won't get any issues before pushing to the GPU. This allows you to quickly ensure that you won't get any `CUDA` device-assist errors, and that the whole training regiment will go well. " 249 | ] 250 | }, 251 | { 252 | "cell_type": "code", 253 | "execution_count": null, 254 | "metadata": {}, 255 | "outputs": [], 256 | "source": [ 257 | "#export\n", 258 | "@patch\n", 259 | "def _do_one_batch(self:Learner):\n", 260 | " try:\n", 261 | " self.pred = self.model(*self.xb)\n", 262 | " except RuntimeError as e:\n", 263 | " catch_pred_errors(e, self.model)\n", 264 | " self('after_pred')\n", 265 | " if len(self.yb):\n", 266 | " try:\n", 267 | " self.loss_grad = self.loss_func(self.pred, *self.yb)\n", 268 | " except Exception as e:\n", 269 | " catch_loss_errors(e, self)\n", 270 | " self.loss = self.loss_grad.clone()\n", 271 | " self('after_loss')\n", 272 | " if not self.training or not len(self.yb): return\n", 273 | " self('before_backward')\n", 274 | " self.loss_grad.backward()\n", 275 | " self._with_events(self.opt.step, 'step', CancelStepException)\n", 276 | " self.opt.zero_grad()" 277 | ] 278 | }, 279 | { 280 | "cell_type": "code", 281 | "execution_count": null, 282 | "metadata": {}, 283 | "outputs": [], 284 | "source": [ 285 | "#export\n", 286 | "@patch\n", 287 | "def _call_one(self:Learner, event_name):\n", 288 | " if not hasattr(event, event_name): raise Exception(f'missing {event_name}')\n", 289 | " for cb in self.cbs.sorted('order'):\n", 290 | " try:\n", 291 | " cb(event_name)\n", 292 | " except Exception as e:\n", 293 | " callback_error(e, cb.__repr__(), event_name)" 294 | ] 295 | }, 296 | { 297 | "cell_type": "code", 298 | "execution_count": null, 299 | "metadata": {}, 300 | "outputs": [], 301 | "source": [ 302 | "#export\n", 303 | "def module_error(e:AttributeError) -> AttributeError:\n", 304 | " \"\"\"\n", 305 | " Raises an error when trying to load in a previous `Learner` and custom functions were not available in the namespace\n", 306 | " \"\"\"\n", 307 | " args = e.args[0]\n", 308 | " err = 'Custom classes or functions exported with your `Learner` are not available in the namespace currently.\\n'\n", 309 | " err += 'Please re-declare them before calling `load_learner`:\\n\\n'\n", 310 | " err += args\n", 311 | " e.args = [err]\n", 312 | " raise e" 313 | ] 314 | }, 315 | { 316 | "cell_type": "code", 317 | "execution_count": null, 318 | "metadata": {}, 319 | "outputs": [], 320 | "source": [ 321 | "#export\n", 322 | "def load_learner(fname, cpu=True, pickle_module=pickle):\n", 323 | " \"Load a `Learner` object in `fname`, optionally putting it on the `cpu`\"\n", 324 | " distrib_barrier()\n", 325 | " try: res = torch.load(fname, map_location='cpu' if cpu else None, pickle_module=pickle_module)\n", 326 | " except AttributeError as e: module_error(e)\n", 327 | " if hasattr(res, 'to_fp32'): res = res.to_fp32()\n", 328 | " if cpu: res.dls.cpu()\n", 329 | " return res" 330 | ] 331 | }, 332 | { 333 | "cell_type": "markdown", 334 | "metadata": {}, 335 | "source": [ 336 | "We have a custom `load_learner` function here that can check if everything exported is available when bringing the model in, if not then it'll raise an explicit error" 337 | ] 338 | }, 339 | { 340 | "cell_type": "code", 341 | "execution_count": null, 342 | "metadata": {}, 343 | "outputs": [], 344 | "source": [] 345 | } 346 | ], 347 | "metadata": { 348 | "kernelspec": { 349 | "display_name": "Python 3", 350 | "language": "python", 351 | "name": "python3" 352 | } 353 | }, 354 | "nbformat": 4, 355 | "nbformat_minor": 2 356 | } 357 | -------------------------------------------------------------------------------- /CHANGELOG.md: -------------------------------------------------------------------------------- 1 | # Release Notes 2 | 3 | ## 0.0.5 4 | ### New Features: 5 | * DataLoader and Dataset errors, essentially FC 6 | 7 | ## 0.0.4 8 | ### New Features: 9 | * `layer_error` now supports multi-input models 10 | 11 | ## 0.0.3 12 | ### New Features: 13 | * Use Hooks to grab the problem layer when using `layer_error` rather than trying to guess. This way we can grab the exact one -------------------------------------------------------------------------------- /CONTRIBUTING.md: -------------------------------------------------------------------------------- 1 | # How to contribute 2 | 3 | ## How to get started 4 | 5 | Before anything else, please install the git hooks that run automatic scripts during each commit and merge to strip the notebooks of superfluous metadata (and avoid merge conflicts). After cloning the repository, run the following command inside it: 6 | ``` 7 | nbdev_install_git_hooks 8 | ``` 9 | 10 | ## Did you find a bug? 11 | 12 | * Ensure the bug was not already reported by searching on GitHub under Issues. 13 | * If you're unable to find an open issue addressing the problem, open a new one. Be sure to include a title and clear description, as much relevant information as possible, and a code sample or an executable test case demonstrating the expected behavior that is not occurring. 14 | * Be sure to add the complete error messages. 15 | 16 | #### Did you write a patch that fixes a bug? 17 | 18 | * Open a new GitHub pull request with the patch. 19 | * Ensure that your PR includes a test that fails without your patch, and pass with it. 20 | * Ensure the PR description clearly describes the problem and solution. Include the relevant issue number if applicable. 21 | 22 | ## PR submission guidelines 23 | 24 | * Keep each PR focused. While it's more convenient, do not combine several unrelated fixes together. Create as many branches as needing to keep each PR focused. 25 | * Do not mix style changes/fixes with "functional" changes. It's very difficult to review such PRs and it most likely get rejected. 26 | * Do not add/remove vertical whitespace. Preserve the original style of the file you edit as much as you can. 27 | * Do not turn an already submitted PR into your development playground. If after you submitted PR, you discovered that more work is needed - close the PR, do the required work and then submit a new PR. Otherwise each of your commits requires attention from maintainers of the project. 28 | * If, however, you submitted a PR and received a request for changes, you should proceed with commits inside that PR, so that the maintainer can see the incremental fixes and won't need to review the whole PR again. In the exception case where you realize it'll take many many commits to complete the requests, then it's probably best to close the PR, do the work and then submit it again. Use common sense where you'd choose one way over another. 29 | 30 | ## Do you want to contribute to the documentation? 31 | 32 | * Docs are automatically created from the notebooks in the nbs folder. 33 | 34 | -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | Apache License 2 | Version 2.0, January 2004 3 | http://www.apache.org/licenses/ 4 | 5 | TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION 6 | 7 | 1. Definitions. 8 | 9 | "License" shall mean the terms and conditions for use, reproduction, 10 | and distribution as defined by Sections 1 through 9 of this document. 11 | 12 | "Licensor" shall mean the copyright owner or entity authorized by 13 | the copyright owner that is granting the License. 14 | 15 | "Legal Entity" shall mean the union of the acting entity and all 16 | other entities that control, are controlled by, or are under common 17 | control with that entity. For the purposes of this definition, 18 | "control" means (i) the power, direct or indirect, to cause the 19 | direction or management of such entity, whether by contract or 20 | otherwise, or (ii) ownership of fifty percent (50%) or more of the 21 | outstanding shares, or (iii) beneficial ownership of such entity. 22 | 23 | "You" (or "Your") shall mean an individual or Legal Entity 24 | exercising permissions granted by this License. 25 | 26 | "Source" form shall mean the preferred form for making modifications, 27 | including but not limited to software source code, documentation 28 | source, and configuration files. 29 | 30 | "Object" form shall mean any form resulting from mechanical 31 | transformation or translation of a Source form, including but 32 | not limited to compiled object code, generated documentation, 33 | and conversions to other media types. 34 | 35 | "Work" shall mean the work of authorship, whether in Source or 36 | Object form, made available under the License, as indicated by a 37 | copyright notice that is included in or attached to the work 38 | (an example is provided in the Appendix below). 39 | 40 | "Derivative Works" shall mean any work, whether in Source or Object 41 | form, that is based on (or derived from) the Work and for which the 42 | editorial revisions, annotations, elaborations, or other modifications 43 | represent, as a whole, an original work of authorship. For the purposes 44 | of this License, Derivative Works shall not include works that remain 45 | separable from, or merely link (or bind by name) to the interfaces of, 46 | the Work and Derivative Works thereof. 47 | 48 | "Contribution" shall mean any work of authorship, including 49 | the original version of the Work and any modifications or additions 50 | to that Work or Derivative Works thereof, that is intentionally 51 | submitted to Licensor for inclusion in the Work by the copyright owner 52 | or by an individual or Legal Entity authorized to submit on behalf of 53 | the copyright owner. For the purposes of this definition, "submitted" 54 | means any form of electronic, verbal, or written communication sent 55 | to the Licensor or its representatives, including but not limited to 56 | communication on electronic mailing lists, source code control systems, 57 | and issue tracking systems that are managed by, or on behalf of, the 58 | Licensor for the purpose of discussing and improving the Work, but 59 | excluding communication that is conspicuously marked or otherwise 60 | designated in writing by the copyright owner as "Not a Contribution." 61 | 62 | "Contributor" shall mean Licensor and any individual or Legal Entity 63 | on behalf of whom a Contribution has been received by Licensor and 64 | subsequently incorporated within the Work. 65 | 66 | 2. Grant of Copyright License. Subject to the terms and conditions of 67 | this License, each Contributor hereby grants to You a perpetual, 68 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 69 | copyright license to reproduce, prepare Derivative Works of, 70 | publicly display, publicly perform, sublicense, and distribute the 71 | Work and such Derivative Works in Source or Object form. 72 | 73 | 3. Grant of Patent License. Subject to the terms and conditions of 74 | this License, each Contributor hereby grants to You a perpetual, 75 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 76 | (except as stated in this section) patent license to make, have made, 77 | use, offer to sell, sell, import, and otherwise transfer the Work, 78 | where such license applies only to those patent claims licensable 79 | by such Contributor that are necessarily infringed by their 80 | Contribution(s) alone or by combination of their Contribution(s) 81 | with the Work to which such Contribution(s) was submitted. If You 82 | institute patent litigation against any entity (including a 83 | cross-claim or counterclaim in a lawsuit) alleging that the Work 84 | or a Contribution incorporated within the Work constitutes direct 85 | or contributory patent infringement, then any patent licenses 86 | granted to You under this License for that Work shall terminate 87 | as of the date such litigation is filed. 88 | 89 | 4. Redistribution. You may reproduce and distribute copies of the 90 | Work or Derivative Works thereof in any medium, with or without 91 | modifications, and in Source or Object form, provided that You 92 | meet the following conditions: 93 | 94 | (a) You must give any other recipients of the Work or 95 | Derivative Works a copy of this License; and 96 | 97 | (b) You must cause any modified files to carry prominent notices 98 | stating that You changed the files; and 99 | 100 | (c) You must retain, in the Source form of any Derivative Works 101 | that You distribute, all copyright, patent, trademark, and 102 | attribution notices from the Source form of the Work, 103 | excluding those notices that do not pertain to any part of 104 | the Derivative Works; and 105 | 106 | (d) If the Work includes a "NOTICE" text file as part of its 107 | distribution, then any Derivative Works that You distribute must 108 | include a readable copy of the attribution notices contained 109 | within such NOTICE file, excluding those notices that do not 110 | pertain to any part of the Derivative Works, in at least one 111 | of the following places: within a NOTICE text file distributed 112 | as part of the Derivative Works; within the Source form or 113 | documentation, if provided along with the Derivative Works; or, 114 | within a display generated by the Derivative Works, if and 115 | wherever such third-party notices normally appear. The contents 116 | of the NOTICE file are for informational purposes only and 117 | do not modify the License. You may add Your own attribution 118 | notices within Derivative Works that You distribute, alongside 119 | or as an addendum to the NOTICE text from the Work, provided 120 | that such additional attribution notices cannot be construed 121 | as modifying the License. 122 | 123 | You may add Your own copyright statement to Your modifications and 124 | may provide additional or different license terms and conditions 125 | for use, reproduction, or distribution of Your modifications, or 126 | for any such Derivative Works as a whole, provided Your use, 127 | reproduction, and distribution of the Work otherwise complies with 128 | the conditions stated in this License. 129 | 130 | 5. Submission of Contributions. Unless You explicitly state otherwise, 131 | any Contribution intentionally submitted for inclusion in the Work 132 | by You to the Licensor shall be under the terms and conditions of 133 | this License, without any additional terms or conditions. 134 | Notwithstanding the above, nothing herein shall supersede or modify 135 | the terms of any separate license agreement you may have executed 136 | with Licensor regarding such Contributions. 137 | 138 | 6. Trademarks. This License does not grant permission to use the trade 139 | names, trademarks, service marks, or product names of the Licensor, 140 | except as required for reasonable and customary use in describing the 141 | origin of the Work and reproducing the content of the NOTICE file. 142 | 143 | 7. Disclaimer of Warranty. Unless required by applicable law or 144 | agreed to in writing, Licensor provides the Work (and each 145 | Contributor provides its Contributions) on an "AS IS" BASIS, 146 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or 147 | implied, including, without limitation, any warranties or conditions 148 | of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A 149 | PARTICULAR PURPOSE. You are solely responsible for determining the 150 | appropriateness of using or redistributing the Work and assume any 151 | risks associated with Your exercise of permissions under this License. 152 | 153 | 8. Limitation of Liability. In no event and under no legal theory, 154 | whether in tort (including negligence), contract, or otherwise, 155 | unless required by applicable law (such as deliberate and grossly 156 | negligent acts) or agreed to in writing, shall any Contributor be 157 | liable to You for damages, including any direct, indirect, special, 158 | incidental, or consequential damages of any character arising as a 159 | result of this License or out of the use or inability to use the 160 | Work (including but not limited to damages for loss of goodwill, 161 | work stoppage, computer failure or malfunction, or any and all 162 | other commercial damages or losses), even if such Contributor 163 | has been advised of the possibility of such damages. 164 | 165 | 9. Accepting Warranty or Additional Liability. While redistributing 166 | the Work or Derivative Works thereof, You may choose to offer, 167 | and charge a fee for, acceptance of support, warranty, indemnity, 168 | or other liability obligations and/or rights consistent with this 169 | License. However, in accepting such obligations, You may act only 170 | on Your own behalf and on Your sole responsibility, not on behalf 171 | of any other Contributor, and only if You agree to indemnify, 172 | defend, and hold each Contributor harmless for any liability 173 | incurred by, or claims asserted against, such Contributor by reason 174 | of your accepting any such warranty or additional liability. 175 | 176 | END OF TERMS AND CONDITIONS 177 | 178 | APPENDIX: How to apply the Apache License to your work. 179 | 180 | To apply the Apache License to your work, attach the following 181 | boilerplate notice, with the fields enclosed by brackets "[]" 182 | replaced with your own identifying information. (Don't include 183 | the brackets!) The text should be enclosed in the appropriate 184 | comment syntax for the file format. We also recommend that a 185 | file or class name and description of purpose be included on the 186 | same "printed page" as the copyright notice for easier 187 | identification within third-party archives. 188 | 189 | Copyright [yyyy] [name of copyright owner] 190 | 191 | Licensed under the Apache License, Version 2.0 (the "License"); 192 | you may not use this file except in compliance with the License. 193 | You may obtain a copy of the License at 194 | 195 | http://www.apache.org/licenses/LICENSE-2.0 196 | 197 | Unless required by applicable law or agreed to in writing, software 198 | distributed under the License is distributed on an "AS IS" BASIS, 199 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 200 | See the License for the specific language governing permissions and 201 | limitations under the License. 202 | -------------------------------------------------------------------------------- /MANIFEST.in: -------------------------------------------------------------------------------- 1 | include settings.ini 2 | include LICENSE 3 | include CONTRIBUTING.md 4 | include README.md 5 | recursive-exclude * __pycache__ 6 | -------------------------------------------------------------------------------- /Makefile: -------------------------------------------------------------------------------- 1 | .ONESHELL: 2 | SHELL := /bin/bash 3 | SRC = $(wildcard ./*.ipynb) 4 | 5 | all: fastdebug docs 6 | 7 | fastdebug: $(SRC) 8 | nbdev_build_lib 9 | touch fastdebug 10 | 11 | sync: 12 | nbdev_update_lib 13 | 14 | docs_serve: docs 15 | cd docs && bundle exec jekyll serve 16 | 17 | docs: $(SRC) 18 | nbdev_build_docs 19 | touch docs 20 | 21 | test: 22 | nbdev_test_nbs 23 | 24 | release: pypi conda_release 25 | nbdev_bump_version 26 | 27 | conda_release: 28 | fastrelease_conda_package 29 | 30 | pypi: dist 31 | twine upload --repository pypi dist/* 32 | 33 | dist: clean 34 | python setup.py sdist bdist_wheel 35 | 36 | clean: 37 | rm -rf dist -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | 2 | # fastdebug 3 | > A helpful library for improving torch and fastai errors 4 | 5 | 6 | ## Install 7 | 8 | `pip install fastdebug` 9 | 10 | ## How to use 11 | 12 | `fastdebug` is designed around improving the quality of life when dealing with Pytorch and fastai errors, while also including some new sanity checks (fastai only) 13 | 14 | ### Pytorch 15 | 16 | Pytorch now has: 17 | * `device_error` 18 | * `layer_error` 19 | 20 | Both can be imported with: 21 | ```python 22 | from fastdebug.error.torch import device_error, layer_error 23 | ``` 24 | 25 | `device_error` prints out a much more readable error for when two tensors aren't on the same device: 26 | 27 | ```python 28 | inp = torch.rand().cuda() 29 | model = model.cpu() 30 | try: 31 | _ = model(inp) 32 | except Exception as e: 33 | device_error(e, 'Input type', 'Model weights') 34 | ``` 35 | And our new log: 36 | ```bash 37 | --------------------------------------------------------------------------- 38 | RuntimeError Traceback (most recent call last) 39 | in () 40 | 2 model(x) 41 | 3 except Exception as e: 42 | ----> 4 device_error(e, 'Input type', 'Model weights') 43 | 44 | 10 frames 45 | /usr/local/lib/python3.7/dist-packages/torch/tensor.py in __torch_function__(cls, func, types, args, kwargs) 46 | 993 47 | 994 with _C.DisableTorchFunction(): 48 | --> 995 ret = func(*args, **kwargs) 49 | 996 return _convert(ret, cls) 50 | 997 51 | 52 | RuntimeError: Mismatch between weight types 53 | 54 | Input type has type: (torch.cuda.FloatTensor) 55 | Model weights have type: (torch.FloatTensor) 56 | 57 | Both should be the same. 58 | ``` 59 | 60 | And with `layer_error`, if there is a shape mismatch it will attempt to find the right layer it was at: 61 | ```python 62 | inp = torch.rand(5,2, 3) 63 | try: 64 | m(inp) 65 | except Exception as e: 66 | layer_error(e, m) 67 | ``` 68 | 69 | ```python 70 | --------------------------------------------------------------------------- 71 | RuntimeError Traceback (most recent call last) 72 | in () 73 | 3 m(inp) 74 | 4 except Exception as e: 75 | ----> 5 layer_error(e, m) 76 | 77 | in layer_error(e, model) 78 | 8 i, layer = get_layer_by_shape(model, shape) 79 | 9 e.args = [f'Size mismatch between input tensors and what the model expects\n\n{args}\n\tat layer {i}: {layer}'] 80 | ---> 10 raise e 81 | 82 | in () 83 | 1 inp = torch.rand(5,2, 3) 84 | 2 try: 85 | ----> 3 m(inp) 86 | 4 except Exception as e: 87 | 5 layer_error(e, m) 88 | 89 | /mnt/d/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs) 90 | 725 result = self._slow_forward(*input, **kwargs) 91 | 726 else: 92 | --> 727 result = self.forward(*input, **kwargs) 93 | 728 for hook in itertools.chain( 94 | 729 _global_forward_hooks.values(), 95 | 96 | /mnt/d/lib/python3.7/site-packages/torch/nn/modules/container.py in forward(self, input) 97 | 115 def forward(self, input): 98 | 116 for module in self: 99 | --> 117 input = module(input) 100 | 118 return input 101 | 119 102 | 103 | /mnt/d/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs) 104 | 725 result = self._slow_forward(*input, **kwargs) 105 | 726 else: 106 | --> 727 result = self.forward(*input, **kwargs) 107 | 728 for hook in itertools.chain( 108 | 729 _global_forward_hooks.values(), 109 | 110 | /mnt/d/lib/python3.7/site-packages/torch/nn/modules/conv.py in forward(self, input) 111 | 421 112 | 422 def forward(self, input: Tensor) -> Tensor: 113 | --> 423 return self._conv_forward(input, self.weight) 114 | 424 115 | 425 class Conv3d(_ConvNd): 116 | 117 | /mnt/d/lib/python3.7/site-packages/torch/nn/modules/conv.py in _conv_forward(self, input, weight) 118 | 418 _pair(0), self.dilation, self.groups) 119 | 419 return F.conv2d(input, weight, self.bias, self.stride, 120 | --> 420 self.padding, self.dilation, self.groups) 121 | 421 122 | 422 def forward(self, input: Tensor) -> Tensor: 123 | 124 | RuntimeError: Size mismatch between input tensors and what the model expects 125 | 126 | Model expected 4-dimensional input for 4-dimensional weight [3, 3, 1, 1], but got 3-dimensional input of size [5, 2, 3] instead 127 | at layer 1: Conv2d(3, 3, kernel_size=(1, 1), stride=(1, 1)) 128 | ``` 129 | 130 | ### fastai 131 | 132 | Along with the additions above (and are used during `fit`), fastai now has a `Learner.sanity_check` function, which allows you to quickly perform a basic check to ensure that your call to `fit` won't raise any exceptions. They are performed on the CPU for a partial epoch to make sure that `CUDA` device-assist errors can be preemptively found. 133 | 134 | To use it simply do: 135 | ```python 136 | from fastdebug.fastai import * 137 | from fastai.vision.all import * 138 | 139 | learn = Learner(...) 140 | learn.sanity_check() 141 | ``` 142 | 143 | This is also now an argument in `Learner`, set to `False` by default, so that after making your `Learner` a quick check is ensured. 144 | 145 | ```python 146 | learn = Learner(..., sanity_check=True) 147 | ``` 148 | -------------------------------------------------------------------------------- /docker-compose.yml: -------------------------------------------------------------------------------- 1 | version: "3" 2 | services: 3 | fastai: &fastai 4 | restart: unless-stopped 5 | working_dir: /data 6 | image: fastai/codespaces 7 | logging: 8 | driver: json-file 9 | options: 10 | max-size: 50m 11 | stdin_open: true 12 | tty: true 13 | volumes: 14 | - .:/data/ 15 | 16 | notebook: 17 | <<: *fastai 18 | command: bash -c "pip install -e . && jupyter notebook --allow-root --no-browser --ip=0.0.0.0 --port=8080 --NotebookApp.token='' --NotebookApp.password=''" 19 | ports: 20 | - "8080:8080" 21 | 22 | watcher: 23 | <<: *fastai 24 | command: watchmedo shell-command --command nbdev_build_docs --pattern *.ipynb --recursive --drop 25 | network_mode: host # for GitHub Codespaces https://github.com/features/codespaces/ 26 | 27 | jekyll: 28 | <<: *fastai 29 | ports: 30 | - "4000:4000" 31 | command: > 32 | bash -c "pip install . 33 | && nbdev_build_docs && cd docs 34 | && bundle i 35 | && chmod -R u+rwx . && bundle exec jekyll serve --host 0.0.0.0" 36 | -------------------------------------------------------------------------------- /docs/.gitignore: -------------------------------------------------------------------------------- 1 | _site/ 2 | -------------------------------------------------------------------------------- /docs/_config.yml: -------------------------------------------------------------------------------- 1 | repository: muellerzr/fastdebug 2 | output: web 3 | topnav_title: fastdebug 4 | site_title: fastdebug 5 | company_name: Zachary Mueller 6 | description: A library that improves the debugging messages for Pytorch and fastai 7 | # Set to false to disable KaTeX math 8 | use_math: true 9 | # Add Google analytics id if you have one and want to use it here 10 | google_analytics: 11 | # See http://nbdev.fast.ai/search for help with adding Search 12 | google_search: 13 | 14 | host: 127.0.0.1 15 | # the preview server used. Leave as is. 16 | port: 4000 17 | # the port where the preview is rendered. 18 | 19 | exclude: 20 | - .idea/ 21 | - .gitignore 22 | - vendor 23 | 24 | exclude: [vendor] 25 | 26 | highlighter: rouge 27 | markdown: kramdown 28 | kramdown: 29 | input: GFM 30 | auto_ids: true 31 | hard_wrap: false 32 | syntax_highlighter: rouge 33 | 34 | collections: 35 | tooltips: 36 | output: false 37 | 38 | defaults: 39 | - 40 | scope: 41 | path: "" 42 | type: "pages" 43 | values: 44 | layout: "page" 45 | comments: true 46 | search: true 47 | sidebar: home_sidebar 48 | topnav: topnav 49 | - 50 | scope: 51 | path: "" 52 | type: "tooltips" 53 | values: 54 | layout: "page" 55 | comments: true 56 | search: true 57 | tooltip: true 58 | 59 | sidebars: 60 | - home_sidebar 61 | 62 | plugins: 63 | - jekyll-remote-theme 64 | 65 | remote_theme: fastai/nbdev-jekyll-theme 66 | baseurl: /fastdebug/ -------------------------------------------------------------------------------- /docs/_data/sidebars/home_sidebar.yml: -------------------------------------------------------------------------------- 1 | 2 | ################################################# 3 | ### THIS FILE WAS AUTOGENERATED! DO NOT EDIT! ### 4 | ################################################# 5 | # Instead edit ../../sidebar.json 6 | entries: 7 | - folders: 8 | - folderitems: 9 | - output: web,pdf 10 | title: Overview 11 | url: / 12 | - output: web,pdf 13 | title: Pytorch Errors 14 | url: torch.html 15 | - output: web,pdf 16 | title: DataLoader Errors 17 | url: fastai.dataloader.html 18 | - output: web,pdf 19 | title: Learner Errors 20 | url: fastai.learner.html 21 | - output: web,pdf 22 | title: Transform Errors 23 | url: fastai.transform.html 24 | - output: web,pdf 25 | title: Dataset Errors 26 | url: fastai.datasets.html 27 | output: web 28 | title: fastdebug 29 | output: web 30 | title: Sidebar 31 | -------------------------------------------------------------------------------- /docs/_data/topnav.yml: -------------------------------------------------------------------------------- 1 | topnav: 2 | - title: Topnav 3 | items: 4 | - title: github 5 | external_url: https://github.com/muellerzr/fastdebug/tree/master/ 6 | 7 | #Topnav dropdowns 8 | topnav_dropdowns: 9 | - title: Topnav dropdowns 10 | folders: -------------------------------------------------------------------------------- /docs/error.fastai.html: -------------------------------------------------------------------------------- 1 | --- 2 | 3 | title: fastai Errors 4 | 5 | 6 | keywords: fastai 7 | sidebar: home_sidebar 8 | 9 | summary: "In-place fastai specific errors to ease debugging" 10 | description: "In-place fastai specific errors to ease debugging" 11 | nb_path: "01_error.fastai.ipynb" 12 | --- 13 | 22 | 23 |
24 | 25 | {% raw %} 26 | 27 |
28 | 29 |
30 |
31 | 32 |
33 | 34 |
35 |
/mnt/d/lib/python3.7/site-packages/torch/cuda/__init__.py:52: UserWarning: CUDA initialization: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx (Triggered internally at  /pytorch/c10/cuda/CUDAFunctions.cpp:100.)
 36 |   return torch._C._cuda_getDeviceCount() > 0
 37 | 
38 |
39 |
40 | 41 |
42 |
43 | 44 |
45 | {% endraw %} 46 | 47 | {% raw %} 48 | 49 |
50 | 51 |
52 | {% endraw %} 53 | 54 |
55 |
56 |

This notebook contains a series of various errors that can be used when running with fastai. It should be noted here that there is no other imports or magic you need to do to use this section of the library other then: from fastdebug import *. It will automatically load in what's needed.

57 |

As a style choice, we are choosing to do the .* notation as this loads in not only all of our errors, but also replaces sections of fastai's code to inject some error handling (as we'll see later)

58 | 59 |
60 |
61 |
62 |
63 |
64 |

Error Types

65 |
66 |
67 |
68 | {% raw %} 69 | 70 |
71 | 72 |
73 |
74 | 75 |
76 | 77 | 78 |
79 |

loss_func_error[source]

loss_func_error(e:Exception, learn)

80 |
81 |

Error that should be run when there is an issue when working with the loss function

82 |

Raises with a message stating the shapes of the inputs and targs, and the error

83 | 84 |
85 | 86 |
87 | 88 |
89 |
90 | 91 |
92 | {% endraw %} 93 | 94 | {% raw %} 95 | 96 |
97 | 98 |
99 | {% endraw %} 100 | 101 | {% raw %} 102 | 103 |
104 | 105 |
106 |
107 | 108 |
109 | 110 | 111 |
112 |

callback_error[source]

callback_error(e:Exception, cb:str, event_name:str)

113 |
114 |

Raises an error from when a Callback event failed, showing what event, the name of the Callback and the trace

115 | 116 |
117 | 118 |
119 | 120 |
121 |
122 | 123 |
124 | {% endraw %} 125 | 126 | {% raw %} 127 | 128 |
129 | 130 |
131 | {% endraw %} 132 | 133 | {% raw %} 134 | 135 |
136 | 137 |
138 |
139 | 140 |
141 | 142 | 143 |
144 |

catch_pred_errors[source]

catch_pred_errors(e:Exception, model)

145 |
146 |

Catches any errors relating to prediction that are either related to the device or model layers. Else raise e

147 | 148 |
149 | 150 |
151 | 152 |
153 |
154 | 155 |
156 | {% endraw %} 157 | 158 | {% raw %} 159 | 160 |
161 | 162 |
163 | {% endraw %} 164 | 165 | {% raw %} 166 | 167 |
168 | 169 |
170 |
171 | 172 |
173 | 174 | 175 |
176 |

catch_loss_errors[source]

catch_loss_errors(e:Exception, learn)

177 |
178 |

Catches any errors that occur with the loss function and its calculation

179 | 180 |
181 | 182 |
183 | 184 |
185 |
186 | 187 |
188 | {% endraw %} 189 | 190 | {% raw %} 191 | 192 |
193 | 194 |
195 | {% endraw %} 196 | 197 |
198 |
199 |

Modifications and Enhancements to the fastai Source Code and Learner:

200 |
201 |
202 |
203 | {% raw %} 204 | 205 |
206 | 207 |
208 | {% endraw %} 209 | 210 | {% raw %} 211 | 212 |
213 | 214 |
215 | {% endraw %} 216 | 217 | {% raw %} 218 | 219 |
220 | 221 |
222 |
223 | 224 |
225 | 226 | 227 |
228 |

Learner.__init__[source]

Learner.__init__(dls, model, loss_func=None, opt_func=Adam, lr=0.001, splitter=trainable_params, cbs=None, metrics=None, path=None, model_dir='models', wd=None, wd_bn_bias=False, train_bn=True, moms=(0.95, 0.85, 0.95), sanity_check=False, show_table=False)

229 |
230 |

Group together a model, some dls and a loss_func to handle training, potentially run a sanity check

231 | 232 |
233 | 234 |
235 | 236 |
237 |
238 | 239 |
240 | {% endraw %} 241 | 242 | {% raw %} 243 | 244 |
245 | 246 |
247 |
248 | 249 |
250 | 251 | 252 |
253 |

Learner.sanity_check[source]

Learner.sanity_check(show_table=False)

254 |
255 |

Performs a short epoch and uses all the callbacks in self.cbs on the CPU to ensure nothing is broken

256 | 257 |
258 | 259 |
260 | 261 |
262 |
263 | 264 |
265 | {% endraw %} 266 | 267 |
268 |
269 |

With sanity_check, you can make sure that you've set everything up properly and you won't get any issues before pushing to the GPU. This allows you to quickly ensure that you won't get any CUDA device-assist errors, and that the whole training regiment will go well.

270 | 271 |
272 |
273 |
274 | {% raw %} 275 | 276 |
277 | 278 |
279 | {% endraw %} 280 | 281 | {% raw %} 282 | 283 |
284 | 285 |
286 | {% endraw %} 287 | 288 |
289 | 290 | 291 | -------------------------------------------------------------------------------- /docs/fastai.data.html: -------------------------------------------------------------------------------- 1 | --- 2 | 3 | title: DataLoader Errors 4 | 5 | 6 | keywords: fastai 7 | sidebar: home_sidebar 8 | 9 | summary: "Errors and exceptions for any step of the `DataLoader` process" 10 | description: "Errors and exceptions for any step of the `DataLoader` process" 11 | nb_path: "01_fastai.data.ipynb" 12 | --- 13 | 22 | 23 |
24 | 25 | {% raw %} 26 | 27 |
28 | 29 |
30 |
31 | 32 |
33 | 34 |
35 |
/mnt/d/lib/python3.7/site-packages/torch/cuda/__init__.py:52: UserWarning: CUDA initialization: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx (Triggered internally at  /pytorch/c10/cuda/CUDAFunctions.cpp:100.)
 36 |   return torch._C._cuda_getDeviceCount() > 0
 37 | 
38 |
39 |
40 | 41 |
42 |
43 | 44 |
45 | {% endraw %} 46 | 47 |
48 |
49 |

This includes after_item, after_batch, and collating. Anything in relation to the Datasets or anything before the DataLoader process can be found in fastdebug.fastai.dataset

50 | 51 |
52 |
53 |
54 | {% raw %} 55 | 56 |
57 | 58 |
59 | {% endraw %} 60 | 61 | {% raw %} 62 | 63 |
64 | 65 |
66 |
67 | 68 |
69 | 70 | 71 |
72 |

collate_error[source]

collate_error(e:Exception, batch)

73 |
74 |

Raises an explicit error when the batch could not collate, stating 75 | what items in the batch are different sizes and their types

76 | 77 |
78 | 79 |
80 | 81 |
82 |
83 | 84 |
85 | {% endraw %} 86 | 87 | {% raw %} 88 | 89 |
90 | 91 |
92 | {% endraw %} 93 | 94 | {% raw %} 95 | 96 |
97 | 98 |
99 |
100 | 101 |
102 | 103 | 104 |
105 |

DataLoader.create_batch[source]

DataLoader.create_batch(b)

106 |
107 |

Collate a list of items into a batch.

108 | 109 |
110 | 111 |
112 | 113 |
114 |
115 | 116 |
117 | {% endraw %} 118 | 119 | {% raw %} 120 | 121 |
122 | 123 |
124 | {% endraw %} 125 | 126 |
127 |
128 |

collate_error is @patch'd into DataLoader's create_batch function through importing this module, so if there is any possible reason why the data cannot be collated into the batch, it is presented to the user.

129 |

An example is below, where we forgot to include an item transform that resizes all our images to the same size:

130 | 131 |
132 |
133 |
134 | {% raw %} 135 | 136 |
137 |
138 | 139 |
140 |
141 |
from fastai.vision.all import *
142 | path = untar_data(URLs.PETS)/'images'
143 | dls = ImageDataLoaders.from_name_func(
144 |     path, get_image_files(path), valid_pct=0.2,
145 |     label_func=lambda x: x[0].isupper())
146 | 
147 | x,y = dls.train.one_batch()
148 | 
149 | 150 |
151 |
152 |
153 | 154 |
155 |
156 | 157 |
158 | 159 |
160 |
161 | ---------------------------------------------------------------------------
162 | RuntimeError                              Traceback (most recent call last)
163 | <ipython-input-8-c493bee87237> in <module>()
164 |       6     label_func=lambda x: x[0].isupper())
165 |       7 
166 | ----> 8 x,y = dls.train.one_batch()
167 | 
168 | /usr/local/lib/python3.7/dist-packages/fastai/data/load.py in one_batch(self)
169 |     148     def one_batch(self):
170 |     149         if self.n is not None and len(self)==0: raise ValueError(f'This DataLoader does not contain any batches')
171 | --> 150         with self.fake_l.no_multiproc(): res = first(self)
172 |     151         if hasattr(self, 'it'): delattr(self, 'it')
173 |     152         return res
174 | 
175 | /usr/local/lib/python3.7/dist-packages/fastcore/basics.py in first(x, f, negate, **kwargs)
176 |     545     x = iter(x)
177 |     546     if f: x = filter_ex(x, f=f, negate=negate, gen=True, **kwargs)
178 | --> 547     return next(x, None)
179 |     548 
180 |     549 # Cell
181 | 
182 | /usr/local/lib/python3.7/dist-packages/fastai/data/load.py in __iter__(self)
183 |     107         self.before_iter()
184 |     108         self.__idxs=self.get_idxs() # called in context of main process (not workers/subprocesses)
185 | --> 109         for b in _loaders[self.fake_l.num_workers==0](self.fake_l):
186 |     110             # fix issue 2899. If the process start method isn't fork, the data will be copied to cuda in learner one_batch.
187 |     111             if self.device is not None and multiprocessing.get_start_method().lower() == "fork":
188 | 
189 | /usr/local/lib/python3.7/dist-packages/torch/utils/data/dataloader.py in __next__(self)
190 |     433         if self._sampler_iter is None:
191 |     434             self._reset()
192 | --> 435         data = self._next_data()
193 |     436         self._num_yielded += 1
194 |     437         if self._dataset_kind == _DatasetKind.Iterable and \
195 | 
196 | /usr/local/lib/python3.7/dist-packages/torch/utils/data/dataloader.py in _next_data(self)
197 |     473     def _next_data(self):
198 |     474         index = self._next_index()  # may raise StopIteration
199 | --> 475         data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
200 |     476         if self._pin_memory:
201 |     477             data = _utils.pin_memory.pin_memory(data)
202 | 
203 | /usr/local/lib/python3.7/dist-packages/torch/utils/data/_utils/fetch.py in fetch(self, possibly_batched_index)
204 |      32                 raise StopIteration
205 |      33         else:
206 | ---> 34             data = next(self.dataset_iter)
207 |      35         return self.collate_fn(data)
208 |      36 
209 | 
210 | /usr/local/lib/python3.7/dist-packages/fastai/data/load.py in create_batches(self, samps)
211 |     118         if self.dataset is not None: self.it = iter(self.dataset)
212 |     119         res = filter(lambda o:o is not None, map(self.do_item, samps))
213 | --> 120         yield from map(self.do_batch, self.chunkify(res))
214 |     121 
215 |     122     def new(self, dataset=None, cls=None, **kwargs):
216 | 
217 | /usr/local/lib/python3.7/dist-packages/fastai/data/load.py in do_batch(self, b)
218 |     144         else: raise IndexError("Cannot index an iterable dataset numerically - must use `None`.")
219 |     145     def create_batch(self, b): return (fa_collate,fa_convert)[self.prebatched](b)
220 | --> 146     def do_batch(self, b): return self.retain(self.create_batch(self.before_batch(b)), b)
221 |     147     def to(self, device): self.device = device
222 |     148     def one_batch(self):
223 | 
224 | <ipython-input-7-a9809be51294> in create_batch(self, b)
225 |       8     except Exception as e:
226 |       9         if not self.prebatched:
227 | ---> 10             collate_error(e, b)
228 |      11         else: raise e
229 | 
230 | <ipython-input-6-f0b390dbe89c> in collate_error(e, batch)
231 |      23                     err += f'Please include a transform in `after_item` that ensures all data of type {type_a} is the same size'
232 |      24                     e.args = [err]
233 | ---> 25                     raise e
234 | 
235 | <ipython-input-7-a9809be51294> in create_batch(self, b)
236 |       5     func = (fa_collate,fa_convert)[self.prebatched]
237 |       6     try:
238 | ----> 7         return func(b)
239 |       8     except Exception as e:
240 |       9         if not self.prebatched:
241 | 
242 | /usr/local/lib/python3.7/dist-packages/fastai/data/load.py in fa_collate(t)
243 |      48     b = t[0]
244 |      49     return (default_collate(t) if isinstance(b, _collate_types)
245 | ---> 50             else type(t[0])([fa_collate(s) for s in zip(*t)]) if isinstance(b, Sequence)
246 |      51             else default_collate(t))
247 |      52 
248 | 
249 | /usr/local/lib/python3.7/dist-packages/fastai/data/load.py in <listcomp>(.0)
250 |      48     b = t[0]
251 |      49     return (default_collate(t) if isinstance(b, _collate_types)
252 | ---> 50             else type(t[0])([fa_collate(s) for s in zip(*t)]) if isinstance(b, Sequence)
253 |      51             else default_collate(t))
254 |      52 
255 | 
256 | /usr/local/lib/python3.7/dist-packages/fastai/data/load.py in fa_collate(t)
257 |      47     "A replacement for PyTorch `default_collate` which maintains types and handles `Sequence`s"
258 |      48     b = t[0]
259 | ---> 49     return (default_collate(t) if isinstance(b, _collate_types)
260 |      50             else type(t[0])([fa_collate(s) for s in zip(*t)]) if isinstance(b, Sequence)
261 |      51             else default_collate(t))
262 | 
263 | /usr/local/lib/python3.7/dist-packages/torch/utils/data/_utils/collate.py in default_collate(batch)
264 |      53             storage = elem.storage()._new_shared(numel)
265 |      54             out = elem.new(storage)
266 | ---> 55         return torch.stack(batch, 0, out=out)
267 |      56     elif elem_type.__module__ == 'numpy' and elem_type.__name__ != 'str_' \
268 |      57             and elem_type.__name__ != 'string_':
269 | 
270 | /usr/local/lib/python3.7/dist-packages/fastai/torch_core.py in __torch_function__(self, func, types, args, kwargs)
271 |     327         convert=False
272 |     328         if _torch_handled(args, self._opt, func): convert,types = type(self),(torch.Tensor,)
273 | --> 329         res = super().__torch_function__(func, types, args=args, kwargs=kwargs)
274 |     330         if convert: res = convert(res)
275 |     331         if isinstance(res, TensorBase): res.set_meta(self, as_copy=True)
276 | 
277 | /usr/local/lib/python3.7/dist-packages/torch/tensor.py in __torch_function__(cls, func, types, args, kwargs)
278 |     993 
279 |     994         with _C.DisableTorchFunction():
280 | --> 995             ret = func(*args, **kwargs)
281 |     996             return _convert(ret, cls)
282 |     997 
283 | 
284 | RuntimeError: Error when trying to collate the data into batches with fa_collate, at least two tensors in the batch are not the same size.
285 | 
286 | Mismatch found within the 0th axis of the batch and is of type TensorImage:
287 | The first item has shape: torch.Size([3, 500, 333])
288 | The second item has shape: torch.Size([3, 333, 500])
289 | 
290 | Please include a transform in `after_item` that ensures all data of type TensorImage is the same size
291 |
292 |
293 | 294 |
295 |
296 | 297 |
298 | {% endraw %} 299 | 300 | {% raw %} 301 | 302 |
303 | 304 |
305 |
306 | 307 |
308 | 309 | 310 |
311 |

TfmdDL.new[source]

TfmdDL.new(dataset=None)

312 |
313 |

Create a new version of self with a few changed attributes

314 | 315 |
316 | 317 |
318 | 319 |
320 |
321 | 322 |
323 | {% endraw %} 324 | 325 | {% raw %} 326 | 327 |
328 | 329 |
330 | {% endraw %} 331 | 332 |
333 | 334 | 335 | -------------------------------------------------------------------------------- /docs/fastai.learner.html: -------------------------------------------------------------------------------- 1 | --- 2 | 3 | title: Learner Errors 4 | 5 | 6 | keywords: fastai 7 | sidebar: home_sidebar 8 | 9 | summary: "In-place fastai specific errors to ease debugging" 10 | description: "In-place fastai specific errors to ease debugging" 11 | nb_path: "02_fastai.learner.ipynb" 12 | --- 13 | 22 | 23 |
24 | 25 | {% raw %} 26 | 27 |
28 | 29 |
30 |
31 | 32 |
33 | 34 |
35 |
/mnt/d/lib/python3.7/site-packages/torch/cuda/__init__.py:52: UserWarning: CUDA initialization: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx (Triggered internally at  /pytorch/c10/cuda/CUDAFunctions.cpp:100.)
 36 |   return torch._C._cuda_getDeviceCount() > 0
 37 | 
38 |
39 |
40 | 41 |
42 |
43 | 44 |
45 | {% endraw %} 46 | 47 | {% raw %} 48 | 49 |
50 | 51 |
52 | {% endraw %} 53 | 54 |
55 |
56 |

This notebook contains a series of various errors that can be used when running with fastai. It should be noted here that there is no other imports or magic you need to do to use this section of the library other then: from fastdebug import *. It will automatically load in what's needed.

57 |

As a style choice, we are choosing to do the .* notation as this loads in not only all of our errors, but also replaces sections of fastai's code to inject some error handling (as we'll see later)

58 | 59 |
60 |
61 |
62 |
63 |
64 |

Error Types

65 |
66 |
67 |
68 | {% raw %} 69 | 70 |
71 | 72 |
73 |
74 | 75 |
76 | 77 | 78 |
79 |

loss_func_error[source]

loss_func_error(e:Exception, learn)

80 |
81 |

Error that should be run when there is an issue when working with the loss function

82 |

Raises with a message stating the shapes of the inputs and targs, and the error

83 | 84 |
85 | 86 |
87 | 88 |
89 |
90 | 91 |
92 | {% endraw %} 93 | 94 | {% raw %} 95 | 96 |
97 | 98 |
99 | {% endraw %} 100 | 101 | {% raw %} 102 | 103 |
104 | 105 |
106 |
107 | 108 |
109 | 110 | 111 |
112 |

callback_error[source]

callback_error(e:Exception, cb:str, event_name:str)

113 |
114 |

Raises an error from when a Callback event failed, showing what event, the name of the Callback and the trace

115 | 116 |
117 | 118 |
119 | 120 |
121 |
122 | 123 |
124 | {% endraw %} 125 | 126 | {% raw %} 127 | 128 |
129 | 130 |
131 | {% endraw %} 132 | 133 | {% raw %} 134 | 135 |
136 | 137 |
138 |
139 | 140 |
141 | 142 | 143 |
144 |

catch_pred_errors[source]

catch_pred_errors(e:Exception, model)

145 |
146 |

Catches any errors relating to prediction that are either related to the device or model layers. Else raise e

147 | 148 |
149 | 150 |
151 | 152 |
153 |
154 | 155 |
156 | {% endraw %} 157 | 158 | {% raw %} 159 | 160 |
161 | 162 |
163 | {% endraw %} 164 | 165 | {% raw %} 166 | 167 |
168 | 169 |
170 |
171 | 172 |
173 | 174 | 175 |
176 |

catch_loss_errors[source]

catch_loss_errors(e:Exception, learn)

177 |
178 |

Catches any errors that occur with the loss function and its calculation

179 | 180 |
181 | 182 |
183 | 184 |
185 |
186 | 187 |
188 | {% endraw %} 189 | 190 | {% raw %} 191 | 192 |
193 | 194 |
195 | {% endraw %} 196 | 197 |
198 |
199 |

Modifications and Enhancements to the fastai Source Code and Learner:

200 |
201 |
202 |
203 | {% raw %} 204 | 205 |
206 | 207 |
208 | {% endraw %} 209 | 210 | {% raw %} 211 | 212 |
213 | 214 |
215 | {% endraw %} 216 | 217 | {% raw %} 218 | 219 |
220 | 221 |
222 |
223 | 224 |
225 | 226 | 227 |
228 |

Learner.__init__[source]

Learner.__init__(dls, model, loss_func=None, opt_func=Adam, lr=0.001, splitter=trainable_params, cbs=None, metrics=None, path=None, model_dir='models', wd=None, wd_bn_bias=False, train_bn=True, moms=(0.95, 0.85, 0.95), sanity_check=False, show_table=False)

229 |
230 |

Group together a model, some dls and a loss_func to handle training, potentially run a sanity check

231 | 232 |
233 | 234 |
235 | 236 |
237 |
238 | 239 |
240 | {% endraw %} 241 | 242 | {% raw %} 243 | 244 |
245 | 246 |
247 |
248 | 249 |
250 | 251 | 252 |
253 |

Learner.sanity_check[source]

Learner.sanity_check(show_table=False)

254 |
255 |

Performs a short epoch and uses all the callbacks in self.cbs on the CPU to ensure nothing is broken

256 | 257 |
258 | 259 |
260 | 261 |
262 |
263 | 264 |
265 | {% endraw %} 266 | 267 |
268 |
269 |

With sanity_check, you can make sure that you've set everything up properly and you won't get any issues before pushing to the GPU. This allows you to quickly ensure that you won't get any CUDA device-assist errors, and that the whole training regiment will go well.

270 | 271 |
272 |
273 |
274 | {% raw %} 275 | 276 |
277 | 278 |
279 | {% endraw %} 280 | 281 | {% raw %} 282 | 283 |
284 | 285 |
286 | {% endraw %} 287 | 288 | {% raw %} 289 | 290 |
291 | 292 |
293 |
294 | 295 |
296 | 297 | 298 |
299 |

module_error[source]

module_error(e:AttributeError)

300 |
301 |

Raises an error when trying to load in a previous Learner and custom functions were not available in the namespace

302 | 303 |
304 | 305 |
306 | 307 |
308 |
309 | 310 |
311 | {% endraw %} 312 | 313 | {% raw %} 314 | 315 |
316 | 317 |
318 | {% endraw %} 319 | 320 | {% raw %} 321 | 322 |
323 | 324 |
325 |
326 | 327 |
328 | 329 | 330 |
331 |

load_learner[source]

load_learner(fname, cpu=True, pickle_module=pickle)

332 |
333 |

Load a Learner object in fname, optionally putting it on the cpu

334 | 335 |
336 | 337 |
338 | 339 |
340 |
341 | 342 |
343 | {% endraw %} 344 | 345 | {% raw %} 346 | 347 |
348 | 349 |
350 | {% endraw %} 351 | 352 |
353 |
354 |

We have a custom load_learner function here that can check if everything exported is available when bringing the model in, if not then it'll raise an explicit error

355 | 356 |
357 |
358 |
359 |
360 | 361 | 362 | -------------------------------------------------------------------------------- /docs/feed.xml: -------------------------------------------------------------------------------- 1 | --- 2 | search: exclude 3 | layout: none 4 | --- 5 | 6 | 7 | 8 | 9 | {{ site.title | xml_escape }} 10 | {{ site.description | xml_escape }} 11 | {{ site.url }}/ 12 | 13 | {{ site.time | date_to_rfc822 }} 14 | {{ site.time | date_to_rfc822 }} 15 | Jekyll v{{ jekyll.version }} 16 | {% for post in site.posts limit:10 %} 17 | 18 | {{ post.title | xml_escape }} 19 | {{ post.content | xml_escape }} 20 | {{ post.date | date_to_rfc822 }} 21 | {{ post.url | prepend: site.url }} 22 | {{ post.url | prepend: site.url }} 23 | {% for tag in post.tags %} 24 | {{ tag | xml_escape }} 25 | {% endfor %} 26 | {% for tag in page.tags %} 27 | {{ cat | xml_escape }} 28 | {% endfor %} 29 | 30 | {% endfor %} 31 | 32 | 33 | -------------------------------------------------------------------------------- /docs/index.html: -------------------------------------------------------------------------------- 1 | --- 2 | 3 | title: fastdebug 4 | 5 | 6 | keywords: fastai 7 | sidebar: home_sidebar 8 | 9 | summary: "A helpful library for improving torch and fastai errors" 10 | description: "A helpful library for improving torch and fastai errors" 11 | nb_path: "index.ipynb" 12 | --- 13 | 22 | 23 |
24 | 25 | {% raw %} 26 | 27 |
28 | 29 |
30 | {% endraw %} 31 | 32 |
33 |
34 |

Install

35 |
36 |
37 |
38 |
39 |
40 |

pip install fastdebug

41 | 42 |
43 |
44 |
45 |
46 |
47 |

How to use

48 |
49 |
50 |
51 |
52 |
53 |

fastdebug is designed around improving the quality of life when dealing with Pytorch and fastai errors, while also including some new sanity checks (fastai only)

54 | 55 |
56 |
57 |
58 |
59 |
60 |

Pytorch

Pytorch now has:

61 | 65 |

Both can be imported with:

66 |
from fastdebug.error.torch import device_error, layer_error
 67 | 
68 |

device_error prints out a much more readable error for when two tensors aren't on the same device:

69 | 70 |
71 |
72 |
73 |
74 |
75 |
inp = torch.rand().cuda()
 76 | model = model.cpu()
 77 | try:
 78 |     _ = model(inp)
 79 | except Exception as e:
 80 |     device_error(e, 'Input type', 'Model weights')
 81 | 
82 |

And our new log:

83 |
---------------------------------------------------------------------------
 84 | RuntimeError                              Traceback (most recent call last)
 85 | <ipython-input-28-981e0ace9c38> in <module>()
 86 |       2     model(x)
 87 |       3 except Exception as e:
 88 | ----> 4     device_error(e, 'Input type', 'Model weights')
 89 | 
 90 | 10 frames
 91 | /usr/local/lib/python3.7/dist-packages/torch/tensor.py in __torch_function__(cls, func, types, args, kwargs)
 92 |     993 
 93 |     994         with _C.DisableTorchFunction():
 94 | --> 995             ret = func(*args, **kwargs)
 95 |     996             return _convert(ret, cls)
 96 |     997 
 97 | 
 98 | RuntimeError: Mismatch between weight types
 99 | 
100 | Input type has type:         (torch.cuda.FloatTensor)
101 | Model weights have type:     (torch.FloatTensor)
102 | 
103 | Both should be the same.
104 | 
105 | 106 |
107 |
108 |
109 |
110 |
111 |

And with layer_error, if there is a shape mismatch it will attempt to find the right layer it was at:

112 |
inp = torch.rand(5,2, 3)
113 | try:
114 |     m(inp)
115 | except Exception as e:
116 |     layer_error(e, m)
117 | 
118 |
---------------------------------------------------------------------------
119 | RuntimeError                              Traceback (most recent call last)
120 | <ipython-input-84-d4ab91131841> in <module>()
121 |       3     m(inp)
122 |       4 except Exception as e:
123 | ----> 5     layer_error(e, m)
124 | 
125 | <ipython-input-83-ca2dc02cfff4> in layer_error(e, model)
126 |       8     i, layer = get_layer_by_shape(model, shape)
127 |       9     e.args = [f'Size mismatch between input tensors and what the model expects\n\n{args}\n\tat layer {i}: {layer}']
128 | ---> 10     raise e
129 | 
130 | <ipython-input-84-d4ab91131841> in <module>()
131 |       1 inp = torch.rand(5,2, 3)
132 |       2 try:
133 | ----> 3     m(inp)
134 |       4 except Exception as e:
135 |       5     layer_error(e, m)
136 | 
137 | /mnt/d/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
138 |     725             result = self._slow_forward(*input, **kwargs)
139 |     726         else:
140 | --> 727             result = self.forward(*input, **kwargs)
141 |     728         for hook in itertools.chain(
142 |     729                 _global_forward_hooks.values(),
143 | 
144 | /mnt/d/lib/python3.7/site-packages/torch/nn/modules/container.py in forward(self, input)
145 |     115     def forward(self, input):
146 |     116         for module in self:
147 | --> 117             input = module(input)
148 |     118         return input
149 |     119 
150 | 
151 | /mnt/d/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
152 |     725             result = self._slow_forward(*input, **kwargs)
153 |     726         else:
154 | --> 727             result = self.forward(*input, **kwargs)
155 |     728         for hook in itertools.chain(
156 |     729                 _global_forward_hooks.values(),
157 | 
158 | /mnt/d/lib/python3.7/site-packages/torch/nn/modules/conv.py in forward(self, input)
159 |     421 
160 |     422     def forward(self, input: Tensor) -> Tensor:
161 | --> 423         return self._conv_forward(input, self.weight)
162 |     424 
163 |     425 class Conv3d(_ConvNd):
164 | 
165 | /mnt/d/lib/python3.7/site-packages/torch/nn/modules/conv.py in _conv_forward(self, input, weight)
166 |     418                             _pair(0), self.dilation, self.groups)
167 |     419         return F.conv2d(input, weight, self.bias, self.stride,
168 | --> 420                         self.padding, self.dilation, self.groups)
169 |     421 
170 |     422     def forward(self, input: Tensor) -> Tensor:
171 | 
172 | RuntimeError: Size mismatch between input tensors and what the model expects
173 | 
174 | Model expected 4-dimensional input for 4-dimensional weight [3, 3, 1, 1], but got 3-dimensional input of size [5, 2, 3] instead
175 |     at layer 1: Conv2d(3, 3, kernel_size=(1, 1), stride=(1, 1))
176 | 
177 | 178 |
179 |
180 |
181 |
182 |
183 |

fastai

Along with the additions above (and are used during fit), fastai now has a Learner.sanity_check function, which allows you to quickly perform a basic check to ensure that your call to fit won't raise any exceptions. They are performed on the CPU for a partial epoch to make sure that CUDA device-assist errors can be preemptively found.

184 |

To use it simply do:

185 |
from fastdebug.fastai import *
186 | from fastai.vision.all import *
187 | 
188 | learn = Learner(...)
189 | learn.sanity_check()
190 | 
191 |

This is also now an argument in Learner, set to False by default, so that after making your Learner a quick check is ensured.

192 | 193 |
194 |
195 |
196 |
197 |
198 |
learn = Learner(..., sanity_check=True)
199 | 
200 | 201 |
202 |
203 |
204 |
205 | 206 | 207 | -------------------------------------------------------------------------------- /docs/sidebar.json: -------------------------------------------------------------------------------- 1 | { 2 | "fastdebug": { 3 | "Overview": "/", 4 | "Pytorch Errors": "torch.html", 5 | "DataLoader Errors": "fastai.dataloader.html", 6 | "Learner Errors": "fastai.learner.html", 7 | "Transform Errors": "fastai.transform.html", 8 | "Dataset Errors": "fastai.datasets.html" 9 | } 10 | } -------------------------------------------------------------------------------- /docs/sitemap.xml: -------------------------------------------------------------------------------- 1 | --- 2 | layout: none 3 | search: exclude 4 | --- 5 | 6 | 7 | 8 | {% for post in site.posts %} 9 | {% unless post.search == "exclude" %} 10 | 11 | {{site.url}}{{post.url}} 12 | 13 | {% endunless %} 14 | {% endfor %} 15 | 16 | 17 | {% for page in site.pages %} 18 | {% unless page.search == "exclude" %} 19 | 20 | {{site.url}}{{ page.url}} 21 | 22 | {% endunless %} 23 | {% endfor %} 24 | -------------------------------------------------------------------------------- /fastdebug/__init__.py: -------------------------------------------------------------------------------- 1 | __version__ = "0.1.3" 2 | 3 | from fastdebug.torch import * 4 | from fastdebug.fastai import * -------------------------------------------------------------------------------- /fastdebug/_nbdev.py: -------------------------------------------------------------------------------- 1 | # AUTOGENERATED BY NBDEV! DO NOT EDIT! 2 | 3 | __all__ = ["index", "modules", "custom_doc_links", "git_url"] 4 | 5 | index = {"device_error": "00_torch.ipynb", 6 | "hook_fn": "00_torch.ipynb", 7 | "PreHook": "00_torch.ipynb", 8 | "ForwardHooks": "00_torch.ipynb", 9 | "hook_outputs": "00_torch.ipynb", 10 | "layer_error": "00_torch.ipynb", 11 | "collate_error": "01_fastai.dataloader.ipynb", 12 | "DataLoader.create_batch": "01_fastai.dataloader.ipynb", 13 | "TfmdDL.new": "03_fastai.transform.ipynb", 14 | "loss_func_error": "02_fastai.learner.ipynb", 15 | "callback_error": "02_fastai.learner.ipynb", 16 | "catch_pred_errors": "02_fastai.learner.ipynb", 17 | "catch_loss_errors": "02_fastai.learner.ipynb", 18 | "Learner.sanity_check": "02_fastai.learner.ipynb", 19 | "Learner.__init__": "02_fastai.learner.ipynb", 20 | "module_error": "02_fastai.learner.ipynb", 21 | "load_learner": "02_fastai.learner.ipynb", 22 | "transform_error": "03_fastai.transform.ipynb", 23 | "Transform.__call__": "03_fastai.transform.ipynb", 24 | "Transform.decode": "03_fastai.transform.ipynb", 25 | "TfmdLists.__init__": "04_fastai.datasets.ipynb", 26 | "subset_error": "04_fastai.datasets.ipynb", 27 | "TfmdLists.subset": "04_fastai.datasets.ipynb", 28 | "TfmdLists.setup": "04_fastai.datasets.ipynb"} 29 | 30 | modules = ["torch.py", 31 | "fastai/dataloader.py", 32 | "fastai/learner.py", 33 | "fastai/transform.py", 34 | "fastai/datasets.py"] 35 | 36 | doc_url = "https://muellerzr.github.io/fastdebug/" 37 | 38 | git_url = "https://github.com/muellerzr/fastdebug/tree/master/" 39 | 40 | def custom_doc_links(name): return None 41 | -------------------------------------------------------------------------------- /fastdebug/fastai/__init__.py: -------------------------------------------------------------------------------- 1 | from .dataloader import * 2 | from .learner import * 3 | from .transform import * 4 | from .datasets import * -------------------------------------------------------------------------------- /fastdebug/fastai/dataloader.py: -------------------------------------------------------------------------------- 1 | # AUTOGENERATED! DO NOT EDIT! File to edit: 01_fastai.dataloader.ipynb (unless otherwise specified). 2 | 3 | __all__ = ['collate_error'] 4 | 5 | # Cell 6 | import inflect 7 | from fastcore.basics import patch 8 | from fastai.data.core import TfmdDL 9 | from fastai.data.load import DataLoader, fa_collate, fa_convert 10 | 11 | # Cell 12 | def collate_error(e:Exception, batch): 13 | """ 14 | Raises an explicit error when the batch could not collate, stating 15 | what items in the batch are different sizes and their types 16 | """ 17 | p = inflect.engine() 18 | err = f'Error when trying to collate the data into batches with fa_collate, ' 19 | err += 'at least two tensors in the batch are not the same size.\n\n' 20 | # we need to iterate through the entire batch and find a mismatch 21 | length = len(batch[0]) 22 | for idx in range(length): # for each type in the batch 23 | for i, item in enumerate(batch): 24 | if i == 0: 25 | shape_a = item[idx].shape 26 | type_a = item[idx].__class__.__name__ 27 | elif item[idx].shape != shape_a: 28 | shape_b = item[idx].shape 29 | if shape_a != shape_b: 30 | err += f'Mismatch found within the {p.ordinal(idx)} axis of the batch and is of type {type_a}:\n' 31 | err += f'The first item has shape: {shape_a}\n' 32 | err += f'The {p.number_to_words(p.ordinal(i+1))} item has shape: {shape_b}\n\n' 33 | err += f'Please include a transform in `after_item` that ensures all data of type {type_a} is the same size' 34 | e.args = [err] 35 | raise e 36 | 37 | # Cell 38 | @patch 39 | def create_batch(self:DataLoader, b): 40 | "Collate a list of items into a batch." 41 | func = (fa_collate,fa_convert)[self.prebatched] 42 | try: 43 | return func(b) 44 | except Exception as e: 45 | if not self.prebatched: 46 | collate_error(e, b) 47 | else: raise e 48 | 49 | # Cell 50 | @patch 51 | def new(self:TfmdDL, dataset=None, cls=None, **kwargs): 52 | res = super(TfmdDL, self).new(dataset, cls, do_setup=False, **kwargs) 53 | if not hasattr(self, '_n_inp') or not hasattr(self, '_types'): 54 | try: 55 | self._one_pass() 56 | res._n_inp,res._types = self._n_inp,self._types 57 | except Exception as e: 58 | print("Could not do one pass in your dataloader, there is something wrong in it") 59 | raise e 60 | else: res._n_inp,res._types = self._n_inp,self._types 61 | return res -------------------------------------------------------------------------------- /fastdebug/fastai/datasets.py: -------------------------------------------------------------------------------- 1 | # AUTOGENERATED! DO NOT EDIT! File to edit: 04_fastai.datasets.ipynb (unless otherwise specified). 2 | 3 | __all__ = ['subset_error'] 4 | 5 | # Cell 6 | from fastcore.basics import patch, store_attr 7 | from fastcore.foundation import L, mask2idxs 8 | from fastcore.transform import Pipeline 9 | from fastcore.xtras import is_listy 10 | 11 | from fastai.imports import pv 12 | from fastai.data.core import TfmdLists 13 | 14 | # Cell 15 | @patch 16 | def __init__(self:TfmdLists, items, tfms, use_list=None, do_setup=True, split_idx=None, train_setup=True, 17 | splits=None, types=None, verbose=False, dl_type=None): 18 | if items is None or len(items) == 0: raise IndexError('Items passed in either has a length of zero or is None') 19 | super(TfmdLists, self).__init__(items, use_list=use_list) 20 | if dl_type is not None: self._dl_type = dl_type 21 | self.splits = L([slice(None),[]] if splits is None else splits).map(mask2idxs) 22 | if isinstance(tfms,TfmdLists): tfms = tfms.tfms 23 | if isinstance(tfms,Pipeline): do_setup=False 24 | self.tfms = Pipeline(tfms, split_idx=split_idx) 25 | store_attr('types,split_idx') 26 | if do_setup: 27 | pv(f"Setting up {self.tfms}", verbose) 28 | self.setup(train_setup=train_setup) 29 | 30 | # Cell 31 | def subset_error(e:IndexError, i:int) -> IndexError: 32 | """ 33 | IndexError when attempting to grab a non-existant subset in the dataset at index `i` 34 | """ 35 | args = e.args[0] 36 | err = f'Tried to grab subset {i} in the Dataset, but it contains no items.\n\n' 37 | err += args 38 | e.args = [err] 39 | raise e 40 | 41 | # Cell 42 | @patch 43 | def subset(self:TfmdLists, i:int): 44 | "New `TfmdLists` with same tfms that only includes items in `i`th split" 45 | try: return self._new(self._get(self.splits[i]), split_idx=i) 46 | except IndexError as e: subset_error(e, i) 47 | 48 | # Cell 49 | @patch 50 | def setup(self:TfmdLists, train_setup=True): 51 | "Transform setup with self" 52 | self.tfms.setup(self, train_setup) 53 | if len(self) != 0: 54 | x = super(TfmdLists, self).__getitem__(0) if self.splits is None else super(TfmdLists, self).__getitem__(self.splits[0])[0] 55 | self.types = [] 56 | for f in self.tfms.fs: 57 | self.types.append(getattr(f, 'input_types', type(x))) 58 | x = f(x) 59 | self.types.append(type(x)) 60 | t = getattr(self, 'types', []) 61 | if t is None or len(t) == 0: raise Exception("The stored dataset contains no items and `self.types` has not been setup yet") 62 | types = L(t if is_listy(t) else [t] for t in self.types).concat().unique() 63 | self.pretty_types = '\n'.join([f' - {t}' for t in types]) -------------------------------------------------------------------------------- /fastdebug/fastai/learner.py: -------------------------------------------------------------------------------- 1 | # AUTOGENERATED! DO NOT EDIT! File to edit: 02_fastai.learner.ipynb (unless otherwise specified). 2 | 3 | __all__ = ['loss_func_error', 'callback_error', 'catch_pred_errors', 'catch_loss_errors', 'module_error', 4 | 'load_learner'] 5 | 6 | # Cell 7 | from ..torch import layer_error, device_error 8 | 9 | from fastai.data.all import * 10 | from fastai.optimizer import * 11 | from fastai.learner import * 12 | from fastai.callback.core import event 13 | from fastai.callback.training import ShortEpochCallback 14 | from fastai.torch_core import default_device 15 | 16 | 17 | from fastcore.basics import patch 18 | from fastcore.meta import delegates 19 | 20 | # Cell 21 | def loss_func_error(e:Exception, learn) -> Exception: 22 | """ 23 | Error that should be run when there is an issue when working with the loss function 24 | 25 | Raises with a message stating the shapes of the inputs and targs, and the error 26 | """ 27 | err = f'There was an issue with calculating the loss with `{getattr(learn.loss_func, "__name__", learn.loss_func)}`' 28 | err += f'\n\nPrediction shape(s): {[p.shape for p in listify(learn.pred)]}' 29 | err += f'\nLabel Shape(s): {[y.shape for y in learn.yb]}' 30 | err += f'\nError: {e.args[0]}' 31 | e.args = [err] 32 | raise e 33 | 34 | # Cell 35 | def callback_error(e:Exception, cb:str, event_name:str) -> Exception: 36 | """ 37 | Raises an error from when a Callback event failed, showing what event, the name of the Callback and the trace 38 | """ 39 | e.args = [f"Exception raised in the {cb} Callback during {event_name}:\n\n{e.args[0]}"] 40 | raise e 41 | 42 | # Cell 43 | def catch_pred_errors(e:Exception, model) -> Exception: 44 | "Catches any errors relating to prediction that are either related to the device or model layers. Else raise `e`" 45 | if "Input type" in e.args[0]: device_error(e, 'Input', 'Model weights') 46 | elif "Expected" in e.args[0]: layer_error(e, model) 47 | else: raise e # anything else 48 | 49 | # Cell 50 | def catch_loss_errors(e:Exception, learn): 51 | "Catches any errors that occur with the loss function and its calculation" 52 | if "Input type" in e.args[0]: device_error(e, 'Model prediction', 'Truths') 53 | else: loss_func_error(e, learn) 54 | 55 | # Cell 56 | @patch 57 | def sanity_check(self:Learner, show_table=False): 58 | "Performs a short epoch and uses all the callbacks in `self.cbs` on the CPU to ensure nothing is broken" 59 | device = getattr(self.dls, 'device', default_device()) 60 | if hasattr(self.dls, 'device'): 61 | self.dls.device = 'cpu' 62 | else: 63 | # Using raw torch 64 | self.model.to('cpu') 65 | self.save('tmp') 66 | cbs = [ShortEpochCallback(short_valid=False)] 67 | if show_table: 68 | with self.no_bar(), self.no_logging(): 69 | self.fit(1, cbs=cbs) 70 | else: 71 | self.fit(1, cbs=cbs) 72 | if hasattr(self.dls, 'device'): 73 | self.dls.device = device 74 | else: 75 | self.model.to(device) 76 | self.load('tmp') 77 | 78 | # Cell 79 | @patch 80 | @delegates(Learner.sanity_check) 81 | def __init__(self:Learner, dls, model, loss_func=None, opt_func=Adam, lr=defaults.lr, splitter=trainable_params, cbs=None, 82 | metrics=None, path=None, model_dir='models', wd=None, wd_bn_bias=False, train_bn=True, 83 | moms=(0.95,0.85,0.95), sanity_check=False, **kwargs): 84 | "Group together a `model`, some `dls` and a `loss_func` to handle training, potentially run a sanity check" 85 | path = Path(path) if path is not None else getattr(dls, 'path', Path('.')) 86 | if loss_func is None: 87 | loss_func = getattr(dls.train_ds, 'loss_func', None) 88 | assert loss_func is not None, "Could not infer loss function from the data, please pass a loss function." 89 | self.dls,self.model = dls,model 90 | store_attr(but='dls,model,cbs') 91 | self.training,self.create_mbar,self.logger,self.opt,self.cbs = False,True,print,None,L() 92 | self.add_cbs(L(defaults.callbacks)+L(cbs)) 93 | self("after_create") 94 | if sanity_check: self.sanity_check(**kwargs) 95 | 96 | # Cell 97 | @patch 98 | def _do_one_batch(self:Learner): 99 | try: 100 | self.pred = self.model(*self.xb) 101 | except RuntimeError as e: 102 | catch_pred_errors(e, self.model) 103 | self('after_pred') 104 | if len(self.yb): 105 | try: 106 | self.loss_grad = self.loss_func(self.pred, *self.yb) 107 | except Exception as e: 108 | catch_loss_errors(e, self) 109 | self.loss = self.loss_grad.clone() 110 | self('after_loss') 111 | if not self.training or not len(self.yb): return 112 | self('before_backward') 113 | self.loss_grad.backward() 114 | self._with_events(self.opt.step, 'step', CancelStepException) 115 | self.opt.zero_grad() 116 | 117 | # Cell 118 | @patch 119 | def _call_one(self:Learner, event_name): 120 | if not hasattr(event, event_name): raise Exception(f'missing {event_name}') 121 | for cb in self.cbs.sorted('order'): 122 | try: 123 | cb(event_name) 124 | except Exception as e: 125 | callback_error(e, cb.__repr__(), event_name) 126 | 127 | # Cell 128 | def module_error(e:AttributeError) -> AttributeError: 129 | """ 130 | Raises an error when trying to load in a previous `Learner` and custom functions were not available in the namespace 131 | """ 132 | args = e.args[0] 133 | err = 'Custom classes or functions exported with your `Learner` are not available in the namespace currently.\n' 134 | err += 'Please re-declare them before calling `load_learner`:\n\n' 135 | err += args 136 | e.args = [err] 137 | raise e 138 | 139 | # Cell 140 | def load_learner(fname, cpu=True, pickle_module=pickle): 141 | "Load a `Learner` object in `fname`, optionally putting it on the `cpu`" 142 | distrib_barrier() 143 | try: res = torch.load(fname, map_location='cpu' if cpu else None, pickle_module=pickle_module) 144 | except AttributeError as e: module_error(e) 145 | if hasattr(res, 'to_fp32'): res = res.to_fp32() 146 | if cpu: res.dls.cpu() 147 | return res -------------------------------------------------------------------------------- /fastdebug/fastai/transform.py: -------------------------------------------------------------------------------- 1 | # AUTOGENERATED! DO NOT EDIT! File to edit: 03_fastai.transform.ipynb (unless otherwise specified). 2 | 3 | __all__ = ['transform_error'] 4 | 5 | # Cell 6 | from fastai.data.core import TfmdDL 7 | from fastai.data.load import DataLoader 8 | 9 | from fastcore.basics import patch 10 | from fastcore.meta import delegates 11 | from fastcore.transform import _get_name, Transform 12 | 13 | # Cell 14 | def transform_error(e:Exception, nm:str, event:str) -> Exception: 15 | """ 16 | Raises Exception `e` stemming from a Transform with more information 17 | 18 | - `nm`: The name of the Transform 19 | - `event`: The event called (such as `encodes` or `decodes`) 20 | """ 21 | err = f'There was an issue calling the {event} on transform {nm}:\n\n' 22 | err += e.args[0] 23 | e.args = [err] 24 | raise e 25 | 26 | # Cell 27 | @patch 28 | def __call__(self:Transform, x, **kwargs): 29 | try: 30 | return self._call('encodes', x, **kwargs) 31 | except Exception as e: 32 | transform_error(e, _get_name(self), 'encodes') 33 | 34 | # Cell 35 | @patch 36 | def decode(self:Transform, x, **kwargs): 37 | "Delegate to decodes to undo transform" 38 | try: 39 | return self._call('decodes', x, **kwargs) 40 | except Exception as e: 41 | transform_error(e, _get_name(self), 'decodes') 42 | 43 | # Cell 44 | @patch 45 | @delegates(DataLoader.new) 46 | def new(self:TfmdDL, dataset=None, cls=None, **kwargs): 47 | "Create a new version of self with a few changed attributes" 48 | res = super(TfmdDL, self).new(dataset, cls, do_setup=False, **kwargs) 49 | if not hasattr(self, '_n_inp') or not hasattr(self, '_types'): 50 | try: 51 | self._one_pass() 52 | res._n_inp,res._types = self._n_inp,self._types 53 | except Exception as e: 54 | print("Could not do one pass in your DataLoader, there is something wrong in it. Please see the stack trace below:") 55 | raise e 56 | else: res._n_inp,res._types = self._n_inp,self._types 57 | return res -------------------------------------------------------------------------------- /fastdebug/torch.py: -------------------------------------------------------------------------------- 1 | # AUTOGENERATED! DO NOT EDIT! File to edit: 00_torch.ipynb (unless otherwise specified). 2 | 3 | __all__ = ['device_error', 'hook_fn', 'PreHook', 'ForwardHooks', 'hook_outputs', 'layer_error'] 4 | 5 | # Cell 6 | import torch 7 | import re 8 | from fastai.callback.hook import Hook 9 | from fastai.torch_core import to_detach 10 | from fastai.layers import flatten_model 11 | 12 | from fastcore.basics import store_attr 13 | 14 | # Cell 15 | def device_error(e:Exception, a:str, b:str) -> Exception: 16 | """ 17 | Verbose error for if `a` and `b` are on different devices 18 | Should be used when checking if a model is on the same device, or two tensors 19 | """ 20 | inp, weight, _ = e.args[0].replace('( ', '').split(')') 21 | inp = inp.replace('Input type', f'{a} has type: \t\t') 22 | weight = weight.replace(' and weight type', f'{b} have type: \t') 23 | err = f'Mismatch between weight types\n\n{inp})\n{weight})\n\nBoth should be the same.' 24 | e.args = [err] 25 | raise e 26 | 27 | # Cell 28 | def hook_fn(m, i): 29 | "Simple hook fn to return the layer" 30 | return m 31 | 32 | # Cell 33 | class PreHook(Hook): 34 | "Creates and registers a hook on `m` with `hook_func` as a forward pre_hook" 35 | def __init__(self, m, hook_func, is_forward=True, detach=True, cpu=False, gather=False): 36 | store_attr('hook_func,detach,cpu,gather') 37 | f = m.register_forward_pre_hook if is_forward else m.register_backward_pre_hook 38 | self.hook = f(self.hook_fn) 39 | self.stored,self.removed = None, False 40 | 41 | def hook_fn(self, module, inp): 42 | "Applies `hook_fn` to `module` and `inp`" 43 | if self.detach: 44 | inp = to_detach(inp, cpu=self.cpu, gather=self.gather) 45 | self.stored = self.hook_func(module, inp) 46 | 47 | # Cell 48 | class ForwardHooks(): 49 | "Create several forward-hooks on the modules in `ms` with `hook_func`" 50 | def __init__(self, ms, hook_func, is_forward=True, detach=True, cpu=False): 51 | self.hooks = [] 52 | for i, m in enumerate(flatten_model(ms)): 53 | self.hooks.append(PreHook(m, hook_func, is_forward, detach, cpu)) 54 | 55 | # Cell 56 | def hook_outputs(modules, detach=True, cpu=False, grad=False): 57 | "Return `Hooks` that store activations of all `modules` in `self.stored`" 58 | return ForwardHooks(modules, hook_fn, detach=detach, cpu=cpu, is_forward=not grad) 59 | 60 | # Cell 61 | def layer_error(e:Exception, model, *inp) -> Exception: 62 | """ 63 | Verbose error for when there is a size mismatch between some input and the model. 64 | `model` should be any torch model 65 | `inp` is the input that went to the model 66 | """ 67 | args = e.args[0].replace("Expected", "Model expected") 68 | hooks = hook_outputs(model) 69 | try: 70 | _ = model(*inp) 71 | except: 72 | pass 73 | finally: 74 | layers,num = [], 0 75 | for i, layer in enumerate(hooks.hooks): 76 | if layer.stored is not None: 77 | layers.append(layer.stored) 78 | num += 1 79 | layer = layers[-1] 80 | [h.remove() for h in hooks.hooks] 81 | e.args = [f'Size mismatch between input tensors and what the model expects\n{"-"*76}\nLayer: {i}, {layer}\nError: {args}'] 82 | raise e -------------------------------------------------------------------------------- /index.ipynb: -------------------------------------------------------------------------------- 1 | { 2 | "cells": [ 3 | { 4 | "cell_type": "code", 5 | "execution_count": null, 6 | "metadata": {}, 7 | "outputs": [], 8 | "source": [ 9 | "#hide\n", 10 | "from fastdebug.torch import *\n", 11 | "from fastdebug.fastai import *" 12 | ] 13 | }, 14 | { 15 | "cell_type": "markdown", 16 | "metadata": {}, 17 | "source": [ 18 | "# fastdebug\n", 19 | "\n", 20 | "> A helpful library for improving torch and fastai errors" 21 | ] 22 | }, 23 | { 24 | "cell_type": "markdown", 25 | "metadata": {}, 26 | "source": [ 27 | "## Install" 28 | ] 29 | }, 30 | { 31 | "cell_type": "markdown", 32 | "metadata": {}, 33 | "source": [ 34 | "`pip install fastdebug`" 35 | ] 36 | }, 37 | { 38 | "cell_type": "markdown", 39 | "metadata": {}, 40 | "source": [ 41 | "## How to use" 42 | ] 43 | }, 44 | { 45 | "cell_type": "markdown", 46 | "metadata": {}, 47 | "source": [ 48 | "`fastdebug` is designed around improving the quality of life when dealing with Pytorch and fastai errors, while also including some new sanity checks (fastai only)" 49 | ] 50 | }, 51 | { 52 | "cell_type": "markdown", 53 | "metadata": {}, 54 | "source": [ 55 | "### Pytorch\n", 56 | "\n", 57 | "Pytorch now has:\n", 58 | "* `device_error`\n", 59 | "* `layer_error`\n", 60 | "\n", 61 | "Both can be imported with:\n", 62 | "```python\n", 63 | "from fastdebug.error.torch import device_error, layer_error\n", 64 | "```\n", 65 | "\n", 66 | "`device_error` prints out a much more readable error for when two tensors aren't on the same device:" 67 | ] 68 | }, 69 | { 70 | "cell_type": "markdown", 71 | "metadata": {}, 72 | "source": [ 73 | "```python\n", 74 | "inp = torch.rand().cuda()\n", 75 | "model = model.cpu()\n", 76 | "try:\n", 77 | " _ = model(inp)\n", 78 | "except Exception as e:\n", 79 | " device_error(e, 'Input type', 'Model weights')\n", 80 | "```\n", 81 | "And our new log:\n", 82 | "```bash\n", 83 | "---------------------------------------------------------------------------\n", 84 | "RuntimeError Traceback (most recent call last)\n", 85 | " in ()\n", 86 | " 2 model(x)\n", 87 | " 3 except Exception as e:\n", 88 | "----> 4 device_error(e, 'Input type', 'Model weights')\n", 89 | "\n", 90 | "10 frames\n", 91 | "/usr/local/lib/python3.7/dist-packages/torch/tensor.py in __torch_function__(cls, func, types, args, kwargs)\n", 92 | " 993 \n", 93 | " 994 with _C.DisableTorchFunction():\n", 94 | "--> 995 ret = func(*args, **kwargs)\n", 95 | " 996 return _convert(ret, cls)\n", 96 | " 997 \n", 97 | "\n", 98 | "RuntimeError: Mismatch between weight types\n", 99 | "\n", 100 | "Input type has type: \t\t (torch.cuda.FloatTensor)\n", 101 | "Model weights have type: \t (torch.FloatTensor)\n", 102 | "\n", 103 | "Both should be the same.\n", 104 | "```" 105 | ] 106 | }, 107 | { 108 | "cell_type": "markdown", 109 | "metadata": {}, 110 | "source": [ 111 | "And with `layer_error`, if there is a shape mismatch it will attempt to find the right layer it was at:\n", 112 | "```python\n", 113 | "inp = torch.rand(5,2, 3)\n", 114 | "try:\n", 115 | " m(inp)\n", 116 | "except Exception as e:\n", 117 | " layer_error(e, m)\n", 118 | "```\n", 119 | "\n", 120 | "```python\n", 121 | "---------------------------------------------------------------------------\n", 122 | "RuntimeError Traceback (most recent call last)\n", 123 | " in ()\n", 124 | " 3 m(inp)\n", 125 | " 4 except Exception as e:\n", 126 | "----> 5 layer_error(e, m)\n", 127 | "\n", 128 | " in layer_error(e, model)\n", 129 | " 8 i, layer = get_layer_by_shape(model, shape)\n", 130 | " 9 e.args = [f'Size mismatch between input tensors and what the model expects\\n\\n{args}\\n\\tat layer {i}: {layer}']\n", 131 | "---> 10 raise e\n", 132 | "\n", 133 | " in ()\n", 134 | " 1 inp = torch.rand(5,2, 3)\n", 135 | " 2 try:\n", 136 | "----> 3 m(inp)\n", 137 | " 4 except Exception as e:\n", 138 | " 5 layer_error(e, m)\n", 139 | "\n", 140 | "/mnt/d/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)\n", 141 | " 725 result = self._slow_forward(*input, **kwargs)\n", 142 | " 726 else:\n", 143 | "--> 727 result = self.forward(*input, **kwargs)\n", 144 | " 728 for hook in itertools.chain(\n", 145 | " 729 _global_forward_hooks.values(),\n", 146 | "\n", 147 | "/mnt/d/lib/python3.7/site-packages/torch/nn/modules/container.py in forward(self, input)\n", 148 | " 115 def forward(self, input):\n", 149 | " 116 for module in self:\n", 150 | "--> 117 input = module(input)\n", 151 | " 118 return input\n", 152 | " 119 \n", 153 | "\n", 154 | "/mnt/d/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)\n", 155 | " 725 result = self._slow_forward(*input, **kwargs)\n", 156 | " 726 else:\n", 157 | "--> 727 result = self.forward(*input, **kwargs)\n", 158 | " 728 for hook in itertools.chain(\n", 159 | " 729 _global_forward_hooks.values(),\n", 160 | "\n", 161 | "/mnt/d/lib/python3.7/site-packages/torch/nn/modules/conv.py in forward(self, input)\n", 162 | " 421 \n", 163 | " 422 def forward(self, input: Tensor) -> Tensor:\n", 164 | "--> 423 return self._conv_forward(input, self.weight)\n", 165 | " 424 \n", 166 | " 425 class Conv3d(_ConvNd):\n", 167 | "\n", 168 | "/mnt/d/lib/python3.7/site-packages/torch/nn/modules/conv.py in _conv_forward(self, input, weight)\n", 169 | " 418 _pair(0), self.dilation, self.groups)\n", 170 | " 419 return F.conv2d(input, weight, self.bias, self.stride,\n", 171 | "--> 420 self.padding, self.dilation, self.groups)\n", 172 | " 421 \n", 173 | " 422 def forward(self, input: Tensor) -> Tensor:\n", 174 | "\n", 175 | "RuntimeError: Size mismatch between input tensors and what the model expects\n", 176 | "\n", 177 | "Model expected 4-dimensional input for 4-dimensional weight [3, 3, 1, 1], but got 3-dimensional input of size [5, 2, 3] instead\n", 178 | "\tat layer 1: Conv2d(3, 3, kernel_size=(1, 1), stride=(1, 1))\n", 179 | "```" 180 | ] 181 | }, 182 | { 183 | "cell_type": "markdown", 184 | "metadata": {}, 185 | "source": [ 186 | "### fastai\n", 187 | "\n", 188 | "Along with the additions above (and are used during `fit`), fastai now has a `Learner.sanity_check` function, which allows you to quickly perform a basic check to ensure that your call to `fit` won't raise any exceptions. They are performed on the CPU for a partial epoch to make sure that `CUDA` device-assist errors can be preemptively found.\n", 189 | "\n", 190 | "To use it simply do:\n", 191 | "```python\n", 192 | "from fastdebug.fastai import *\n", 193 | "from fastai.vision.all import *\n", 194 | "\n", 195 | "learn = Learner(...)\n", 196 | "learn.sanity_check()\n", 197 | "```\n", 198 | "\n", 199 | "This is also now an argument in `Learner`, set to `False` by default, so that after making your `Learner` a quick check is ensured." 200 | ] 201 | }, 202 | { 203 | "cell_type": "markdown", 204 | "metadata": {}, 205 | "source": [ 206 | "```python\n", 207 | "learn = Learner(..., sanity_check=True)\n", 208 | "```" 209 | ] 210 | } 211 | ], 212 | "metadata": { 213 | "kernelspec": { 214 | "display_name": "Python 3", 215 | "language": "python", 216 | "name": "python3" 217 | } 218 | }, 219 | "nbformat": 4, 220 | "nbformat_minor": 2 221 | } 222 | -------------------------------------------------------------------------------- /settings.ini: -------------------------------------------------------------------------------- 1 | [DEFAULT] 2 | # All sections below are required unless otherwise specified 3 | host = github 4 | lib_name = fastdebug 5 | # For Enterprise Git add variable repo_name and company name 6 | # repo_name = analytics 7 | # company_name = nike 8 | 9 | user = muellerzr 10 | description = A library that improves the debugging messages for Pytorch and fastai 11 | keywords = fastai pytorch debugging 12 | author = Zachary Mueller 13 | author_email = muellerzr@gmail.com 14 | copyright = Zachary Mueller 15 | branch = master 16 | version = 0.1.4 17 | min_python = 3.6 18 | audience = Developers 19 | language = English 20 | # Set to True if you want to create a more fancy sidebar.json than the default 21 | custom_sidebar = False 22 | # Add licenses and see current list in `setup.py` 23 | license = apache2 24 | # From 1-7: Planning Pre-Alpha Alpha Beta Production Mature Inactive 25 | status = 2 26 | 27 | requirements = torch>=1.7.0 fastai>=2.0.0 inflect>=5.3.0 28 | 29 | # Optional. Same format as setuptools console_scripts 30 | # console_scripts = 31 | # Optional. Same format as setuptools dependency-links 32 | # dep_links = 33 | 34 | ### 35 | # You probably won't need to change anything under here, 36 | # unless you have some special requirements 37 | ### 38 | 39 | # Change to, e.g. "nbs", to put your notebooks in nbs dir instead of repo root 40 | nbs_path = . 41 | doc_path = docs 42 | 43 | # Whether to look for library notebooks recursively in the `nbs_path` dir 44 | recursive = False 45 | 46 | # Anything shown as '%(...)s' is substituted with that setting automatically 47 | doc_host = https://%(user)s.github.io 48 | #For Enterprise Git pages use: 49 | #doc_host = https://pages.github.%(company_name)s.com. 50 | 51 | 52 | doc_baseurl = /%(lib_name)s/ 53 | # For Enterprise Github pages docs use: 54 | # doc_baseurl = /%(repo_name)s/%(lib_name)s/ 55 | 56 | git_url = https://github.com/%(user)s/%(lib_name)s/tree/%(branch)s/ 57 | # For Enterprise Github use: 58 | #git_url = https://github.%(company_name)s.com/%(repo_name)s/%(lib_name)s/tree/%(branch)s/ 59 | 60 | 61 | 62 | lib_path = %(lib_name)s 63 | title = %(lib_name)s 64 | 65 | #Optional advanced parameters 66 | #Monospace docstings: adds
 tags around the doc strings, preserving newlines/indentation.
67 | #monospace_docstrings = False
68 | #Test flags: introduce here the test flags you want to use separated by |
69 | tst_flags = slow|failing
70 | #Custom sidebar: customize sidebar.json yourself for advanced sidebars (False/True)
71 | #custom_sidebar = 
72 | #Cell spacing: if you want cell blocks in code separated by more than one new line
73 | #cell_spacing = 
74 | #Custom jekyll styles: if you want more jekyll styles than tip/important/warning, set them here
75 | #jekyll_styles = note,warning,tip,important
76 | 


--------------------------------------------------------------------------------
/setup.py:
--------------------------------------------------------------------------------
 1 | from pkg_resources import parse_version
 2 | from configparser import ConfigParser
 3 | import setuptools,re,sys
 4 | assert parse_version(setuptools.__version__)>=parse_version('36.2')
 5 | 
 6 | # note: all settings are in settings.ini; edit there, not here
 7 | config = ConfigParser(delimiters=['='])
 8 | config.read('settings.ini')
 9 | cfg = config['DEFAULT']
10 | 
11 | cfg_keys = 'version description keywords author author_email'.split()
12 | expected = cfg_keys + "lib_name user branch license status min_python audience language".split()
13 | for o in expected: assert o in cfg, "missing expected setting: {}".format(o)
14 | setup_cfg = {o:cfg[o] for o in cfg_keys}
15 | 
16 | if len(sys.argv)>1 and sys.argv[1]=='version':
17 |     print(setup_cfg['version'])
18 |     exit()
19 | 
20 | licenses = {
21 |     'apache2': ('Apache Software License 2.0','OSI Approved :: Apache Software License'),
22 |     'mit': ('MIT License', 'OSI Approved :: MIT License'),
23 |     'gpl2': ('GNU General Public License v2', 'OSI Approved :: GNU General Public License v2 (GPLv2)'),
24 |     'gpl3': ('GNU General Public License v3', 'OSI Approved :: GNU General Public License v3 (GPLv3)'),
25 |     'bsd3': ('BSD License', 'OSI Approved :: BSD License'),
26 | }
27 | statuses = [ '1 - Planning', '2 - Pre-Alpha', '3 - Alpha',
28 |     '4 - Beta', '5 - Production/Stable', '6 - Mature', '7 - Inactive' ]
29 | py_versions = '2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 3.0 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8'.split()
30 | 
31 | lic = licenses.get(cfg['license'].lower(), (cfg['license'], None))
32 | min_python = cfg['min_python']
33 | 
34 | requirements = ['pip', 'packaging']
35 | if cfg.get('requirements'): requirements += cfg.get('requirements','').split()
36 | if cfg.get('pip_requirements'): requirements += cfg.get('pip_requirements','').split()
37 | dev_requirements = (cfg.get('dev_requirements') or '').split()
38 | 
39 | long_description = open('README.md').read()
40 | # ![png](docs/images/output_13_0.png)
41 | for ext in ['png', 'svg']:
42 |     long_description = re.sub(r'!\['+ext+'\]\((.*)\)', '!['+ext+']('+'https://raw.githubusercontent.com/{}/{}'.format(cfg['user'],cfg['lib_name'])+'/'+cfg['branch']+'/\\1)', long_description)
43 |     long_description = re.sub(r'src=\"(.*)\.'+ext+'\"', 'src=\"https://raw.githubusercontent.com/{}/{}'.format(cfg['user'],cfg['lib_name'])+'/'+cfg['branch']+'/\\1.'+ext+'\"', long_description)
44 | 
45 | setuptools.setup(
46 |     name = cfg['lib_name'],
47 |     license = lic[0],
48 |     classifiers = [
49 |         'Development Status :: ' + statuses[int(cfg['status'])],
50 |         'Intended Audience :: ' + cfg['audience'].title(),
51 |         'Natural Language :: ' + cfg['language'].title(),
52 |     ] + ['Programming Language :: Python :: '+o for o in py_versions[py_versions.index(min_python):]] + (['License :: ' + lic[1] ] if lic[1] else []),
53 |     url = cfg['git_url'],
54 |     packages = setuptools.find_packages(),
55 |     include_package_data = True,
56 |     install_requires = requirements,
57 |     extras_require={ 'dev': dev_requirements },
58 |     python_requires  = '>=' + cfg['min_python'],
59 |     long_description = long_description,
60 |     long_description_content_type = 'text/markdown',
61 |     zip_safe = False,
62 |     entry_points = { 'console_scripts': cfg.get('console_scripts','').split() },
63 |     **setup_cfg)
64 | 
65 | 


--------------------------------------------------------------------------------