{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data augmentation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#export\n",
    "from exp.nb_09 import *"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## PIL transforms"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "path = datasets.untar_data(datasets.URLs.IMAGENETTE_160)\n",
    "tfms = [make_rgb, ResizeFixed(128), to_byte_tensor, to_float_tensor]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from nvidia.dali.pipeline import Pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import nvidia.dali.ops as ops\n",
    "import nvidia.dali.types as types"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "bs=8"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "class SimplePipeline(Pipeline):\n",
    "    def __init__(self, batch_size=8, num_threads=8, device_id=0):\n",
    "        super(SimplePipeline, self).__init__(batch_size, num_threads, device_id, seed = 12)\n",
    "        self.input = ops.FileReader(file_root = path/'train')\n",
    "        # self.input = ops.FileReader(file_root = image_dir, file_list = image_dir + '/file_list.txt')\n",
    "        self.resize = ops.Resize(device = \"cpu\", resize_x=128, resize_y=128, image_type = types.RGB,\n",
    "            interp_type = types.INTERP_LINEAR)\n",
    "        self.decode = ops.HostDecoder(output_type = types.RGB)\n",
    "\n",
    "    def define_graph(self):\n",
    "        jpegs, labels = self.input(name='r')\n",
    "        images = self.decode(jpegs)\n",
    "        images = self.resize(images)\n",
    "        return (images, labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pipe = SimplePipeline()\n",
    "pipe.build()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pipe_out = pipe.run()\n",
    "print(pipe_out)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "images, labels = pipe_out\n",
    "images.is_dense_tensor(), labels.is_dense_tensor()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "t = images.as_tensor()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from nvidia.dali.plugin.pytorch import DALIGenericIterator,DALIClassificationIterator,feed_ndarray"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "it = DALIGenericIterator(pipe, ['data','label'], pipe.epoch_size('r'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "it = DALIClassificationIterator(pipe, pipe.epoch_size('r'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "its = iter(it)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "t = next(it)[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "t['label'].cuda().long().type()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "t['data'].type()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels_tensor = labels.as_tensor()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels_tensor.shape()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.array(labels_tensor)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.gridspec as gridspec"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def show_images(image_batch):\n",
    "    columns = 4\n",
    "    rows = len(image_batch) // (columns)\n",
    "    fig = plt.figure(figsize = (32,(32 // columns) * rows))\n",
    "    gs = gridspec.GridSpec(rows, columns)\n",
    "    for j in range(rows*columns):\n",
    "        plt.subplot(gs[j])\n",
    "        plt.axis(\"off\")\n",
    "        plt.imshow(image_batch.at(j))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "show_images(images)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}