Add files using upload-large-folder tool
Browse files
Builder Script/builder.script.trainner.ipynb
ADDED
|
@@ -0,0 +1,565 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"cells": [
|
| 3 |
+
{
|
| 4 |
+
"cell_type": "code",
|
| 5 |
+
"execution_count": null,
|
| 6 |
+
"id": "97b4efc3-1879-4441-af52-de470fbc3ae8",
|
| 7 |
+
"metadata": {},
|
| 8 |
+
"outputs": [],
|
| 9 |
+
"source": [
|
| 10 |
+
"!pip install -q evaluate datasets accelerate\n",
|
| 11 |
+
"!pip install -q transformers\n",
|
| 12 |
+
"!pip install -q huggingface_hub"
|
| 13 |
+
]
|
| 14 |
+
},
|
| 15 |
+
{
|
| 16 |
+
"cell_type": "code",
|
| 17 |
+
"execution_count": null,
|
| 18 |
+
"id": "ae923886-86f3-431d-b701-1200110b429c",
|
| 19 |
+
"metadata": {},
|
| 20 |
+
"outputs": [],
|
| 21 |
+
"source": [
|
| 22 |
+
"!pip install -q imbalanced-learn\n",
|
| 23 |
+
"#Skip the installation if your runtime is in Google Colab notebooks."
|
| 24 |
+
]
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"cell_type": "code",
|
| 28 |
+
"execution_count": null,
|
| 29 |
+
"id": "126923c7-d53f-42d8-8f06-2ea05609ab0e",
|
| 30 |
+
"metadata": {},
|
| 31 |
+
"outputs": [],
|
| 32 |
+
"source": [
|
| 33 |
+
"!pip install -q numpy\n",
|
| 34 |
+
"#Skip the installation if your runtime is in Google Colab notebooks."
|
| 35 |
+
]
|
| 36 |
+
},
|
| 37 |
+
{
|
| 38 |
+
"cell_type": "code",
|
| 39 |
+
"execution_count": null,
|
| 40 |
+
"id": "9e628805-b90b-4b98-ae97-9f8a8142767f",
|
| 41 |
+
"metadata": {},
|
| 42 |
+
"outputs": [],
|
| 43 |
+
"source": [
|
| 44 |
+
"!pip install -q pillow==11.0.0\n",
|
| 45 |
+
"#Skip the installation if your runtime is in Google Colab notebooks."
|
| 46 |
+
]
|
| 47 |
+
},
|
| 48 |
+
{
|
| 49 |
+
"cell_type": "code",
|
| 50 |
+
"execution_count": null,
|
| 51 |
+
"id": "b58fab4c-211f-4b7b-b7c4-dd76e20c1beb",
|
| 52 |
+
"metadata": {},
|
| 53 |
+
"outputs": [],
|
| 54 |
+
"source": [
|
| 55 |
+
"!pip install -q torchvision \n",
|
| 56 |
+
"#Skip the installation if your runtime is in Google Colab notebooks."
|
| 57 |
+
]
|
| 58 |
+
},
|
| 59 |
+
{
|
| 60 |
+
"cell_type": "code",
|
| 61 |
+
"execution_count": null,
|
| 62 |
+
"id": "d7454ffa-885e-44ba-8259-d8c45f8ec72b",
|
| 63 |
+
"metadata": {},
|
| 64 |
+
"outputs": [],
|
| 65 |
+
"source": [
|
| 66 |
+
"!pip install -q matplotlib\n",
|
| 67 |
+
"!pip install -q scikit-learn\n",
|
| 68 |
+
"#Skip the installation if your runtime is in Google Colab notebooks."
|
| 69 |
+
]
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"cell_type": "code",
|
| 73 |
+
"execution_count": null,
|
| 74 |
+
"id": "4987ed31-c012-434b-9ea7-78da17061d5d",
|
| 75 |
+
"metadata": {},
|
| 76 |
+
"outputs": [],
|
| 77 |
+
"source": [
|
| 78 |
+
"import warnings\n",
|
| 79 |
+
"warnings.filterwarnings(\"ignore\")\n",
|
| 80 |
+
"\n",
|
| 81 |
+
"import gc\n",
|
| 82 |
+
"import numpy as np\n",
|
| 83 |
+
"import pandas as pd\n",
|
| 84 |
+
"import itertools\n",
|
| 85 |
+
"from collections import Counter\n",
|
| 86 |
+
"import matplotlib.pyplot as plt\n",
|
| 87 |
+
"from sklearn.metrics import accuracy_score, roc_auc_score, confusion_matrix, classification_report, f1_score\n",
|
| 88 |
+
"from imblearn.over_sampling import RandomOverSampler\n",
|
| 89 |
+
"import evaluate\n",
|
| 90 |
+
"from datasets import Dataset, Image, ClassLabel\n",
|
| 91 |
+
"from transformers import (\n",
|
| 92 |
+
" TrainingArguments,\n",
|
| 93 |
+
" Trainer,\n",
|
| 94 |
+
" ViTImageProcessor,\n",
|
| 95 |
+
" ViTForImageClassification,\n",
|
| 96 |
+
" DefaultDataCollator\n",
|
| 97 |
+
")\n",
|
| 98 |
+
"import torch\n",
|
| 99 |
+
"from torch.utils.data import DataLoader\n",
|
| 100 |
+
"from torchvision.transforms import (\n",
|
| 101 |
+
" CenterCrop,\n",
|
| 102 |
+
" Compose,\n",
|
| 103 |
+
" Normalize,\n",
|
| 104 |
+
" RandomRotation,\n",
|
| 105 |
+
" RandomResizedCrop,\n",
|
| 106 |
+
" RandomHorizontalFlip,\n",
|
| 107 |
+
" RandomAdjustSharpness,\n",
|
| 108 |
+
" Resize,\n",
|
| 109 |
+
" ToTensor\n",
|
| 110 |
+
")\n",
|
| 111 |
+
"\n",
|
| 112 |
+
"#.......................................................................\n",
|
| 113 |
+
"\n",
|
| 114 |
+
"#Retain this part if you're working outside Google Colab notebooks.\n",
|
| 115 |
+
"from PIL import Image, ExifTags\n",
|
| 116 |
+
"\n",
|
| 117 |
+
"#.......................................................................\n",
|
| 118 |
+
"\n",
|
| 119 |
+
"from PIL import Image as PILImage\n",
|
| 120 |
+
"from PIL import ImageFile\n",
|
| 121 |
+
"# Enable loading truncated images\n",
|
| 122 |
+
"ImageFile.LOAD_TRUNCATED_IMAGES = True"
|
| 123 |
+
]
|
| 124 |
+
},
|
| 125 |
+
{
|
| 126 |
+
"cell_type": "code",
|
| 127 |
+
"execution_count": null,
|
| 128 |
+
"id": "236bc802-54ba-44d1-b35b-62f548832935",
|
| 129 |
+
"metadata": {},
|
| 130 |
+
"outputs": [],
|
| 131 |
+
"source": [
|
| 132 |
+
"from datasets import load_dataset\n",
|
| 133 |
+
"dataset = load_dataset(\"--your--dataset--goes--here--\", split=\"train\")"
|
| 134 |
+
]
|
| 135 |
+
},
|
| 136 |
+
{
|
| 137 |
+
"cell_type": "code",
|
| 138 |
+
"execution_count": null,
|
| 139 |
+
"id": "d57e17cc-72b2-4fde-9855-751cf3440624",
|
| 140 |
+
"metadata": {},
|
| 141 |
+
"outputs": [],
|
| 142 |
+
"source": [
|
| 143 |
+
"from pathlib import Path\n",
|
| 144 |
+
"\n",
|
| 145 |
+
"file_names = []\n",
|
| 146 |
+
"labels = []\n",
|
| 147 |
+
"\n",
|
| 148 |
+
"for example in dataset:\n",
|
| 149 |
+
" file_path = str(example['image']) \n",
|
| 150 |
+
" label = example['label'] \n",
|
| 151 |
+
"\n",
|
| 152 |
+
" file_names.append(file_path) \n",
|
| 153 |
+
" labels.append(label) \n",
|
| 154 |
+
"\n",
|
| 155 |
+
"print(len(file_names), len(labels))"
|
| 156 |
+
]
|
| 157 |
+
},
|
| 158 |
+
{
|
| 159 |
+
"cell_type": "code",
|
| 160 |
+
"execution_count": null,
|
| 161 |
+
"id": "e52c85d2-a245-47c5-9403-5a9cf4e4269d",
|
| 162 |
+
"metadata": {},
|
| 163 |
+
"outputs": [],
|
| 164 |
+
"source": [
|
| 165 |
+
"df = pd.DataFrame.from_dict({\"image\": file_names, \"label\": labels})\n",
|
| 166 |
+
"print(df.shape)"
|
| 167 |
+
]
|
| 168 |
+
},
|
| 169 |
+
{
|
| 170 |
+
"cell_type": "code",
|
| 171 |
+
"execution_count": null,
|
| 172 |
+
"id": "beba86dd-0605-4ebf-8ebb-97d6ad9e5edd",
|
| 173 |
+
"metadata": {},
|
| 174 |
+
"outputs": [],
|
| 175 |
+
"source": [
|
| 176 |
+
"df.head()\n",
|
| 177 |
+
"df['label'].unique()"
|
| 178 |
+
]
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"cell_type": "code",
|
| 182 |
+
"execution_count": null,
|
| 183 |
+
"id": "6defc1e9-4f46-49b6-addc-f422c38fe7e8",
|
| 184 |
+
"metadata": {},
|
| 185 |
+
"outputs": [],
|
| 186 |
+
"source": [
|
| 187 |
+
"y = df[['label']]\n",
|
| 188 |
+
"df = df.drop(['label'], axis=1)\n",
|
| 189 |
+
"ros = RandomOverSampler(random_state=83)\n",
|
| 190 |
+
"df, y_resampled = ros.fit_resample(df, y)\n",
|
| 191 |
+
"del y\n",
|
| 192 |
+
"df['label'] = y_resampled\n",
|
| 193 |
+
"del y_resampled\n",
|
| 194 |
+
"gc.collect()"
|
| 195 |
+
]
|
| 196 |
+
},
|
| 197 |
+
{
|
| 198 |
+
"cell_type": "code",
|
| 199 |
+
"execution_count": null,
|
| 200 |
+
"id": "129d278c-3899-49d2-b06f-a0b2f22f4c4e",
|
| 201 |
+
"metadata": {},
|
| 202 |
+
"outputs": [],
|
| 203 |
+
"source": [
|
| 204 |
+
"dataset[0][\"image\"]\n",
|
| 205 |
+
"dataset[99][\"image\"]"
|
| 206 |
+
]
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"cell_type": "code",
|
| 210 |
+
"execution_count": null,
|
| 211 |
+
"id": "bffc8755-c4ac-41be-b8ab-f9a6e0dbcca3",
|
| 212 |
+
"metadata": {},
|
| 213 |
+
"outputs": [],
|
| 214 |
+
"source": [
|
| 215 |
+
"labels_subset = labels[:5]\n",
|
| 216 |
+
"print(labels_subset)"
|
| 217 |
+
]
|
| 218 |
+
},
|
| 219 |
+
{
|
| 220 |
+
"cell_type": "code",
|
| 221 |
+
"execution_count": null,
|
| 222 |
+
"id": "d003f439-09d1-41e6-9f34-213c4ee38593",
|
| 223 |
+
"metadata": {},
|
| 224 |
+
"outputs": [],
|
| 225 |
+
"source": [
|
| 226 |
+
"labels_list = ['Issue In Deepfake', 'High Quality Deepfake']\n",
|
| 227 |
+
"\n",
|
| 228 |
+
"label2id, id2label = {}, {}\n",
|
| 229 |
+
"for i, label in enumerate(labels_list):\n",
|
| 230 |
+
" label2id[label] = i\n",
|
| 231 |
+
" id2label[i] = label\n",
|
| 232 |
+
"\n",
|
| 233 |
+
"ClassLabels = ClassLabel(num_classes=len(labels_list), names=labels_list)\n",
|
| 234 |
+
"\n",
|
| 235 |
+
"print(\"Mapping of IDs to Labels:\", id2label, '\\n')\n",
|
| 236 |
+
"print(\"Mapping of Labels to IDs:\", label2id)"
|
| 237 |
+
]
|
| 238 |
+
},
|
| 239 |
+
{
|
| 240 |
+
"cell_type": "code",
|
| 241 |
+
"execution_count": null,
|
| 242 |
+
"id": "2fbf1f1b-5936-48be-bc99-6897fea94794",
|
| 243 |
+
"metadata": {},
|
| 244 |
+
"outputs": [],
|
| 245 |
+
"source": [
|
| 246 |
+
"def map_label2id(example):\n",
|
| 247 |
+
" example['label'] = ClassLabels.str2int(example['label'])\n",
|
| 248 |
+
" return example\n",
|
| 249 |
+
"\n",
|
| 250 |
+
"dataset = dataset.map(map_label2id, batched=True)\n",
|
| 251 |
+
"\n",
|
| 252 |
+
"dataset = dataset.cast_column('label', ClassLabels)\n",
|
| 253 |
+
"\n",
|
| 254 |
+
"dataset = dataset.train_test_split(test_size=0.4, shuffle=True, stratify_by_column=\"label\")\n",
|
| 255 |
+
"\n",
|
| 256 |
+
"train_data = dataset['train']\n",
|
| 257 |
+
"\n",
|
| 258 |
+
"test_data = dataset['test']"
|
| 259 |
+
]
|
| 260 |
+
},
|
| 261 |
+
{
|
| 262 |
+
"cell_type": "code",
|
| 263 |
+
"execution_count": null,
|
| 264 |
+
"id": "d8a4f7ca-4dff-4446-acaf-f3e7630b678d",
|
| 265 |
+
"metadata": {},
|
| 266 |
+
"outputs": [],
|
| 267 |
+
"source": [
|
| 268 |
+
"model_str = \"google/vit-base-patch16-224-in21k\"\n",
|
| 269 |
+
"processor = ViTImageProcessor.from_pretrained(model_str)\n",
|
| 270 |
+
"\n",
|
| 271 |
+
"image_mean, image_std = processor.image_mean, processor.image_std\n",
|
| 272 |
+
"size = processor.size[\"height\"]\n",
|
| 273 |
+
"\n",
|
| 274 |
+
"_train_transforms = Compose(\n",
|
| 275 |
+
" [\n",
|
| 276 |
+
" Resize((size, size)),\n",
|
| 277 |
+
" RandomRotation(90),\n",
|
| 278 |
+
" RandomAdjustSharpness(2),\n",
|
| 279 |
+
" ToTensor(),\n",
|
| 280 |
+
" Normalize(mean=image_mean, std=image_std)\n",
|
| 281 |
+
" ]\n",
|
| 282 |
+
")\n",
|
| 283 |
+
"\n",
|
| 284 |
+
"_val_transforms = Compose(\n",
|
| 285 |
+
" [\n",
|
| 286 |
+
" Resize((size, size)),\n",
|
| 287 |
+
" ToTensor(),\n",
|
| 288 |
+
" Normalize(mean=image_mean, std=image_std)\n",
|
| 289 |
+
" ]\n",
|
| 290 |
+
")\n",
|
| 291 |
+
"\n",
|
| 292 |
+
"def train_transforms(examples):\n",
|
| 293 |
+
" examples['pixel_values'] = [_train_transforms(image.convert(\"RGB\")) for image in examples['image']]\n",
|
| 294 |
+
" return examples\n",
|
| 295 |
+
"\n",
|
| 296 |
+
"def val_transforms(examples):\n",
|
| 297 |
+
" examples['pixel_values'] = [_val_transforms(image.convert(\"RGB\")) for image in examples['image']]\n",
|
| 298 |
+
" return examples\n",
|
| 299 |
+
"\n",
|
| 300 |
+
"train_data.set_transform(train_transforms)\n",
|
| 301 |
+
"test_data.set_transform(val_transforms)"
|
| 302 |
+
]
|
| 303 |
+
},
|
| 304 |
+
{
|
| 305 |
+
"cell_type": "code",
|
| 306 |
+
"execution_count": null,
|
| 307 |
+
"id": "0c8a93ca-e4ff-42e2-b58d-445afa0cfee0",
|
| 308 |
+
"metadata": {},
|
| 309 |
+
"outputs": [],
|
| 310 |
+
"source": [
|
| 311 |
+
"def collate_fn(examples):\n",
|
| 312 |
+
" pixel_values = torch.stack([example[\"pixel_values\"] for example in examples])\n",
|
| 313 |
+
" labels = torch.tensor([example['label'] for example in examples])\n",
|
| 314 |
+
" return {\"pixel_values\": pixel_values, \"labels\": labels}"
|
| 315 |
+
]
|
| 316 |
+
},
|
| 317 |
+
{
|
| 318 |
+
"cell_type": "code",
|
| 319 |
+
"execution_count": null,
|
| 320 |
+
"id": "11e0c254-ebb1-4100-a389-9e661d0810ff",
|
| 321 |
+
"metadata": {},
|
| 322 |
+
"outputs": [],
|
| 323 |
+
"source": [
|
| 324 |
+
"model = ViTForImageClassification.from_pretrained(model_str, num_labels=len(labels_list))\n",
|
| 325 |
+
"model.config.id2label = id2label\n",
|
| 326 |
+
"model.config.label2id = label2id\n",
|
| 327 |
+
"\n",
|
| 328 |
+
"print(model.num_parameters(only_trainable=True) / 1e6)"
|
| 329 |
+
]
|
| 330 |
+
},
|
| 331 |
+
{
|
| 332 |
+
"cell_type": "code",
|
| 333 |
+
"execution_count": null,
|
| 334 |
+
"id": "bea51959-9abc-4afc-aee6-0e774f8db9c2",
|
| 335 |
+
"metadata": {},
|
| 336 |
+
"outputs": [],
|
| 337 |
+
"source": [
|
| 338 |
+
"accuracy = evaluate.load(\"accuracy\")\n",
|
| 339 |
+
"\n",
|
| 340 |
+
"def compute_metrics(eval_pred):\n",
|
| 341 |
+
" predictions = eval_pred.predictions\n",
|
| 342 |
+
" label_ids = eval_pred.label_ids\n",
|
| 343 |
+
"\n",
|
| 344 |
+
" predicted_labels = predictions.argmax(axis=1)\n",
|
| 345 |
+
" acc_score = accuracy.compute(predictions=predicted_labels, references=label_ids)['accuracy']\n",
|
| 346 |
+
" \n",
|
| 347 |
+
" return {\n",
|
| 348 |
+
" \"accuracy\": acc_score\n",
|
| 349 |
+
" }"
|
| 350 |
+
]
|
| 351 |
+
},
|
| 352 |
+
{
|
| 353 |
+
"cell_type": "code",
|
| 354 |
+
"execution_count": null,
|
| 355 |
+
"id": "d5ea0bbc-51a3-4b98-823e-10819ffda292",
|
| 356 |
+
"metadata": {},
|
| 357 |
+
"outputs": [],
|
| 358 |
+
"source": [
|
| 359 |
+
"args = TrainingArguments(\n",
|
| 360 |
+
" output_dir=\"deepfake_vit\",\n",
|
| 361 |
+
" logging_dir='./logs',\n",
|
| 362 |
+
" evaluation_strategy=\"epoch\",\n",
|
| 363 |
+
" learning_rate=2e-5,\n",
|
| 364 |
+
" per_device_train_batch_size=32,\n",
|
| 365 |
+
" per_device_eval_batch_size=8,\n",
|
| 366 |
+
" num_train_epochs=4,\n",
|
| 367 |
+
" weight_decay=0.02,\n",
|
| 368 |
+
" warmup_steps=50,\n",
|
| 369 |
+
" remove_unused_columns=False,\n",
|
| 370 |
+
" save_strategy='epoch',\n",
|
| 371 |
+
" load_best_model_at_end=True,\n",
|
| 372 |
+
" save_total_limit=1,\n",
|
| 373 |
+
" report_to=\"none\"\n",
|
| 374 |
+
")"
|
| 375 |
+
]
|
| 376 |
+
},
|
| 377 |
+
{
|
| 378 |
+
"cell_type": "code",
|
| 379 |
+
"execution_count": null,
|
| 380 |
+
"id": "0a965131-c670-43b1-a153-c1a4df611189",
|
| 381 |
+
"metadata": {},
|
| 382 |
+
"outputs": [],
|
| 383 |
+
"source": [
|
| 384 |
+
"trainer = Trainer(\n",
|
| 385 |
+
" model,\n",
|
| 386 |
+
" args,\n",
|
| 387 |
+
" train_dataset=train_data,\n",
|
| 388 |
+
" eval_dataset=test_data,\n",
|
| 389 |
+
" data_collator=collate_fn,\n",
|
| 390 |
+
" compute_metrics=compute_metrics,\n",
|
| 391 |
+
" tokenizer=processor,\n",
|
| 392 |
+
")"
|
| 393 |
+
]
|
| 394 |
+
},
|
| 395 |
+
{
|
| 396 |
+
"cell_type": "code",
|
| 397 |
+
"execution_count": null,
|
| 398 |
+
"id": "ad42ea98-86d6-420e-befe-2ef77eadd76d",
|
| 399 |
+
"metadata": {},
|
| 400 |
+
"outputs": [],
|
| 401 |
+
"source": [
|
| 402 |
+
"trainer.evaluate()"
|
| 403 |
+
]
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"cell_type": "code",
|
| 407 |
+
"execution_count": null,
|
| 408 |
+
"id": "df43c341-0e55-41ef-a274-731c88b9b5d5",
|
| 409 |
+
"metadata": {},
|
| 410 |
+
"outputs": [],
|
| 411 |
+
"source": [
|
| 412 |
+
"trainer.train()"
|
| 413 |
+
]
|
| 414 |
+
},
|
| 415 |
+
{
|
| 416 |
+
"cell_type": "code",
|
| 417 |
+
"execution_count": null,
|
| 418 |
+
"id": "28866dda",
|
| 419 |
+
"metadata": {},
|
| 420 |
+
"outputs": [],
|
| 421 |
+
"source": [
|
| 422 |
+
"trainer.evaluate()"
|
| 423 |
+
]
|
| 424 |
+
},
|
| 425 |
+
{
|
| 426 |
+
"cell_type": "code",
|
| 427 |
+
"execution_count": null,
|
| 428 |
+
"id": "0ec258d9",
|
| 429 |
+
"metadata": {},
|
| 430 |
+
"outputs": [],
|
| 431 |
+
"source": [
|
| 432 |
+
"outputs = trainer.predict(test_data)\n",
|
| 433 |
+
"print(outputs.metrics)"
|
| 434 |
+
]
|
| 435 |
+
},
|
| 436 |
+
{
|
| 437 |
+
"cell_type": "code",
|
| 438 |
+
"execution_count": null,
|
| 439 |
+
"id": "c12a6b10",
|
| 440 |
+
"metadata": {},
|
| 441 |
+
"outputs": [],
|
| 442 |
+
"source": [
|
| 443 |
+
"y_true = outputs.label_ids\n",
|
| 444 |
+
"y_pred = outputs.predictions.argmax(1)\n",
|
| 445 |
+
"\n",
|
| 446 |
+
"def plot_confusion_matrix(cm, classes, title='Confusion Matrix', cmap=plt.cm.Blues, figsize=(10, 8)):\n",
|
| 447 |
+
" \n",
|
| 448 |
+
" plt.figure(figsize=figsize)\n",
|
| 449 |
+
"\n",
|
| 450 |
+
" plt.imshow(cm, interpolation='nearest', cmap=cmap)\n",
|
| 451 |
+
" plt.title(title)\n",
|
| 452 |
+
" plt.colorbar()\n",
|
| 453 |
+
"\n",
|
| 454 |
+
" tick_marks = np.arange(len(classes))\n",
|
| 455 |
+
" plt.xticks(tick_marks, classes, rotation=90)\n",
|
| 456 |
+
" plt.yticks(tick_marks, classes)\n",
|
| 457 |
+
"\n",
|
| 458 |
+
" fmt = '.0f'\n",
|
| 459 |
+
" thresh = cm.max() / 2.0\n",
|
| 460 |
+
" for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):\n",
|
| 461 |
+
" plt.text(j, i, format(cm[i, j], fmt), horizontalalignment=\"center\", color=\"white\" if cm[i, j] > thresh else \"black\")\n",
|
| 462 |
+
"\n",
|
| 463 |
+
" plt.ylabel('True label')\n",
|
| 464 |
+
" plt.xlabel('Predicted label')\n",
|
| 465 |
+
" plt.tight_layout()\n",
|
| 466 |
+
" plt.show()\n",
|
| 467 |
+
"\n",
|
| 468 |
+
"accuracy = accuracy_score(y_true, y_pred)\n",
|
| 469 |
+
"f1 = f1_score(y_true, y_pred, average='macro')\n",
|
| 470 |
+
"\n",
|
| 471 |
+
"print(f\"Accuracy: {accuracy:.4f}\")\n",
|
| 472 |
+
"print(f\"F1 Score: {f1:.4f}\")\n",
|
| 473 |
+
"\n",
|
| 474 |
+
"if len(labels_list) <= 150:\n",
|
| 475 |
+
" cm = confusion_matrix(y_true, y_pred)\n",
|
| 476 |
+
" plot_confusion_matrix(cm, labels_list, figsize=(8, 6))\n",
|
| 477 |
+
"\n",
|
| 478 |
+
"print()\n",
|
| 479 |
+
"print(\"Classification report:\")\n",
|
| 480 |
+
"print()\n",
|
| 481 |
+
"print(classification_report(y_true, y_pred, target_names=labels_list, digits=4))"
|
| 482 |
+
]
|
| 483 |
+
},
|
| 484 |
+
{
|
| 485 |
+
"cell_type": "code",
|
| 486 |
+
"execution_count": null,
|
| 487 |
+
"id": "9889438c",
|
| 488 |
+
"metadata": {},
|
| 489 |
+
"outputs": [],
|
| 490 |
+
"source": [
|
| 491 |
+
"trainer.save_model()"
|
| 492 |
+
]
|
| 493 |
+
},
|
| 494 |
+
{
|
| 495 |
+
"cell_type": "code",
|
| 496 |
+
"execution_count": null,
|
| 497 |
+
"id": "688e3d62",
|
| 498 |
+
"metadata": {},
|
| 499 |
+
"outputs": [],
|
| 500 |
+
"source": [
|
| 501 |
+
"#upload to hub\n",
|
| 502 |
+
"from huggingface_hub import notebook_login\n",
|
| 503 |
+
"notebook_login()"
|
| 504 |
+
]
|
| 505 |
+
},
|
| 506 |
+
{
|
| 507 |
+
"cell_type": "code",
|
| 508 |
+
"execution_count": null,
|
| 509 |
+
"id": "fad56df2",
|
| 510 |
+
"metadata": {},
|
| 511 |
+
"outputs": [],
|
| 512 |
+
"source": [
|
| 513 |
+
"from huggingface_hub import HfApi\n",
|
| 514 |
+
"\n",
|
| 515 |
+
"api = HfApi()\n",
|
| 516 |
+
"repo_id = f\"prithivMLmods/deepfake_vit\"\n",
|
| 517 |
+
"\n",
|
| 518 |
+
"try:\n",
|
| 519 |
+
" api.create_repo(repo_id)\n",
|
| 520 |
+
" print(f\"Repo {repo_id} created\")\n",
|
| 521 |
+
"\n",
|
| 522 |
+
"except:\n",
|
| 523 |
+
" \n",
|
| 524 |
+
" print(f\"Repo {repo_id} already exists\")"
|
| 525 |
+
]
|
| 526 |
+
},
|
| 527 |
+
{
|
| 528 |
+
"cell_type": "code",
|
| 529 |
+
"execution_count": null,
|
| 530 |
+
"id": "f5e1559f",
|
| 531 |
+
"metadata": {},
|
| 532 |
+
"outputs": [],
|
| 533 |
+
"source": [
|
| 534 |
+
"api.upload_folder(\n",
|
| 535 |
+
" folder_path=\"deepfake_vit\", \n",
|
| 536 |
+
" path_in_repo=\".\", \n",
|
| 537 |
+
" repo_id=repo_id, \n",
|
| 538 |
+
" repo_type=\"model\", \n",
|
| 539 |
+
" revision=\"main\"\n",
|
| 540 |
+
")"
|
| 541 |
+
]
|
| 542 |
+
}
|
| 543 |
+
],
|
| 544 |
+
"metadata": {
|
| 545 |
+
"kernelspec": {
|
| 546 |
+
"display_name": "Python 3",
|
| 547 |
+
"language": "python",
|
| 548 |
+
"name": "python3"
|
| 549 |
+
},
|
| 550 |
+
"language_info": {
|
| 551 |
+
"codemirror_mode": {
|
| 552 |
+
"name": "ipython",
|
| 553 |
+
"version": 3
|
| 554 |
+
},
|
| 555 |
+
"file_extension": ".py",
|
| 556 |
+
"mimetype": "text/x-python",
|
| 557 |
+
"name": "python",
|
| 558 |
+
"nbconvert_exporter": "python",
|
| 559 |
+
"pygments_lexer": "ipython3",
|
| 560 |
+
"version": "3.12.7"
|
| 561 |
+
}
|
| 562 |
+
},
|
| 563 |
+
"nbformat": 4,
|
| 564 |
+
"nbformat_minor": 5
|
| 565 |
+
}
|