pytorch hook.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 264,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "from loguru import logger\n",
    "\n",
    "class Add(nn.Module):\n",
    "    def __init__(self):\n",
    "        super().__init__()\n",
    "        \n",
    "    def forward(self, x1):\n",
    "        output = torch.where(x1==1., torch.nan, x1)\n",
    "        output = nn.ReLU()(output)\n",
    "        return output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 265,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[ 1.,  2.,  3.,  4.,  5.],\n",
       "        [ 6.,  7.,  8.,  9., 10.]])"
      ]
     },
     "execution_count": 265,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "input = torch.arange(1, 11).reshape(2, 5).to(torch.float32)\n",
    "input"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 266,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[ 1.,  2.,  3.,  4.,  5.],\n",
       "        [ 6.,  7.,  8., inf, 10.]])"
      ]
     },
     "execution_count": 266,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "input[1, 3] = float('inf')\n",
    "input"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 267,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Add()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 268,
   "metadata": {},
   "outputs": [],
   "source": [
    "def nan_hook(module, inp, output):\n",
    "    if not isinstance(output, tuple):\n",
    "        outputs = [output]\n",
    "    else:\n",
    "        outputs = output\n",
    "\n",
    "    for i, out in enumerate(outputs):\n",
    "        nan_mask = torch.isnan(out)\n",
    "        if nan_mask.any():\n",
    "            logger.debug(f\"Found NAN in output {i} at indices: {nan_mask.nonzero()} where: {out[nan_mask.nonzero()[:, 0].unique(sorted=True)]}\")\n",
    "\n",
    "def inf_hook(module, inp, output):\n",
    "    if not isinstance(output, tuple):\n",
    "        outputs = [output]\n",
    "    else:\n",
    "        outputs = output\n",
    "\n",
    "    for i, out in enumerate(outputs):\n",
    "        inf_mask = torch.isinf(out)\n",
    "        if inf_mask.any():\n",
    "            logger.debug(f\"Found INF in output {i} at indices: {inf_mask.nonzero()} where: {out[inf_mask.nonzero()[:, 0].unique(sorted=True)]}\")\n",
    "\n",
    "for submodule in model.modules():\n",
    "    submodule.register_forward_hook(inf_hook)\n",
    "    submodule.register_forward_hook(nan_hook)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 269,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[32m2023-11-08 10:31:54.312\u001b[0m | \u001b[34m\u001b[1mDEBUG   \u001b[0m | \u001b[36m__main__\u001b[0m:\u001b[36minf_hook\u001b[0m:\u001b[36m21\u001b[0m - \u001b[34m\u001b[1mFound INF in output 0 at indices: tensor([[1, 3]]) where: tensor([[ 6.,  7.,  8., inf, 10.]])\u001b[0m\n",
      "\u001b[32m2023-11-08 10:31:54.314\u001b[0m | \u001b[34m\u001b[1mDEBUG   \u001b[0m | \u001b[36m__main__\u001b[0m:\u001b[36mnan_hook\u001b[0m:\u001b[36m10\u001b[0m - \u001b[34m\u001b[1mFound NAN in output 0 at indices: tensor([[0, 0]]) where: tensor([[nan, 2., 3., 4., 5.]])\u001b[0m\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "tensor([[nan,  2.,  3.,  4.,  5.],\n",
       "        [ 6.,  7.,  8., inf, 10.]])"
      ]
     },
     "execution_count": 269,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model(input)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}