{"metadata":{"kernelspec":{"name":"python3","display_name":"Python 3","language":"python"},"language_info":{"name":"python","version":"3.12.13","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"nvidiaTeslaT4","dataSources":[{"databundleVersionId":875431,"isSourceIdPinned":false,"mountSlug":"competitions/aptos2019-blindness-detection","sourceId":14774,"sourceType":"competition"}],"dockerImageVersionId":31401,"isGpuEnabled":true,"isInternetEnabled":true,"language":"python","sourceType":"notebook"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"# This Python 3 environment comes with many helpful analytics libraries installed\n# It is defined by the kaggle/python Docker image: https://github.com/kaggle/docker-python\n# For example, here's several helpful packages to load\n\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n\n# Input data files are available in the read-only \"../input/\" directory\n# For example, running this (by clicking run or pressing Shift+Enter) will list all files under the input directory\n\nimport os\nfor dirname, _, filenames in os.walk('/kaggle/input'):\n    for filename in filenames:\n        print(os.path.join(dirname, filename))\n\n# You can write up to 20GB to the current directory (/kaggle/working/) that gets preserved as output when you create a version using \"Save & Run All\" \n# You can also write temporary files to /kaggle/temp/, but they won't be saved outside of the current session\n\n# Use the kagglehub client library to attach Kaggle resources like competitions, datasets, and models to your session\n# Learn more about kagglehub: https://github.com/Kaggle/kagglehub/blob/main/README.md\n\nimport kagglehub\n# kagglehub.dataset_download('<owner>/<dataset-slug>')","metadata":{"_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","execution":{"iopub.status.busy":"2026-06-01T06:05:26.004488Z","iopub.execute_input":"2026-06-01T06:05:26.004715Z","iopub.status.idle":"2026-06-01T06:05:43.252557Z","shell.execute_reply.started":"2026-06-01T06:05:26.004691Z","shell.execute_reply":"2026-06-01T06:05:43.251973Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"import os\nimport random\nimport warnings\nwarnings.filterwarnings('ignore')\n\nimport numpy as np\nimport pandas as pd\nimport matplotlib.pyplot as plt\nimport seaborn as sns\nimport cv2\nfrom PIL import Image\nfrom tqdm import tqdm\n\nimport torch\nimport torch.nn as nn\nimport torch.nn.functional as F\nimport torch.optim as optim\nfrom torch.utils.data import Dataset, DataLoader\nimport torchvision.transforms as T\nimport torchvision.models as models\nfrom torch.optim.lr_scheduler import OneCycleLR\n\nfrom sklearn.metrics import (\n    accuracy_score, precision_score, recall_score,\n    f1_score, confusion_matrix, classification_report,\n    cohen_kappa_score\n)\nfrom sklearn.model_selection import train_test_split, GridSearchCV\nfrom sklearn.svm import SVC\nfrom sklearn.preprocessing import StandardScaler\nfrom sklearn.utils.class_weight import compute_class_weight\n\nprint(\"All libraries imported!\")","metadata":{"execution":{"iopub.status.busy":"2026-06-01T06:05:43.253940Z","iopub.execute_input":"2026-06-01T06:05:43.254280Z","iopub.status.idle":"2026-06-01T06:05:57.284981Z","shell.execute_reply.started":"2026-06-01T06:05:43.254257Z","shell.execute_reply":"2026-06-01T06:05:57.284322Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Cell 2: Optimized Configuration (Fast Training)\n\nSEED = 42\nrandom.seed(SEED)\nnp.random.seed(SEED)\ntorch.manual_seed(SEED)\ntorch.cuda.manual_seed_all(SEED)\n\nDEVICE = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\nprint(f'Device: {DEVICE}')\n\n# Optimized for speed\nIMG_SIZE = 224  # Standard size (faster than 260)\nBATCH_SIZE = 64  # Larger batch = faster training\nEPOCHS = 30  # Reduced from 60\nBACKBONE_LR = 2e-4  # Slightly higher\nHEAD_LR = 1e-3\nWEIGHT_DECAY = 1e-4\nPATIENCE = 8  # Early stopping\nMIXUP_PROB = 0.2  # Reduced\nLABEL_SMOOTHING = 0.1\n\nBASE = '/kaggle/input/competitions/aptos2019-blindness-detection'\nCLASS_NAMES = ['No DR', 'Mild', 'Moderate', 'Severe', 'Proliferative']\nNUM_CLASSES = 5\nMEAN = [0.485, 0.456, 0.406]\nSTD = [0.229, 0.224, 0.225]\n\nprint(\"Optimized configuration loaded!\")\nprint(f\"Epochs: {EPOCHS}, Batch Size: {BATCH_SIZE}, Image Size: {IMG_SIZE}\")","metadata":{"execution":{"iopub.status.busy":"2026-06-01T06:05:57.285816Z","iopub.execute_input":"2026-06-01T06:05:57.286296Z","iopub.status.idle":"2026-06-01T06:05:57.552523Z","shell.execute_reply.started":"2026-06-01T06:05:57.286272Z","shell.execute_reply":"2026-06-01T06:05:57.551583Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Cell 3: Fast CLAHE (Removed expensive operations)\n\ndef apply_clahe_fast(image_path, img_size=IMG_SIZE, clip_limit=2.0):\n    \"\"\"Fast CLAHE - no sharpening overhead\"\"\"\n    try:\n        img = cv2.imread(image_path)\n        if img is None:\n            raise ValueError('Cannot read')\n        img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n        img = cv2.resize(img, (img_size, img_size))\n        \n        lab = cv2.cvtColor(img, cv2.COLOR_RGB2LAB)\n        l, a, b = cv2.split(lab)\n        clahe = cv2.createCLAHE(clipLimit=clip_limit, tileGridSize=(8, 8))\n        l_c = clahe.apply(l)\n        result = cv2.cvtColor(cv2.merge([l_c, a, b]), cv2.COLOR_LAB2RGB)\n        \n        return Image.fromarray(result)\n    except Exception:\n        return Image.open(image_path).convert('RGB').resize((img_size, img_size))\n\ndef find_image(id_code, folders):\n    for folder in folders:\n        for ext in ['.png', '.jpeg', '.jpg']:\n            p = os.path.join(BASE, folder, str(id_code) + ext)\n            if os.path.exists(p):\n                return p\n    return None\n\nclass FundusDataset(Dataset):\n    def __init__(self, df, transform=None):\n        self.df = df.reset_index(drop=True)\n        self.transform = transform\n\n    def __len__(self):\n        return len(self.df)\n\n    def __getitem__(self, idx):\n        row = self.df.iloc[idx]\n        img = apply_clahe_fast(row['filepath'], IMG_SIZE)\n        if self.transform:\n            img = self.transform(img)\n        return img, int(row['label'])\n\ndef mixup_batch(imgs, labels, alpha=0.2):\n    lam = np.random.beta(alpha, alpha)\n    bs = imgs.size(0)\n    idx = torch.randperm(bs, device=imgs.device)\n    mixed = lam * imgs + (1 - lam) * imgs[idx]\n    return mixed, labels, labels[idx], lam\n\nprint(\"Fast helper functions ready!\")","metadata":{"execution":{"iopub.status.busy":"2026-06-01T06:05:57.553632Z","iopub.execute_input":"2026-06-01T06:05:57.554005Z","iopub.status.idle":"2026-06-01T06:05:57.564504Z","shell.execute_reply.started":"2026-06-01T06:05:57.553974Z","shell.execute_reply":"2026-06-01T06:05:57.563762Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Cell 4: EfficientNet-B0 (Optimized)\n\nclass EfficientNetB0_DR(nn.Module):\n    def __init__(self, num_classes=NUM_CLASSES, dropout=0.4):\n        super().__init__()\n        base = models.efficientnet_b0(\n            weights=models.EfficientNet_B0_Weights.IMAGENET1K_V1\n        )\n        self.features = base.features\n        self.avgpool = base.avgpool\n        \n        # Simpler classifier (faster)\n        self.classifier = nn.Sequential(\n            nn.Dropout(p=dropout),\n            nn.Linear(1280, 512),\n            nn.BatchNorm1d(512),\n            nn.ReLU(inplace=True),\n            nn.Dropout(p=0.3),\n            nn.Linear(512, 256),\n            nn.BatchNorm1d(256),\n            nn.ReLU(inplace=True),\n            nn.Dropout(p=0.2),\n            nn.Linear(256, num_classes),\n        )\n\n    def forward(self, x):\n        x = self.features(x)\n        x = self.avgpool(x)\n        x = x.flatten(1)\n        return self.classifier(x)\n    \n    def extract_features(self, x):\n        x = self.features(x)\n        x = self.avgpool(x)\n        x = x.flatten(1)\n        return x\n\nprint(\"EfficientNet-B0 defined!\")","metadata":{"execution":{"iopub.status.busy":"2026-06-01T06:05:57.565556Z","iopub.execute_input":"2026-06-01T06:05:57.565859Z","iopub.status.idle":"2026-06-01T06:05:57.589749Z","shell.execute_reply.started":"2026-06-01T06:05:57.565828Z","shell.execute_reply":"2026-06-01T06:05:57.589013Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Cell 5: Efficient Transforms\n\ntrain_tfm = T.Compose([\n    T.RandomHorizontalFlip(p=0.5),\n    T.RandomVerticalFlip(p=0.3),\n    T.RandomRotation(30),\n    T.ColorJitter(brightness=0.2, contrast=0.2, saturation=0.15, hue=0.02),\n    T.RandomAffine(degrees=0, translate=(0.1, 0.1), scale=(0.9, 1.1)),\n    T.ToTensor(),\n    T.Normalize(MEAN, STD),\n])\n\nval_tfm = T.Compose([\n    T.ToTensor(),\n    T.Normalize(MEAN, STD),\n])\n\nprint(\"Efficient transforms defined!\")","metadata":{"execution":{"iopub.status.busy":"2026-06-01T06:05:57.590643Z","iopub.execute_input":"2026-06-01T06:05:57.591364Z","iopub.status.idle":"2026-06-01T06:05:57.612103Z","shell.execute_reply.started":"2026-06-01T06:05:57.591341Z","shell.execute_reply":"2026-06-01T06:05:57.611492Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Cell 6: Load Data (80-20 Split)\n\nTRAIN_FOLDER = ['train_images']\ndf_raw = pd.read_csv(f'{BASE}/train.csv')\ndf_raw['label'] = df_raw['diagnosis']\ndf_raw['filepath'] = df_raw['id_code'].apply(\n    lambda x: find_image(x, TRAIN_FOLDER))\ndf_raw = df_raw[df_raw['filepath'].notna()].reset_index(drop=True)\n\ndf_tr, df_te = train_test_split(\n    df_raw, test_size=0.20, stratify=df_raw['label'], random_state=SEED)\n\nprint(f'Train: {len(df_tr)} | Test: {len(df_te)}')\nprint('\\nClass distribution (train):')\nfor i, name in enumerate(CLASS_NAMES):\n    c = (df_tr['label'] == i).sum()\n    print(f'  {name:15s}: {c:4d}')\n\ntrain_dl = DataLoader(FundusDataset(df_tr, train_tfm),\n                      batch_size=BATCH_SIZE, shuffle=True, num_workers=2, pin_memory=True)\ntest_dl = DataLoader(FundusDataset(df_te, val_tfm),\n                     batch_size=BATCH_SIZE, shuffle=False, num_workers=2, pin_memory=True)\n\ncw = compute_class_weight('balanced', classes=np.arange(NUM_CLASSES), y=df_tr['label'].values)\ncriterion = nn.CrossEntropyLoss(\n    weight=torch.tensor(cw, dtype=torch.float).to(DEVICE),\n    label_smoothing=LABEL_SMOOTHING\n)\n\nprint('\\nClass weights:', {n: f'{w:.3f}' for n, w in zip(CLASS_NAMES, cw)})\nprint('✅ Data ready!')","metadata":{"execution":{"iopub.status.busy":"2026-06-01T06:05:57.614524Z","iopub.execute_input":"2026-06-01T06:05:57.614842Z","iopub.status.idle":"2026-06-01T06:06:02.602546Z","shell.execute_reply.started":"2026-06-01T06:05:57.614798Z","shell.execute_reply":"2026-06-01T06:06:02.601722Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Cell 7: Initialize Model\n\nmodel = EfficientNetB0_DR(NUM_CLASSES).to(DEVICE)\n\ntotal_params = sum(p.numel() for p in model.parameters())\nprint(f'Model parameters: {total_params:,}')\n\noptimizer = optim.AdamW([\n    {'params': model.features.parameters(), 'lr': BACKBONE_LR},\n    {'params': model.classifier.parameters(), 'lr': HEAD_LR},\n], weight_decay=WEIGHT_DECAY)\n\n# OneCycleLR is fast and effective\nscheduler = OneCycleLR(\n    optimizer,\n    max_lr=[BACKBONE_LR, HEAD_LR],\n    steps_per_epoch=len(train_dl),\n    epochs=EPOCHS,\n    pct_start=0.2,\n)\n\nprint(\"Model initialized!\")","metadata":{"execution":{"iopub.status.busy":"2026-06-01T06:06:02.603449Z","iopub.execute_input":"2026-06-01T06:06:02.603733Z","iopub.status.idle":"2026-06-01T06:06:03.092013Z","shell.execute_reply.started":"2026-06-01T06:06:02.603712Z","shell.execute_reply":"2026-06-01T06:06:03.091418Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Cell 8: Fast Training Loop\n\nbest_train_acc = 0.0\ntrain_losses, train_accs = [], []\npatience_ctr = 0\n\nprint(f'\\nTraining for {EPOCHS} epochs...\\n')\nprint(f'{\"Epoch\":>6}  {\"Loss\":>8}  {\"Acc\":>8}  {\"Time\":>8}  Status')\nprint('-' * 50)\n\nimport time\n\nfor epoch in range(1, EPOCHS + 1):\n    start_time = time.time()\n    model.train()\n    tr_loss = 0.0\n    tr_correct = 0\n    total_train = 0\n    \n    for imgs, lbls in train_dl:\n        imgs, lbls = imgs.to(DEVICE), lbls.to(DEVICE)\n        \n        if random.random() < MIXUP_PROB:\n            mixed, lbl_a, lbl_b, lam = mixup_batch(imgs, lbls)\n            optimizer.zero_grad()\n            out = model(mixed)\n            loss = lam * criterion(out, lbl_a) + (1 - lam) * criterion(out, lbl_b)\n        else:\n            optimizer.zero_grad()\n            out = model(imgs)\n            loss = criterion(out, lbls)\n        \n        loss.backward()\n        optimizer.step()\n        scheduler.step()\n        \n        tr_loss += loss.item()\n        tr_correct += (out.argmax(1) == lbls).sum().item()\n        total_train += lbls.size(0)\n    \n    t_loss = tr_loss / len(train_dl)\n    t_acc = tr_correct / total_train * 100\n    \n    train_losses.append(t_loss)\n    train_accs.append(t_acc)\n    \n    epoch_time = time.time() - start_time\n    \n    if t_acc > best_train_acc:\n        best_train_acc = t_acc\n        patience_ctr = 0\n        torch.save(model.state_dict(), '/kaggle/working/effb0_fast_best.pth')\n        status = f'✓ saved (best {t_acc:.2f}%)'\n    else:\n        patience_ctr += 1\n        status = f'patience {patience_ctr}/{PATIENCE}'\n    \n    print(f'{epoch:>6}  {t_loss:>8.4f}  {t_acc:>7.2f}%  {epoch_time:>7.1f}s  {status}')\n    \n    if patience_ctr >= PATIENCE:\n        print(f'\\nEarly stopping at epoch {epoch}')\n        break\n\nprint(f'\\n✅ Best Training Accuracy: {best_train_acc:.2f}%')","metadata":{"execution":{"iopub.status.busy":"2026-06-01T06:06:03.092842Z","iopub.execute_input":"2026-06-01T06:06:03.093170Z","iopub.status.idle":"2026-06-01T07:20:46.521585Z","shell.execute_reply.started":"2026-06-01T06:06:03.093147Z","shell.execute_reply":"2026-06-01T07:20:46.520625Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Cell 9: Plot Training Curves\n\nfig, axes = plt.subplots(1, 2, figsize=(14, 5))\n\naxes[0].plot(train_losses, linewidth=2)\naxes[0].set_xlabel('Epoch')\naxes[0].set_ylabel('Loss')\naxes[0].set_title('Training Loss')\naxes[0].grid(True, alpha=0.3)\n\naxes[1].plot(train_accs, linewidth=2)\naxes[1].axhline(y=best_train_acc, color='r', linestyle='--', label=f'Best: {best_train_acc:.2f}%')\naxes[1].set_xlabel('Epoch')\naxes[1].set_ylabel('Accuracy (%)')\naxes[1].set_title('Training Accuracy')\naxes[1].legend()\naxes[1].grid(True, alpha=0.3)\n\nplt.tight_layout()\nplt.savefig('/kaggle/working/training_curves.png', dpi=150)\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2026-06-01T07:20:46.522925Z","iopub.execute_input":"2026-06-01T07:20:46.523196Z","iopub.status.idle":"2026-06-01T07:20:47.149893Z","shell.execute_reply.started":"2026-06-01T07:20:46.523168Z","shell.execute_reply":"2026-06-01T07:20:47.149280Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Cell 10: Fast Feature Extraction (No TTA - Saves Time)\n\nmodel.load_state_dict(torch.load('/kaggle/working/effb0_fast_best.pth', map_location=DEVICE))\nmodel.eval()\n\nprint(\"\\nExtracting features (fast, no TTA)...\")\n\n@torch.no_grad()\ndef extract_features(model, dataloader, device):\n    all_features = []\n    all_labels = []\n    for images, labels in tqdm(dataloader):\n        images = images.to(device)\n        features = model.extract_features(images).cpu().numpy()\n        all_features.append(features)\n        all_labels.append(labels.numpy())\n    return np.vstack(all_features), np.concatenate(all_labels)\n\nprint(\"\\nExtracting training features...\")\ntrain_features, train_labels = extract_features(model, train_dl, DEVICE)\n\nprint(\"Extracting test features...\")\ntest_features, test_labels = extract_features(model, test_dl, DEVICE)\n\nprint(f\"\\n✅ Done! Train: {train_features.shape}, Test: {test_features.shape}\")","metadata":{"execution":{"iopub.status.busy":"2026-06-01T07:20:47.150672Z","iopub.execute_input":"2026-06-01T07:20:47.150979Z","iopub.status.idle":"2026-06-01T07:23:53.162288Z","shell.execute_reply.started":"2026-06-01T07:20:47.150942Z","shell.execute_reply":"2026-06-01T07:23:53.161528Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Cell 11: Fast SVM Training\n\nprint(\"\\nTraining SVM...\")\n\nscaler = StandardScaler()\ntrain_scaled = scaler.fit_transform(train_features)\ntest_scaled = scaler.transform(test_features)\n\n# Small grid search for speed\nparam_grid = {\n    'C': [1, 5, 10, 50],\n    'gamma': ['scale', 0.01],\n    'kernel': ['rbf'],\n    'class_weight': ['balanced']\n}\n\ngrid_search = GridSearchCV(\n    SVC(random_state=SEED, probability=True),\n    param_grid,\n    cv=3,\n    scoring='accuracy',\n    n_jobs=-1,\n    verbose=1\n)\n\ngrid_search.fit(train_scaled, train_labels)\n\nprint(f\"\\n✅ Best parameters: {grid_search.best_params_}\")\nprint(f\"Best CV accuracy: {grid_search.best_score_:.4f}\")\n\nbest_svm = SVC(\n    kernel='rbf',\n    C=grid_search.best_params_['C'],\n    gamma=grid_search.best_params_['gamma'],\n    class_weight='balanced',\n    random_state=SEED\n)\n\nbest_svm.fit(train_scaled, train_labels)\ntest_predictions = best_svm.predict(test_scaled)","metadata":{"execution":{"iopub.status.busy":"2026-06-01T07:23:53.163574Z","iopub.execute_input":"2026-06-01T07:23:53.163814Z","iopub.status.idle":"2026-06-01T07:27:23.284358Z","shell.execute_reply.started":"2026-06-01T07:23:53.163788Z","shell.execute_reply":"2026-06-01T07:27:23.283264Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Cell 12: Results\n\nacc = accuracy_score(test_labels, test_predictions)\nprec = precision_score(test_labels, test_predictions, average='weighted')\nrec = recall_score(test_labels, test_predictions, average='weighted')\nf1 = f1_score(test_labels, test_predictions, average='weighted')\nkappa = cohen_kappa_score(test_labels, test_predictions, weights='quadratic')\n\nprint('\\n' + '='*60)\nprint('  FAST EFFICIENTNET-B0 + SVM  —  5-Class DR')\nprint('='*60)\nprint(f'  Test Accuracy      : {acc*100:.2f}%')\nprint(f'  Weighted Precision : {prec:.4f}')\nprint(f'  Weighted Recall    : {rec:.4f}')\nprint(f'  Weighted F1-Score  : {f1:.4f}')\nprint(f'  Quadratic Kappa    : {kappa:.4f}')\nprint('='*60)\n\nprint('\\n📊 Classification Report:')\nprint(classification_report(test_labels, test_predictions, target_names=CLASS_NAMES))\n\n# Confusion Matrix\ncm = confusion_matrix(test_labels, test_predictions)\nplt.figure(figsize=(10, 8))\nsns.heatmap(cm, annot=True, fmt='d', cmap='Blues',\n            xticklabels=CLASS_NAMES, yticklabels=CLASS_NAMES)\nplt.title(f'Test Accuracy: {acc*100:.2f}% | Kappa: {kappa:.4f}')\nplt.ylabel('True Label')\nplt.xlabel('Predicted Label')\nplt.xticks(rotation=45)\nplt.tight_layout()\nplt.savefig('/kaggle/working/confusion_matrix.png', dpi=150)\nplt.show()\n\nprint(\"\\n\" + \"=\"*60)\nprint(f\"🎯 FINAL TEST ACCURACY: {acc*100:.2f}%\")\nprint(f\"   Original: 82.55%\")\nprint(f\"   Improvement: +{(acc*100 - 82.55):.2f}%\")\nprint(\"=\"*60)","metadata":{"execution":{"iopub.status.busy":"2026-06-01T07:27:23.285623Z","iopub.execute_input":"2026-06-01T07:27:23.286293Z","iopub.status.idle":"2026-06-01T07:27:23.786125Z","shell.execute_reply.started":"2026-06-01T07:27:23.286264Z","shell.execute_reply":"2026-06-01T07:27:23.785389Z"},"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"import joblib\n\njoblib.dump(best_svm, \"/kaggle/working/best_svm.pkl\")\njoblib.dump(scaler, \"/kaggle/working/scaler.pkl\")","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-06-01T07:28:27.025861Z","iopub.execute_input":"2026-06-01T07:28:27.026450Z","iopub.status.idle":"2026-06-01T07:28:27.045595Z","shell.execute_reply.started":"2026-06-01T07:28:27.026418Z","shell.execute_reply":"2026-06-01T07:28:27.044715Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"","metadata":{"trusted":true},"outputs":[],"execution_count":null}]}