L2_linear_I_quantize_dequantize_tensor.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "You can download the `requirements.txt` for this course from the workspace of this lab. `File --> Open...`"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# L2-A - Linear Quantization I: Quantize and De-quantize a Tensor\n",
    "\n",
    "In this lesson, you will learn the fundamentals of linear quantization."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The libraries are already installed in the classroom.  If you're running this notebook on your own machine, you can install the following:\n",
    "\n",
    "```Python\n",
    "!pip install torch==2.1.1\n",
    "```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "height": 30
   },
   "outputs": [],
   "source": [
    "import torch"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Quantization with Random `Scale` and `Zero Point`\n",
    "\n",
    "- Implement Linear Quantization for when the \"scale\" and the \"zero point\" are known/randomly selected."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "height": 234
   },
   "outputs": [],
   "source": [
    "def linear_q_with_scale_and_zero_point(\n",
    "    tensor, scale, zero_point, dtype = torch.int8):\n",
    "\n",
    "    scaled_and_shifted_tensor = tensor / scale + zero_point\n",
    "\n",
    "    rounded_tensor = torch.round(scaled_and_shifted_tensor)\n",
    "\n",
    "    q_min = torch.iinfo(dtype).min\n",
    "    q_max = torch.iinfo(dtype).max\n",
    "\n",
    "    q_tensor = rounded_tensor.clamp(q_min,q_max).to(dtype)\n",
    "    \n",
    "    return q_tensor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "height": 115,
    "id": "Oz0fFY6gMXI6"
   },
   "outputs": [],
   "source": [
    "### a dummy tensor to test the implementation\n",
    "test_tensor=torch.tensor(\n",
    "    [[191.6, -13.5, 728.6],\n",
    "     [92.14, 295.5,  -184],\n",
    "     [0,     684.6, 245.5]]\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "height": 81
   },
   "outputs": [],
   "source": [
    "### these are random values for \"scale\" and \"zero_point\"\n",
    "### to test the implementation\n",
    "scale = 3.5\n",
    "zero_point = -70"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "height": 47
   },
   "outputs": [],
   "source": [
    "quantized_tensor = linear_q_with_scale_and_zero_point(\n",
    "    test_tensor, scale, zero_point)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "height": 30
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[ -15,  -74,  127],\n",
       "        [ -44,   14, -123],\n",
       "        [ -70,  126,    0]], dtype=torch.int8)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "quantized_tensor"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Dequantization with Random `Scale` and `Zero Point`"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- Now, Dequantize the tensor to see how precise the quantization is."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "height": 30
   },
   "outputs": [],
   "source": [
    "dequantized_tensor = scale * (quantized_tensor.float() - zero_point)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "height": 81
   },
   "outputs": [],
   "source": [
    "# this was the original tensor\n",
    "# [[191.6, -13.5, 728.6],\n",
    "#  [92.14, 295.5,  -184],\n",
    "#  [0,     684.6, 245.5]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "height": 30
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[ 192.5000,  -14.0000,  689.5000],\n",
       "        [  91.0000,  294.0000, -185.5000],\n",
       "        [   0.0000,  686.0000,  245.0000]])"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dequantized_tensor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "height": 47
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[ 192.5000,  -14.0000, -206.5000],\n",
       "        [  91.0000,  294.0000, -185.5000],\n",
       "        [   0.0000, -210.0000,  245.0000]])"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "### without casting to float\n",
    "scale * (quantized_tensor - zero_point)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "height": 47
   },
   "outputs": [],
   "source": [
    "def linear_dequantization(quantized_tensor, scale, zero_point):\n",
    "    return scale * (quantized_tensor.float() - zero_point)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- Calculate `dequantized_tensor` using the function `linear_dequantization`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "height": 47
   },
   "outputs": [],
   "source": [
    "dequantized_tensor = linear_dequantization(\n",
    "    quantized_tensor, scale, zero_point)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- Print the results of the `dequantized_tensor`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "height": 30
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[ 192.5000,  -14.0000,  689.5000],\n",
       "        [  91.0000,  294.0000, -185.5000],\n",
       "        [   0.0000,  686.0000,  245.0000]])"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dequantized_tensor"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Quantization Error"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- Load the `plot_quantization_errors` from the helper file.\n",
    "- To access the `helper.py` file, you can click `File --> Open...`, on the top left."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "height": 30
   },
   "outputs": [],
   "source": [
    "from helper import plot_quantization_errors"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- Plot the quantization results."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 425
    },
    "height": 47,
    "id": "5Jr5S5ZBNhkj",
    "outputId": "2edb875f-db33-48ea-aec5-1bdfb7e1f244"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1500x400 with 4 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_quantization_errors(test_tensor, quantized_tensor,\n",
    "                         dequantized_tensor)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Note:** For the plot above, `Quantization Error Tensor = abs(Original Tensor - Dequantized Tensor)`"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- Calculate an \"overall\" quantization error by using [Mean Squared Error](https://en.wikipedia.org/wiki/Mean_squared_error) technique."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "height": 30
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[  0.9000,  -0.5000, -39.1000],\n",
       "        [ -1.1400,  -1.5000,  -1.5000],\n",
       "        [  0.0000,   1.4000,  -0.5000]])"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dequantized_tensor - test_tensor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "height": 30
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[8.0999e-01, 2.5000e-01, 1.5288e+03],\n",
       "        [1.2996e+00, 2.2500e+00, 2.2500e+00],\n",
       "        [0.0000e+00, 1.9601e+00, 2.5000e-01]])"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(dequantized_tensor - test_tensor).square()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "height": 30
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor(170.8753)"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(dequantized_tensor - test_tensor).square().mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "height": 30
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "height": 30
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [
    "kd9Q2SD0MJGw",
    "E1palk4uRQX9",
    "oAexFpXiX1PW",
    "ChnEqFPYMn3p",
    "MFx2m7RmzRd5",
    "LbPjb9OOi0Xp",
    "6WopWDYWQr7X",
    "LYfTqh_VMTzT",
    "4YZP-9XTNkur",
    "dpKLMdCYvT_W",
    "2hoC5tcJznoI",
    "S68UGldKRnJc",
    "Y4Qfalu9vtHv",
    "WJN9IfVLTFNd",
    "qC0X9ux6JEmi",
    "AkwpMs-C5ccj",
    "EDpd5Te632KY",
    "JA1-rcLz4t4D",
    "kROAEGfdDsau",
    "oo4BCLpsDw3t",
    "h2gK-eALFc8U",
    "yDin7Rm6Dzqu",
    "X6J9ZiyHWzHa",
    "qUw1gQUu5yIe",
    "vGll7vBT6BGI",
    "Cl3AUuDuAH5w",
    "8NS1TnQt6E6v",
    "vcRy85lACotg"
   ],
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.11.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}