{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2025-01-27 16:42:47.893272: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:477] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered\n",
      "WARNING: All log messages before absl::InitializeLog() is called are written to STDERR\n",
      "E0000 00:00:1737974567.952170    6163 cuda_dnn.cc:8310] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered\n",
      "E0000 00:00:1737974567.970030    6163 cuda_blas.cc:1418] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered\n",
      "2025-01-27 16:42:48.097247: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.\n",
      "To enable the following instructions: AVX2 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "from torch.utils.data import DataLoader\n",
    "from torchvision import transforms\n",
    "from torchvision.datasets import ImageFolder\n",
    "from transformers import ViTForImageClassification, ViTFeatureExtractor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Using device: cuda\n"
     ]
    }
   ],
   "source": [
    "# Define device\n",
    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
    "print(f\"Using device: {device}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Paths\n",
    "dataset_path = \"/home/shanin/Desktop/SHANIN/MAIN/ALL_CODE/Face_Recognition/FACE_CROP\"  # Replace with your dataset path\n",
    "model_save_path = \"/home/shanin/Desktop/SHANIN/MAIN/ALL_CODE/Face_Recognition/v1.pth\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of ViTForImageClassification were not initialized from the model checkpoint at google/vit-base-patch16-224 and are newly initialized because the shapes did not match:\n",
      "- classifier.bias: found shape torch.Size([1000]) in the checkpoint and torch.Size([1364]) in the model instantiated\n",
      "- classifier.weight: found shape torch.Size([1000, 768]) in the checkpoint and torch.Size([1364, 768]) in the model instantiated\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
     ]
    }
   ],
   "source": [
    "# Data Transformations with 160x160 resize\n",
    "transform = transforms.Compose([\n",
    "    transforms.Resize((224, 224)),  # Resize to 160x160 as per your dataset\n",
    "    transforms.RandomHorizontalFlip(),\n",
    "    transforms.RandomRotation(15),\n",
    "    transforms.ToTensor(),\n",
    "    transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5]),\n",
    "])\n",
    "\n",
    "# Load Dataset\n",
    "dataset = ImageFolder(root=dataset_path, transform=transform)\n",
    "\n",
    "# DataLoader (using all images)\n",
    "dataloader = DataLoader(dataset, batch_size=16, shuffle=True, num_workers=4)\n",
    "\n",
    "# Define ViT Model\n",
    "model = ViTForImageClassification.from_pretrained(\n",
    "    \"google/vit-base-patch16-224\", num_labels=len(dataset.classes), \n",
    "    ignore_mismatched_sizes=True  # This will ignore size mismatch warnings\n",
    ")\n",
    "\n",
    "# Modify the classifier head to match the number of classes in your dataset\n",
    "model.classifier = nn.Linear(model.config.hidden_size, len(dataset.classes))\n",
    "\n",
    "# Move model to device\n",
    "model = model.to(device)\n",
    "\n",
    "# Define Optimizer, Loss Function, and Scheduler\n",
    "optimizer = torch.optim.AdamW(model.parameters(), lr=2e-5, weight_decay=1e-4)\n",
    "criterion = nn.CrossEntropyLoss()\n",
    "scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=5, gamma=0.1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Training Function\n",
    "def train_model(model, dataloader, epochs=100):\n",
    "    model.train()\n",
    "\n",
    "    for epoch in range(epochs):\n",
    "        print(f\"Epoch {epoch + 1}/{epochs}\")\n",
    "        epoch_loss = 0.0\n",
    "\n",
    "        for images, labels in dataloader:\n",
    "            images, labels = images.to(device), labels.to(device)\n",
    "\n",
    "            # Forward pass\n",
    "            optimizer.zero_grad()\n",
    "            outputs = model(images).logits\n",
    "            loss = criterion(outputs, labels)\n",
    "\n",
    "            # Backward pass\n",
    "            loss.backward()\n",
    "            optimizer.step()\n",
    "\n",
    "            epoch_loss += loss.item()\n",
    "\n",
    "        # Step the scheduler\n",
    "        scheduler.step()\n",
    "\n",
    "        print(f\"Epoch Loss: {epoch_loss / len(dataloader):.4f}\")\n",
    "\n",
    "    # Save the trained model\n",
    "    torch.save(model.state_dict(), model_save_path)\n",
    "    print(\"Training complete! Model saved.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "Epoch Loss: 6.4940\n",
      "Epoch 2/100\n",
      "Epoch Loss: 4.4488\n",
      "Epoch 3/100\n",
      "Epoch Loss: 2.7864\n",
      "Epoch 4/100\n",
      "Epoch Loss: 1.6240\n",
      "Epoch 5/100\n",
      "Epoch Loss: 0.9264\n",
      "Epoch 6/100\n",
      "Epoch Loss: 0.5430\n",
      "Epoch 7/100\n",
      "Epoch Loss: 0.4927\n",
      "Epoch 8/100\n",
      "Epoch Loss: 0.4542\n",
      "Epoch 9/100\n",
      "Epoch Loss: 0.4172\n",
      "Epoch 10/100\n",
      "Epoch Loss: 0.3850\n",
      "Epoch 11/100\n",
      "Epoch Loss: 0.3539\n",
      "Epoch 12/100\n",
      "Epoch Loss: 0.3482\n",
      "Epoch 13/100\n",
      "Epoch Loss: 0.3455\n",
      "Epoch 14/100\n",
      "Epoch Loss: 0.3437\n",
      "Epoch 15/100\n",
      "Epoch Loss: 0.3408\n",
      "Epoch 16/100\n",
      "Epoch Loss: 0.3362\n",
      "Epoch 17/100\n",
      "Epoch Loss: 0.3362\n",
      "Epoch 18/100\n",
      "Epoch Loss: 0.3366\n",
      "Epoch 19/100\n",
      "Epoch Loss: 0.3359\n",
      "Epoch 20/100\n",
      "Epoch Loss: 0.3364\n",
      "Epoch 21/100\n",
      "Epoch Loss: 0.3349\n",
      "Epoch 22/100\n",
      "Epoch Loss: 0.3352\n",
      "Epoch 23/100\n",
      "Epoch Loss: 0.3356\n",
      "Epoch 24/100\n",
      "Epoch Loss: 0.3347\n",
      "Epoch 25/100\n",
      "Epoch Loss: 0.3343\n",
      "Epoch 26/100\n",
      "Epoch Loss: 0.3355\n",
      "Epoch 27/100\n",
      "Epoch Loss: 0.3347\n",
      "Epoch 28/100\n",
      "Epoch Loss: 0.3350\n",
      "Epoch 29/100\n",
      "Epoch Loss: 0.3354\n",
      "Epoch 30/100\n",
      "Epoch Loss: 0.3350\n",
      "Epoch 31/100\n",
      "Epoch Loss: 0.3356\n",
      "Epoch 32/100\n",
      "Epoch Loss: 0.3358\n",
      "Epoch 33/100\n",
      "Epoch Loss: 0.3349\n",
      "Epoch 34/100\n",
      "Epoch Loss: 0.3354\n",
      "Epoch 35/100\n",
      "Epoch Loss: 0.3347\n",
      "Epoch 36/100\n",
      "Epoch Loss: 0.3352\n",
      "Epoch 37/100\n",
      "Epoch Loss: 0.3351\n",
      "Epoch 38/100\n",
      "Epoch Loss: 0.3349\n",
      "Epoch 39/100\n",
      "Epoch Loss: 0.3343\n",
      "Epoch 40/100\n",
      "Epoch Loss: 0.3356\n",
      "Epoch 41/100\n",
      "Epoch Loss: 0.3349\n",
      "Epoch 42/100\n",
      "Epoch Loss: 0.3348\n",
      "Epoch 43/100\n",
      "Epoch Loss: 0.3348\n",
      "Epoch 44/100\n",
      "Epoch Loss: 0.3360\n",
      "Epoch 45/100\n",
      "Epoch Loss: 0.3352\n",
      "Epoch 46/100\n",
      "Epoch Loss: 0.3344\n",
      "Epoch 47/100\n",
      "Epoch Loss: 0.3351\n",
      "Epoch 48/100\n",
      "Epoch Loss: 0.3360\n",
      "Epoch 49/100\n",
      "Epoch Loss: 0.3351\n",
      "Epoch 50/100\n",
      "Epoch Loss: 0.3348\n",
      "Epoch 51/100\n",
      "Epoch Loss: 0.3344\n",
      "Epoch 52/100\n",
      "Epoch Loss: 0.3347\n",
      "Epoch 53/100\n",
      "Epoch Loss: 0.3349\n",
      "Epoch 54/100\n",
      "Epoch Loss: 0.3359\n",
      "Epoch 55/100\n",
      "Epoch Loss: 0.3353\n",
      "Epoch 56/100\n",
      "Epoch Loss: 0.3347\n",
      "Epoch 57/100\n",
      "Epoch Loss: 0.3355\n",
      "Epoch 58/100\n",
      "Epoch Loss: 0.3356\n",
      "Epoch 59/100\n",
      "Epoch Loss: 0.3352\n",
      "Epoch 60/100\n",
      "Epoch Loss: 0.3357\n",
      "Epoch 61/100\n",
      "Epoch Loss: 0.3359\n",
      "Epoch 62/100\n",
      "Epoch Loss: 0.3357\n",
      "Epoch 63/100\n",
      "Epoch Loss: 0.3349\n",
      "Epoch 64/100\n",
      "Epoch Loss: 0.3358\n",
      "Epoch 65/100\n",
      "Epoch Loss: 0.3349\n",
      "Epoch 66/100\n",
      "Epoch Loss: 0.3347\n",
      "Epoch 67/100\n",
      "Epoch Loss: 0.3359\n",
      "Epoch 68/100\n",
      "Epoch Loss: 0.3349\n",
      "Epoch 69/100\n",
      "Epoch Loss: 0.3338\n",
      "Epoch 70/100\n",
      "Epoch Loss: 0.3351\n",
      "Epoch 71/100\n",
      "Epoch Loss: 0.3358\n",
      "Epoch 72/100\n",
      "Epoch Loss: 0.3347\n",
      "Epoch 73/100\n",
      "Epoch Loss: 0.3353\n",
      "Epoch 74/100\n",
      "Epoch Loss: 0.3347\n",
      "Epoch 75/100\n",
      "Epoch Loss: 0.3344\n",
      "Epoch 76/100\n",
      "Epoch Loss: 0.3341\n",
      "Epoch 77/100\n",
      "Epoch Loss: 0.3352\n",
      "Epoch 78/100\n",
      "Epoch Loss: 0.3349\n",
      "Epoch 79/100\n",
      "Epoch Loss: 0.3344\n",
      "Epoch 80/100\n",
      "Epoch Loss: 0.3350\n",
      "Epoch 81/100\n",
      "Epoch Loss: 0.3351\n",
      "Epoch 82/100\n",
      "Epoch Loss: 0.3347\n",
      "Epoch 83/100\n",
      "Epoch Loss: 0.3358\n",
      "Epoch 84/100\n",
      "Epoch Loss: 0.3346\n",
      "Epoch 85/100\n",
      "Epoch Loss: 0.3351\n",
      "Epoch 86/100\n",
      "Epoch Loss: 0.3347\n",
      "Epoch 87/100\n",
      "Epoch Loss: 0.3364\n",
      "Epoch 88/100\n",
      "Epoch Loss: 0.3356\n",
      "Epoch 89/100\n",
      "Epoch Loss: 0.3349\n",
      "Epoch 90/100\n",
      "Epoch Loss: 0.3347\n",
      "Epoch 91/100\n",
      "Epoch Loss: 0.3346\n",
      "Epoch 92/100\n",
      "Epoch Loss: 0.3354\n",
      "Epoch 93/100\n",
      "Epoch Loss: 0.3362\n",
      "Epoch 94/100\n",
      "Epoch Loss: 0.3344\n",
      "Epoch 95/100\n",
      "Epoch Loss: 0.3351\n",
      "Epoch 96/100\n",
      "Epoch Loss: 0.3346\n",
      "Epoch 97/100\n",
      "Epoch Loss: 0.3352\n",
      "Epoch 98/100\n",
      "Epoch Loss: 0.3343\n",
      "Epoch 99/100\n",
      "Epoch Loss: 0.3352\n",
      "Epoch 100/100\n",
      "Epoch Loss: 0.3346\n",
      "Training complete! Model saved.\n"
     ]
    }
   ],
   "source": [
    "# Train the Model\n",
    "train_model(model, dataloader, epochs=100)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "minibat",
   "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": "3.12.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}