p1ch3/2_named_tensors.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "../c10/core/TensorImpl.h:860: UserWarning: Named tensors and all their associated APIs are an experimental feature and subject to change. Please do not use them for anything important until they are released as stable.\n"
     ]
    }
   ],
   "source": [
    "import torch\n",
    "_ = torch.tensor([0.2126, 0.7152, 0.0722], names=['c'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "img_t = torch.randn(3, 5, 5) # shape [channels, rows, columns]\n",
    "weights = torch.tensor([0.2126, 0.7152, 0.0722])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "batch_t = torch.randn(2, 3, 5, 5) # shape [batch, channels, rows, columns]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(torch.Size([5, 5]), torch.Size([2, 5, 5]))"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "img_gray_naive = img_t.mean(-3)\n",
    "batch_gray_naive = batch_t.mean(-3)\n",
    "img_gray_naive.shape, batch_gray_naive.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(torch.Size([2, 3, 5, 5]), torch.Size([2, 3, 5, 5]), torch.Size([3, 1, 1]))"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "unsqueezed_weights = weights.unsqueeze(-1).unsqueeze_(-1)\n",
    "img_weights = (img_t * unsqueezed_weights)\n",
    "batch_weights = (batch_t * unsqueezed_weights)\n",
    "img_gray_weighted = img_weights.sum(-3)\n",
    "batch_gray_weighted = batch_weights.sum(-3)\n",
    "batch_weights.shape, batch_t.shape, unsqueezed_weights.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([2, 5, 5])"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "img_gray_weighted_fancy = torch.einsum('...chw,c->...hw', img_t, weights)\n",
    "batch_gray_weighted_fancy = torch.einsum('...chw,c->...hw', batch_t, weights)\n",
    "batch_gray_weighted_fancy.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([0.2126, 0.7152, 0.0722], names=('channels',))"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "weights_named = torch.tensor([0.2126, 0.7152, 0.0722], names=['channels'])\n",
    "weights_named"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "img named: torch.Size([3, 5, 5]) ('channels', 'rows', 'columns')\n",
      "batch named: torch.Size([2, 3, 5, 5]) (None, 'channels', 'rows', 'columns')\n"
     ]
    }
   ],
   "source": [
    "img_named =  img_t.refine_names(..., 'channels', 'rows', 'columns')\n",
    "batch_named = batch_t.refine_names(..., 'channels', 'rows', 'columns')\n",
    "print(\"img named:\", img_named.shape, img_named.names)\n",
    "print(\"batch named:\", batch_named.shape, batch_named.names)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(torch.Size([3, 1, 1]), ('channels', 'rows', 'columns'))"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "weights_aligned = weights_named.align_as(img_named)\n",
    "weights_aligned.shape, weights_aligned.names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(torch.Size([5, 5]), ('rows', 'columns'))"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gray_named = (img_named * weights_aligned).sum('channels')\n",
    "gray_named.shape, gray_named.names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Error when attempting to broadcast dims ['channels', 'rows', 'columns'] and dims ['channels']: dim 'columns' and dim 'channels' are at the same position from the right but do not match.\n"
     ]
    }
   ],
   "source": [
    "try:\n",
    "    gray_named = (img_named[..., :3] * weights_named).sum('channels')\n",
    "except Exception as e:\n",
    "    print(e)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(torch.Size([5, 5]), (None, None))"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gray_plain = gray_named.rename(None)\n",
    "gray_plain.shape, gray_plain.names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "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": "3.7.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}