|
| 1 | +{ |
| 2 | + "nbformat": 4, |
| 3 | + "nbformat_minor": 0, |
| 4 | + "metadata": { |
| 5 | + "colab": { |
| 6 | + "name": "inception.ipynb", |
| 7 | + "provenance": [], |
| 8 | + "collapsed_sections": [] |
| 9 | + }, |
| 10 | + "kernelspec": { |
| 11 | + "display_name": "Python 3", |
| 12 | + "language": "python", |
| 13 | + "name": "python3" |
| 14 | + }, |
| 15 | + "language_info": { |
| 16 | + "file_extension": ".py", |
| 17 | + "mimetype": "text/x-python", |
| 18 | + "name": "python", |
| 19 | + "nbconvert_exporter": "python", |
| 20 | + "pygments_lexer": "ipython3", |
| 21 | + "version": "3.6.8" |
| 22 | + } |
| 23 | + }, |
| 24 | + "cells": [ |
| 25 | + { |
| 26 | + "cell_type": "markdown", |
| 27 | + "metadata": { |
| 28 | + "id": "n8CwINQcEBKz", |
| 29 | + "colab_type": "text" |
| 30 | + }, |
| 31 | + "source": [ |
| 32 | + "# Exporting ImageNet Inception\n", |
| 33 | + "In this notebook, we'll show how to export the [pre-trained Imagenet Inception model](https://tfhub.dev/google/imagenet/inception_v3/classification/3) for serving." |
| 34 | + ] |
| 35 | + }, |
| 36 | + { |
| 37 | + "cell_type": "markdown", |
| 38 | + "metadata": { |
| 39 | + "id": "3221z3P69fgf", |
| 40 | + "colab_type": "text" |
| 41 | + }, |
| 42 | + "source": [ |
| 43 | + "First, we'll install the required packages:" |
| 44 | + ] |
| 45 | + }, |
| 46 | + { |
| 47 | + "cell_type": "code", |
| 48 | + "metadata": { |
| 49 | + "id": "_SdQpq7g9LiI", |
| 50 | + "colab_type": "code", |
| 51 | + "colab": {} |
| 52 | + }, |
| 53 | + "source": [ |
| 54 | + "!pip install tensorflow==1.14.* tensorflow-hub==0.6.* boto3==1.*" |
| 55 | + ], |
| 56 | + "execution_count": 0, |
| 57 | + "outputs": [] |
| 58 | + }, |
| 59 | + { |
| 60 | + "cell_type": "markdown", |
| 61 | + "metadata": { |
| 62 | + "id": "I-k0gUpxDGkU", |
| 63 | + "colab_type": "text" |
| 64 | + }, |
| 65 | + "source": [ |
| 66 | + "Next, we'll download the model from TensorFlow Hub and export it for serving:" |
| 67 | + ] |
| 68 | + }, |
| 69 | + { |
| 70 | + "cell_type": "code", |
| 71 | + "metadata": { |
| 72 | + "id": "z6QLCzB4BKMe", |
| 73 | + "colab_type": "code", |
| 74 | + "colab": {} |
| 75 | + }, |
| 76 | + "source": [ |
| 77 | + "import tensorflow as tf\n", |
| 78 | + "import tensorflow_hub as hub\n", |
| 79 | + "from tensorflow.python.saved_model.signature_def_utils_impl import predict_signature_def\n", |
| 80 | + "\n", |
| 81 | + "export_dir = \"export/1\"\n", |
| 82 | + "builder = tf.saved_model.builder.SavedModelBuilder(export_dir)\n", |
| 83 | + "\n", |
| 84 | + "with tf.Session(graph=tf.Graph()) as sess:\n", |
| 85 | + " module = hub.Module(\"https://tfhub.dev/google/imagenet/inception_v3/classification/3\")\n", |
| 86 | + "\n", |
| 87 | + " input_params = module.get_input_info_dict()\n", |
| 88 | + " image_input = tf.placeholder(\n", |
| 89 | + " name=\"images\", dtype=input_params[\"images\"].dtype, shape=input_params[\"images\"].get_shape()\n", |
| 90 | + " )\n", |
| 91 | + " \n", |
| 92 | + " sess.run([tf.global_variables_initializer(), tf.tables_initializer()])\n", |
| 93 | + "\n", |
| 94 | + " classes = module(image_input)\n", |
| 95 | + " signature = predict_signature_def(inputs={\"images\": image_input}, outputs={\"classes\": classes})\n", |
| 96 | + "\n", |
| 97 | + " builder.add_meta_graph_and_variables(\n", |
| 98 | + " sess, [\"serve\"], signature_def_map={\"predict\": signature}, strip_default_attrs=True\n", |
| 99 | + " )\n", |
| 100 | + "\n", |
| 101 | + "builder.save()" |
| 102 | + ], |
| 103 | + "execution_count": 0, |
| 104 | + "outputs": [] |
| 105 | + }, |
| 106 | + { |
| 107 | + "cell_type": "markdown", |
| 108 | + "metadata": { |
| 109 | + "id": "aGtJiyEnBgwl", |
| 110 | + "colab_type": "text" |
| 111 | + }, |
| 112 | + "source": [ |
| 113 | + "## Upload the model to AWS\n", |
| 114 | + "\n", |
| 115 | + "Cortex loads models from AWS, so we need to upload the exported model." |
| 116 | + ] |
| 117 | + }, |
| 118 | + { |
| 119 | + "cell_type": "markdown", |
| 120 | + "metadata": { |
| 121 | + "id": "fTkjvSKBBmUB", |
| 122 | + "colab_type": "text" |
| 123 | + }, |
| 124 | + "source": [ |
| 125 | + "Set these variables to configure your AWS credentials and model upload path:" |
| 126 | + ] |
| 127 | + }, |
| 128 | + { |
| 129 | + "cell_type": "code", |
| 130 | + "metadata": { |
| 131 | + "id": "4xcDWxqCBPre", |
| 132 | + "colab_type": "code", |
| 133 | + "colab": {}, |
| 134 | + "cellView": "form" |
| 135 | + }, |
| 136 | + "source": [ |
| 137 | + "AWS_ACCESS_KEY_ID = \"\" #@param {type:\"string\"}\n", |
| 138 | + "AWS_SECRET_ACCESS_KEY = \"\" #@param {type:\"string\"}\n", |
| 139 | + "S3_UPLOAD_PATH = \"s3://my-bucket/inception\" #@param {type:\"string\"}\n", |
| 140 | + "\n", |
| 141 | + "import sys\n", |
| 142 | + "import re\n", |
| 143 | + "\n", |
| 144 | + "if AWS_ACCESS_KEY_ID == \"\":\n", |
| 145 | + " print(\"\\033[91m{}\\033[00m\".format(\"ERROR: Please set AWS_ACCESS_KEY_ID\"), file=sys.stderr)\n", |
| 146 | + "\n", |
| 147 | + "elif AWS_SECRET_ACCESS_KEY == \"\":\n", |
| 148 | + " print(\"\\033[91m{}\\033[00m\".format(\"ERROR: Please set AWS_SECRET_ACCESS_KEY\"), file=sys.stderr)\n", |
| 149 | + "\n", |
| 150 | + "else:\n", |
| 151 | + " try:\n", |
| 152 | + " bucket = re.search(\"s3://(.+?)/\", S3_UPLOAD_PATH).group(1)\n", |
| 153 | + " key = re.search(\"s3://.+?/(.+)\", S3_UPLOAD_PATH).group(1)\n", |
| 154 | + " except:\n", |
| 155 | + " print(\"\\033[91m{}\\033[00m\".format(\"ERROR: Invalid s3 path (should be of the form s3://my-bucket/path/to/file)\"), file=sys.stderr)" |
| 156 | + ], |
| 157 | + "execution_count": 0, |
| 158 | + "outputs": [] |
| 159 | + }, |
| 160 | + { |
| 161 | + "cell_type": "markdown", |
| 162 | + "metadata": { |
| 163 | + "id": "czZkjb1IBr-f", |
| 164 | + "colab_type": "text" |
| 165 | + }, |
| 166 | + "source": [ |
| 167 | + "Upload the model to S3:" |
| 168 | + ] |
| 169 | + }, |
| 170 | + { |
| 171 | + "cell_type": "code", |
| 172 | + "metadata": { |
| 173 | + "id": "M0b0IbyaBsim", |
| 174 | + "colab_type": "code", |
| 175 | + "colab": {} |
| 176 | + }, |
| 177 | + "source": [ |
| 178 | + "import os\n", |
| 179 | + "import boto3\n", |
| 180 | + "\n", |
| 181 | + "s3 = boto3.client(\"s3\", aws_access_key_id=AWS_ACCESS_KEY_ID, aws_secret_access_key=AWS_SECRET_ACCESS_KEY)\n", |
| 182 | + "\n", |
| 183 | + "for dirpath, _, filenames in os.walk(\"export\"):\n", |
| 184 | + " for filename in filenames:\n", |
| 185 | + " filepath = os.path.join(dirpath, filename)\n", |
| 186 | + " filekey = os.path.join(key, filepath[len(\"export/\"):])\n", |
| 187 | + " print(\"Uploading s3://{}/{}...\".format(bucket, filekey), end = '')\n", |
| 188 | + " s3.upload_file(filepath, bucket, filekey)\n", |
| 189 | + " print(\" ✓\")", |
| 190 | + "\n", |
| 191 | + "print(\"\\nUploaded model export directory to \" + S3_UPLOAD_PATH)" |
| 192 | + ], |
| 193 | + "execution_count": 0, |
| 194 | + "outputs": [] |
| 195 | + }, |
| 196 | + { |
| 197 | + "cell_type": "markdown", |
| 198 | + "metadata": { |
| 199 | + "id": "pZQWoeZbE7Wc", |
| 200 | + "colab_type": "text" |
| 201 | + }, |
| 202 | + "source": [ |
| 203 | + "<!-- CORTEX_VERSION_MINOR -->\n", |
| 204 | + "That's it! See the [example on GitHub](https://github.com/cortexlabs/cortex/tree/master/examples/image-classifier) for how to deploy the model as an API." |
| 205 | + ] |
| 206 | + } |
| 207 | + ] |
| 208 | +} |
0 commit comments