Pooling notebook for checking einops channel pooling correctness

Pooling.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "af0d71c1",
   "metadata": {},
   "outputs": [],
   "source": [
    "from torch.nn import MaxPool1d\n",
    "import torch.nn.functional as F\n",
    "\n",
    "\n",
    "class ChannelPool(MaxPool1d):\n",
    "    def forward(self, input):\n",
    "        n, c, w, h = input.size()\n",
    "        input = input.view(n, c, w * h).permute(0, 2, 1)\n",
    "        pooled = F.max_pool1d(\n",
    "            input,\n",
    "            self.kernel_size,\n",
    "            self.stride,\n",
    "            self.padding,\n",
    "            self.dilation,\n",
    "            self.ceil_mode,\n",
    "            self.return_indices,\n",
    "        )\n",
    "        _, _, c = pooled.size()\n",
    "        pooled = pooled.permute(0, 2, 1)\n",
    "        return pooled.view(n, c, w, h)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "1ca5fd19",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "ff163bfd",
   "metadata": {},
   "outputs": [],
   "source": [
    "pool = ChannelPool(2)\n",
    "x = torch.randn(2, 4, 16, 16)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "7d2123e5",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([2, 2, 16, 16])"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y = pool(x)\n",
    "y.size()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "5bb44b47",
   "metadata": {},
   "outputs": [],
   "source": [
    "from torch.nn import MaxPool1d\n",
    "import torch.nn.functional as F\n",
    "from einops import rearrange\n",
    "\n",
    "\n",
    "class ChannelPool(MaxPool1d):\n",
    "    def forward(self, input):\n",
    "        n, c, w, h = input.size()\n",
    "        pool = lambda x: F.max_pool1d(\n",
    "            x,\n",
    "            self.kernel_size,\n",
    "            self.stride,\n",
    "            self.padding,\n",
    "            self.dilation,\n",
    "            self.ceil_mode,\n",
    "            self.return_indices,\n",
    "        )\n",
    "        return rearrange(\n",
    "            pool(rearrange(input, \"n c w h -> n (w h) c\")),\n",
    "            \"n (w h) c -> n c w h\",\n",
    "            n=n,\n",
    "            w=w,\n",
    "            h=h,\n",
    "        )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "b34dc47e",
   "metadata": {},
   "outputs": [],
   "source": [
    "pool = ChannelPool(2)\n",
    "_y = pool(x)\n",
    "assert torch.abs(y - _y).max() < 1e-5"
   ]
  }
 ],
 "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": "3.8.10"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}