{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.12.12","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"tpuV5e8","dataSources":[{"sourceId":21154,"databundleVersionId":1243559,"sourceType":"competition"},{"sourceId":1138814,"sourceType":"datasetVersion","datasetId":601927}],"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":false}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"# Flower Classification Swin Large with External Data","metadata":{}},{"cell_type":"markdown","source":"This notebook is designed to show how Swin Large transformer model can be used with TPU.  It is based on my [earlier notebook](https://www.kaggle.com/code/atamazian/fc-ensemble-external-data-effnet-densenet), and it uses [external data](https://www.kaggle.com/kirillblinov/tf-flower-photo-tfrec) to increase model's accuracy.\n\nHave any suggestions? Feel free to comment.\n\n**<span style=\"color:red\">If you liked this kernel, please don't forget to upvote it!</span>**","metadata":{}},{"cell_type":"code","source":"!export PATH=\"${HOME}/.local/bin:${PATH}\" && uv pip uninstall --system jax\n!export PATH=\"${HOME}/.local/bin:${PATH}\" && uv pip install --system tensorflow-tpu==\"2.18.0\" --find-links https://storage.googleapis.com/libtpu-tf-releases/index.html","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-25T15:55:50.159274Z","iopub.execute_input":"2025-11-25T15:55:50.159412Z","iopub.status.idle":"2025-11-25T15:56:04.105037Z","shell.execute_reply.started":"2025-11-25T15:55:50.159397Z","shell.execute_reply":"2025-11-25T15:56:04.104048Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"!pip install -qU tfswin wandb==0.22.3","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-25T15:56:04.105696Z","iopub.execute_input":"2025-11-25T15:56:04.105885Z","iopub.status.idle":"2025-11-25T15:56:17.777121Z","shell.execute_reply.started":"2025-11-25T15:56:04.105865Z","shell.execute_reply":"2025-11-25T15:56:17.776046Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"import math, re, os, random\nimport numpy as np\nimport pandas as pd\nimport wandb\nfrom wandb.integration.keras import WandbMetricsLogger\nfrom matplotlib import pyplot as plt\nfrom sklearn.metrics import f1_score, precision_score, \\\n                            recall_score, confusion_matrix\n\nimport tensorflow as tf\nfrom tensorflow.keras.metrics import F1Score\nfrom tensorflow.keras import layers as L\nfrom tensorflow.keras import callbacks\nfrom tfswin import SwinTransformerLarge224\n\nprint(\"TF version \" + tf.__version__)","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true,"execution":{"iopub.status.busy":"2025-11-25T15:56:17.777663Z","iopub.execute_input":"2025-11-25T15:56:17.777864Z","iopub.status.idle":"2025-11-25T15:56:39.509674Z","shell.execute_reply.started":"2025-11-25T15:56:17.777847Z","shell.execute_reply":"2025-11-25T15:56:39.508637Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"We will use wandb for logging purposes. If you want to write logging results to your Wandb account, use Add-ons -> Secrets to set `wandb_key` variable to your Wandb API key.","metadata":{}},{"cell_type":"code","source":"try:\n    from kaggle_secrets import UserSecretsClient\n    user_secrets = UserSecretsClient()\n    api_key = user_secrets.get_secret('wandb_key')\n    wandb.login(key=api_key)\n    anonymous = None\nexcept:\n    wandb.login(anonymous='must')\n    print('To use your W&B account,\\nGo to Add-ons -> Secrets and provide your \\\n           W&B access token. Use the Label name as WANDB. \\nGet your W&B access \\\n           token from here: https://wandb.ai/authorize')","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-25T15:56:39.510251Z","iopub.execute_input":"2025-11-25T15:56:39.510643Z","iopub.status.idle":"2025-11-25T15:56:41.407505Z","shell.execute_reply.started":"2025-11-25T15:56:39.510626Z","shell.execute_reply":"2025-11-25T15:56:41.406646Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## TPU detection","metadata":{}},{"cell_type":"code","source":"AUTO = tf.data.AUTOTUNE\n\n# Detect and initialize TPU\ntry:\n    tpu = tf.distribute.cluster_resolver.TPUClusterResolver(tpu=\"local\")  \n    tf.config.experimental_connect_to_cluster(tpu)\n    tf.tpu.experimental.initialize_tpu_system(tpu)\n    strategy = tf.distribute.TPUStrategy(tpu)\n    print(f\"Running on TPU: {tpu.master()}\")\nexcept Exception as e:\n    print(f\"Could not initialize TPU: {e}\")\n    strategy = tf.distribute.get_strategy()  # Fallback to default CPU/GPU\n    print(\"Running on CPU or single GPU\")\n\n# Print number of available TPU cores\nprint(\"Number of devices:\", strategy.num_replicas_in_sync, \"🚀\")","metadata":{"_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","trusted":true,"execution":{"iopub.status.busy":"2025-11-25T15:56:41.408068Z","iopub.execute_input":"2025-11-25T15:56:41.408463Z","iopub.status.idle":"2025-11-25T15:56:57.379734Z","shell.execute_reply.started":"2025-11-25T15:56:41.408446Z","shell.execute_reply":"2025-11-25T15:56:57.378692Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## Configuration","metadata":{}},{"cell_type":"code","source":"SEED = 42\n\nIMAGE_SIZE = [224, 224]\nEPOCHS = 10\nBATCH_SIZE = 32 * strategy.num_replicas_in_sync","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-25T15:56:57.380232Z","iopub.execute_input":"2025-11-25T15:56:57.380409Z","iopub.status.idle":"2025-11-25T15:56:57.383459Z","shell.execute_reply.started":"2025-11-25T15:56:57.380391Z","shell.execute_reply":"2025-11-25T15:56:57.382561Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"def set_seed(seed):\n    np.random.seed(seed)\n    random.seed(seed)\n    os.environ['PYTHONHASHSEED'] = str(seed)\n    tf.random.set_seed(seed)\n    \nset_seed(SEED)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-25T15:56:57.384045Z","iopub.execute_input":"2025-11-25T15:56:57.384208Z","iopub.status.idle":"2025-11-25T15:56:57.395076Z","shell.execute_reply.started":"2025-11-25T15:56:57.384192Z","shell.execute_reply":"2025-11-25T15:56:57.394357Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## Data access and classes","metadata":{}},{"cell_type":"markdown","source":"TPUs read data directly from Google Cloud Storage (GCS), so we need to copy the dataset to a GCS bucket co-located with the TPU. To do that, pass the name of a specific dataset to the get_gcs_path function. The name of the dataset is the name of the directory it is mounted in. ","metadata":{}},{"cell_type":"code","source":"DS_PATH = '/kaggle/input/tpu-getting-started'\nDS_PATH_EXT = '/kaggle/input/tf-flower-photo-tfrec'\n\nDATA_PATH_SELECT = { # available image sizes\n    192: DS_PATH + '/tfrecords-jpeg-192x192',\n    224: DS_PATH + '/tfrecords-jpeg-224x224',\n    331: DS_PATH + '/tfrecords-jpeg-331x331',\n    512: DS_PATH + '/tfrecords-jpeg-512x512'\n}\nDATA_PATH = DATA_PATH_SELECT[IMAGE_SIZE[0]]\n\n# External data\nDATA_PATH_SELECT_EXT = {\n    192: '/tfrecords-jpeg-192x192',\n    224: '/tfrecords-jpeg-224x224',\n    331: '/tfrecords-jpeg-331x331',\n    512: '/tfrecords-jpeg-512x512'\n}\nDATA_PATH_EXT = DATA_PATH_SELECT_EXT[IMAGE_SIZE[0]]\n\nIMAGENET_FILES = tf.io.gfile.glob(DS_PATH_EXT + '/imagenet' + DATA_PATH_EXT + '/*.tfrec')\nINATURELIST_FILES = tf.io.gfile.glob(DS_PATH_EXT + '/inaturalist' + DATA_PATH_EXT + '/*.tfrec')\nOPENIMAGE_FILES = tf.io.gfile.glob(DS_PATH_EXT + '/openimage' + DATA_PATH_EXT + '/*.tfrec')\nOXFORD_FILES = tf.io.gfile.glob(DS_PATH_EXT + '/oxford_102' + DATA_PATH_EXT + '/*.tfrec')\nTENSORFLOW_FILES = tf.io.gfile.glob(DS_PATH_EXT + '/tf_flowers' + DATA_PATH_EXT + '/*.tfrec')\n\nADDITIONAL_TRAINING_FILENAMES = IMAGENET_FILES + INATURELIST_FILES + OPENIMAGE_FILES + OXFORD_FILES + TENSORFLOW_FILES  \n\nCLASSES = ['pink primrose',    'hard-leaved pocket orchid', 'canterbury bells', 'sweet pea',     'wild geranium',     'tiger lily',           'moon orchid',              'bird of paradise', 'monkshood',        'globe thistle',         # 00 - 09\n           'snapdragon',       \"colt's foot\",               'king protea',      'spear thistle', 'yellow iris',       'globe-flower',         'purple coneflower',        'peruvian lily',    'balloon flower',   'giant white arum lily', # 10 - 19\n           'fire lily',        'pincushion flower',         'fritillary',       'red ginger',    'grape hyacinth',    'corn poppy',           'prince of wales feathers', 'stemless gentian', 'artichoke',        'sweet william',         # 20 - 29\n           'carnation',        'garden phlox',              'love in the mist', 'cosmos',        'alpine sea holly',  'ruby-lipped cattleya', 'cape flower',              'great masterwort', 'siam tulip',       'lenten rose',           # 30 - 39\n           'barberton daisy',  'daffodil',                  'sword lily',       'poinsettia',    'bolero deep blue',  'wallflower',           'marigold',                 'buttercup',        'daisy',            'common dandelion',      # 40 - 49\n           'petunia',          'wild pansy',                'primula',          'sunflower',     'lilac hibiscus',    'bishop of llandaff',   'gaura',                    'geranium',         'orange dahlia',    'pink-yellow dahlia',    # 50 - 59\n           'cautleya spicata', 'japanese anemone',          'black-eyed susan', 'silverbush',    'californian poppy', 'osteospermum',         'spring crocus',            'iris',             'windflower',       'tree poppy',            # 60 - 69\n           'gazania',          'azalea',                    'water lily',       'rose',          'thorn apple',       'morning glory',        'passion flower',           'lotus',            'toad lily',        'anthurium',             # 70 - 79\n           'frangipani',       'clematis',                  'hibiscus',         'columbine',     'desert-rose',       'tree mallow',          'magnolia',                 'cyclamen ',        'watercress',       'canna lily',            # 80 - 89\n           'hippeastrum ',     'bee balm',                  'pink quill',       'foxglove',      'bougainvillea',     'camellia',             'mallow',                   'mexican petunia',  'bromelia',         'blanket flower',        # 90 - 99\n           'trumpet creeper',  'blackberry lily',           'common tulip',     'wild rose']                          # 100 - 102\n\nTRAINING_FILENAMES = tf.io.gfile.glob(DATA_PATH + '/train/*.tfrec')\nVALIDATION_FILENAMES = tf.io.gfile.glob(DATA_PATH + '/val/*.tfrec')\nTEST_FILENAMES = tf.io.gfile.glob(DATA_PATH + '/test/*.tfrec') # predictions on this dataset should be submitted for the competition \n\nTRAINING_FILENAMES = TRAINING_FILENAMES + ADDITIONAL_TRAINING_FILENAMES","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-25T15:56:57.395664Z","iopub.execute_input":"2025-11-25T15:56:57.395838Z","iopub.status.idle":"2025-11-25T15:56:57.526464Z","shell.execute_reply.started":"2025-11-25T15:56:57.395823Z","shell.execute_reply":"2025-11-25T15:56:57.525423Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## Visualization functions","metadata":{}},{"cell_type":"markdown","source":"A set of functions to visualize data.","metadata":{}},{"cell_type":"code","source":"# numpy and matplotlib defaults\nnp.set_printoptions(threshold=15, linewidth=80)\n\ndef batch_to_numpy_images_and_labels(data):\n    images, labels = data\n    numpy_images = images.numpy()\n    numpy_labels = labels.numpy()\n    if numpy_labels.dtype == object: # binary string in this case, these are image ID strings\n        numpy_labels = [None for _ in enumerate(numpy_images)]\n    # If no labels, only image IDs, return None for labels (this is the case for test data)\n    return numpy_images, numpy_labels\n\ndef title_from_label_and_target(label, correct_label):\n    if correct_label is None:\n        return CLASSES[label], True\n    correct = (label == correct_label)\n    return \"{} [{}{}{}]\".format(CLASSES[label], 'OK' if correct else 'NO', u\"\\u2192\" if not correct else '',\n                                CLASSES[correct_label] if not correct else ''), correct\n\ndef display_one_flower(image, title, subplot, red=False, titlesize=16):\n    plt.subplot(*subplot)\n    plt.axis('off')\n    plt.imshow(image)\n    if len(title) > 0:\n        plt.title(title, fontsize=int(titlesize) if not red else int(titlesize/1.2), color='red' if red else 'black', fontdict={'verticalalignment':'center'}, pad=int(titlesize/1.5))\n    return (subplot[0], subplot[1], subplot[2]+1)\n    \ndef display_batch_of_images(databatch, predictions=None):\n    \"\"\"This will work with:\n    display_batch_of_images(images)\n    display_batch_of_images(images, predictions)\n    display_batch_of_images((images, labels))\n    display_batch_of_images((images, labels), predictions)\n    \"\"\"\n    # data\n    images, labels = batch_to_numpy_images_and_labels(databatch)\n    if labels is None:\n        labels = [None for _ in enumerate(images)]\n        \n    # auto-squaring: this will drop data that does not fit into square or square-ish rectangle\n    rows = int(math.sqrt(len(images)))\n    cols = len(images)//rows\n        \n    # size and spacing\n    FIGSIZE = 13.0\n    SPACING = 0.1\n    subplot=(rows,cols,1)\n    if rows < cols:\n        plt.figure(figsize=(FIGSIZE,FIGSIZE/cols*rows))\n    else:\n        plt.figure(figsize=(FIGSIZE/rows*cols,FIGSIZE))\n    \n    # display\n    for i, (image, label) in enumerate(zip(images[:rows*cols], labels[:rows*cols])):\n        title = '' if label is None else CLASSES[label]\n        correct = True\n        if predictions is not None:\n            title, correct = title_from_label_and_target(predictions[i], label)\n        dynamic_titlesize = FIGSIZE*SPACING/max(rows,cols)*40+3 # magic formula tested to work from 1x1 to 10x10 images\n        subplot = display_one_flower(image, title, subplot, not correct, titlesize=dynamic_titlesize)\n    \n    #layout\n    plt.tight_layout()\n    if label is None and predictions is None:\n        plt.subplots_adjust(wspace=0, hspace=0)\n    else:\n        plt.subplots_adjust(wspace=SPACING, hspace=SPACING)\n    plt.show()\n\ndef display_confusion_matrix(cmat, score, precision, recall):\n    plt.figure(figsize=(15,15))\n    ax = plt.gca()\n    ax.matshow(cmat, cmap='Reds')\n    ax.set_xticks(range(len(CLASSES)))\n    ax.set_xticklabels(CLASSES, fontdict={'fontsize': 7})\n    plt.setp(ax.get_xticklabels(), rotation=45, ha=\"left\", rotation_mode=\"anchor\")\n    ax.set_yticks(range(len(CLASSES)))\n    ax.set_yticklabels(CLASSES, fontdict={'fontsize': 7})\n    plt.setp(ax.get_yticklabels(), rotation=45, ha=\"right\", rotation_mode=\"anchor\")\n    titlestring = \"\"\n    if score is not None:\n        titlestring += 'f1 = {:.3f} '.format(score)\n    if precision is not None:\n        titlestring += '\\nprecision = {:.3f} '.format(precision)\n    if recall is not None:\n        titlestring += '\\nrecall = {:.3f} '.format(recall)\n    if len(titlestring) > 0:\n        ax.text(101, 1, titlestring, fontdict={'fontsize': 18, 'horizontalalignment':'right', 'verticalalignment':'top', 'color':'#804040'})\n    plt.show()\n    \ndef display_training_curves(training, validation, title, subplot):\n    if subplot%10==1: # set up the subplots on the first call\n        plt.subplots(figsize=(10,10), facecolor='#F0F0F0')\n        plt.tight_layout()\n    ax = plt.subplot(subplot)\n    ax.set_facecolor('#F8F8F8')\n    ax.plot(training)\n    ax.plot(validation)\n    ax.set_title('model '+ title)\n    ax.set_ylabel(title)\n    #ax.set_ylim(0.28,1.05)\n    ax.set_xlabel('epoch')\n    ax.legend(['train', 'valid.'])","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-25T15:56:57.526872Z","iopub.execute_input":"2025-11-25T15:56:57.527035Z","iopub.status.idle":"2025-11-25T15:56:57.537356Z","shell.execute_reply.started":"2025-11-25T15:56:57.527020Z","shell.execute_reply":"2025-11-25T15:56:57.536587Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# Random erasing (blockout) augmentation","metadata":{}},{"cell_type":"code","source":"# https://www.kaggle.com/tusharkendre/tpu-flowers\ndef random_erasing(img, sl=0.1, sh=0.2, rl=0.4, p=0.3):\n    h = tf.shape(img)[0]\n    w = tf.shape(img)[1]\n    c = tf.shape(img)[2]\n    origin_area = tf.cast(h*w, tf.float32)\n\n    e_size_l = tf.cast(tf.round(tf.sqrt(origin_area * sl * rl)), tf.int32)\n    e_size_h = tf.cast(tf.round(tf.sqrt(origin_area * sh / rl)), tf.int32)\n\n    e_height_h = tf.minimum(e_size_h, h)\n    e_width_h = tf.minimum(e_size_h, w)\n\n    erase_height = tf.random.uniform(shape=[], minval=e_size_l, maxval=e_height_h, dtype=tf.int32)\n    erase_width = tf.random.uniform(shape=[], minval=e_size_l, maxval=e_width_h, dtype=tf.int32)\n\n    erase_area = tf.zeros(shape=[erase_height, erase_width, c])\n    erase_area = tf.cast(erase_area, tf.uint8)\n\n    pad_h = h - erase_height\n    pad_top = tf.random.uniform(shape=[], minval=0, maxval=pad_h, dtype=tf.int32)\n    pad_bottom = pad_h - pad_top\n\n    pad_w = w - erase_width\n    pad_left = tf.random.uniform(shape=[], minval=0, maxval=pad_w, dtype=tf.int32)\n    pad_right = pad_w - pad_left\n\n    erase_mask = tf.pad([erase_area], [[0,0],[pad_top, pad_bottom], [pad_left, pad_right], [0,0]], constant_values=1)\n    erase_mask = tf.squeeze(erase_mask, axis=0)\n    erased_img = tf.multiply(tf.cast(img,tf.float32), tf.cast(erase_mask, tf.float32))\n\n    return tf.cond(tf.random.uniform([], 0, 1) > p, lambda: tf.cast(img, img.dtype), lambda:  tf.cast(erased_img, img.dtype))","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-25T15:56:57.537959Z","iopub.execute_input":"2025-11-25T15:56:57.538120Z","iopub.status.idle":"2025-11-25T15:56:57.555260Z","shell.execute_reply.started":"2025-11-25T15:56:57.538105Z","shell.execute_reply":"2025-11-25T15:56:57.554491Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## Dataset functions","metadata":{}},{"cell_type":"code","source":"def decode_image(image_data):\n    image = tf.image.decode_jpeg(image_data, channels=3)\n    image = tf.cast(image, tf.float32) / 255.0  # convert image to floats in [0, 1] range\n    image = tf.reshape(image, [*IMAGE_SIZE, 3]) # explicit size needed for TPU\n    return image\n\ndef onehot(image,label):\n    return image,tf.one_hot(label, len(CLASSES))\n\ndef read_labeled_tfrecord(example):\n    LABELED_TFREC_FORMAT = {\n        \"image\": tf.io.FixedLenFeature([], tf.string), # tf.string means bytestring\n        \"class\": tf.io.FixedLenFeature([], tf.int64),  # shape [] means single element\n    }\n    example = tf.io.parse_single_example(example, LABELED_TFREC_FORMAT)\n    image = decode_image(example['image'])\n    label = tf.cast(example['class'], tf.int32)\n    return image, label # returns a dataset of (image, label) pairs\n\ndef read_unlabeled_tfrecord(example):\n    UNLABELED_TFREC_FORMAT = {\n        \"image\": tf.io.FixedLenFeature([], tf.string), # tf.string means bytestring\n        \"id\": tf.io.FixedLenFeature([], tf.string),  # shape [] means single element\n        # class is missing, this competitions's challenge is to predict flower classes for the test dataset\n    }\n    example = tf.io.parse_single_example(example, UNLABELED_TFREC_FORMAT)\n    image = decode_image(example['image'])\n    idnum = example['id']\n    return image, idnum # returns a dataset of image(s)\n\ndef load_dataset(filenames, labeled=True, ordered=False):\n    # Read from TFRecords. For optimal performance, reading from multiple files at once and\n    # disregarding data order. Order does not matter since we will be shuffling the data anyway.\n\n    ignore_order = tf.data.Options()\n    if not ordered:\n        ignore_order.experimental_deterministic = False # disable order, increase speed\n\n    dataset = tf.data.TFRecordDataset(filenames, num_parallel_reads=AUTO) # automatically interleaves reads from multiple files\n    dataset = dataset.with_options(ignore_order) # uses data as soon as it streams in, rather than in its original order\n    dataset = dataset.map(read_labeled_tfrecord if labeled else read_unlabeled_tfrecord, num_parallel_calls=AUTO)\n    # returns a dataset of (image, label) pairs if labeled=True or (image, id) pairs if labeled=False\n    return dataset\n\ndef data_augment(image, label):\n    image = tf.image.random_flip_left_right(image)\n    image = random_erasing(image)\n    return image, label\n\ndef data_hflip(image, idnum):\n    image = tf.image.flip_left_right(image)\n    return image, idnum\n\ndef get_training_dataset(do_onehot=False):\n    dataset = load_dataset(TRAINING_FILENAMES, labeled=True)\n    dataset = dataset.map(data_augment, num_parallel_calls=AUTO)\n    if do_onehot:\n        dataset = dataset.map(onehot, num_parallel_calls=AUTO)\n    dataset = dataset.repeat()\n    dataset = dataset.shuffle(512)#, reshuffle_each_iteration=True)\n    dataset = dataset.batch(BATCH_SIZE)\n    dataset = dataset.prefetch(AUTO) # prefetch next batch while training (autotune prefetch buffer size)\n    return dataset\n\ndef get_validation_dataset(ordered=False, do_onehot=False, repeat=False):\n    dataset = load_dataset(VALIDATION_FILENAMES, labeled=True, ordered=ordered)\n    if do_onehot:\n        dataset = dataset.map(onehot, num_parallel_calls=AUTO)\n    if repeat:\n        dataset = dataset.repeat()\n    dataset = dataset.batch(BATCH_SIZE)\n    dataset = dataset.prefetch(AUTO) # prefetch next batch while training (autotune prefetch buffer size)\n    return dataset\n\ndef get_test_dataset(ordered=False, augmented=False):\n    dataset = load_dataset(TEST_FILENAMES, labeled=False, ordered=ordered)\n    if augmented:\n        dataset = dataset.map(data_hflip, num_parallel_calls=AUTO)\n    dataset = dataset.batch(BATCH_SIZE)\n    dataset = dataset.prefetch(AUTO) # prefetch next batch while training (autotune prefetch buffer size)\n    return dataset\n\ndef count_data_items(filenames):\n    # the number of data items is written in the name of the .tfrec files, i.e. flowers00-230.tfrec = 230 data items\n    n = [int(re.compile(r\"-([0-9]*)\\.\").search(filename).group(1)) for filename in filenames]\n    return np.sum(n)\n\nNUM_TRAINING_IMAGES = count_data_items(TRAINING_FILENAMES)\nNUM_VALIDATION_IMAGES = count_data_items(VALIDATION_FILENAMES)\nNUM_TEST_IMAGES = count_data_items(TEST_FILENAMES)\nSTEPS_PER_EPOCH = NUM_TRAINING_IMAGES // BATCH_SIZE\nVALIDATION_STEPS = NUM_VALIDATION_IMAGES // BATCH_SIZE\nprint(f'Dataset: {NUM_TRAINING_IMAGES} training images, {NUM_VALIDATION_IMAGES} validation images, {NUM_TEST_IMAGES} unlabeled test images')","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-25T15:56:57.555875Z","iopub.execute_input":"2025-11-25T15:56:57.556048Z","iopub.status.idle":"2025-11-25T15:56:57.572195Z","shell.execute_reply.started":"2025-11-25T15:56:57.556033Z","shell.execute_reply":"2025-11-25T15:56:57.571512Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## Dataset visualizations","metadata":{}},{"cell_type":"code","source":"# data dump\nprint(\"Training data shapes:\")\nfor image, label in get_training_dataset().take(3):\n    print(image.numpy().shape, label.numpy().shape)\nprint(\"Training data label examples:\", label.numpy())\nprint(\"Validation data shapes:\")\nfor image, label in get_validation_dataset().take(3):\n    print(image.numpy().shape, label.numpy().shape)\nprint(\"Validation data label examples:\", label.numpy())\nprint(\"Test data shapes:\")\nfor image, idnum in get_test_dataset().take(3):\n    print(image.numpy().shape, idnum.numpy().shape)\nprint(\"Test data IDs:\", idnum.numpy().astype('U')) # U=unicode string","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-25T15:56:57.572595Z","iopub.execute_input":"2025-11-25T15:56:57.572763Z","iopub.status.idle":"2025-11-25T15:56:59.434805Z","shell.execute_reply.started":"2025-11-25T15:56:57.572736Z","shell.execute_reply":"2025-11-25T15:56:59.433779Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Peek at training data\ntraining_dataset = get_training_dataset()\ntraining_dataset = training_dataset.unbatch().batch(20)\ntrain_batch = iter(training_dataset)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-25T15:56:59.435287Z","iopub.execute_input":"2025-11-25T15:56:59.435462Z","iopub.status.idle":"2025-11-25T15:56:59.530287Z","shell.execute_reply.started":"2025-11-25T15:56:59.435446Z","shell.execute_reply":"2025-11-25T15:56:59.529596Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# run this cell again for next set of images\ndisplay_batch_of_images(next(train_batch))","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-25T15:56:59.530915Z","iopub.execute_input":"2025-11-25T15:56:59.531081Z","iopub.status.idle":"2025-11-25T15:57:01.034038Z","shell.execute_reply.started":"2025-11-25T15:56:59.531065Z","shell.execute_reply":"2025-11-25T15:57:01.033001Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# peer at test data\ntest_dataset = get_test_dataset()\ntest_dataset = test_dataset.unbatch().batch(20)\ntest_batch = iter(test_dataset)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-25T15:57:01.034600Z","iopub.execute_input":"2025-11-25T15:57:01.034787Z","iopub.status.idle":"2025-11-25T15:57:01.080196Z","shell.execute_reply.started":"2025-11-25T15:57:01.034771Z","shell.execute_reply":"2025-11-25T15:57:01.079220Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# run this cell again for next set of images\ndisplay_batch_of_images(next(test_batch))","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-25T15:57:01.080647Z","iopub.execute_input":"2025-11-25T15:57:01.080827Z","iopub.status.idle":"2025-11-25T15:57:02.091694Z","shell.execute_reply.started":"2025-11-25T15:57:01.080812Z","shell.execute_reply":"2025-11-25T15:57:02.090541Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## Define and train the model","metadata":{}},{"cell_type":"markdown","source":"### Custom LR scheduler","metadata":{}},{"cell_type":"code","source":"# Learning rate schedule for TPU, GPU and CPU.\n# Using an LR ramp up because fine-tuning a pre-trained model.\n# Starting with a high LR would break the pre-trained weights.\ndef get_lr_callback(plot_schedule=False):\n    LR_START = 0.00001\n    LR_MAX = 0.00005 * strategy.num_replicas_in_sync\n    LR_MIN = 0.00001\n    LR_RAMPUP_EPOCHS = 2\n    LR_SUSTAIN_EPOCHS = 0\n    LR_EXP_DECAY = .8\n\n    def lrfn(epoch):\n        if epoch < LR_RAMPUP_EPOCHS:\n            lr = (LR_MAX - LR_START) / LR_RAMPUP_EPOCHS * epoch + LR_START\n        elif epoch < LR_RAMPUP_EPOCHS + LR_SUSTAIN_EPOCHS:\n            lr = LR_MAX\n        else:\n            lr = (LR_MAX - LR_MIN) * LR_EXP_DECAY**(epoch - LR_RAMPUP_EPOCHS - LR_SUSTAIN_EPOCHS) + LR_MIN\n        return lr\n    \n    if plot_schedule:\n        rng = [i for i in range(25 if EPOCHS < 25 else EPOCHS)]\n        y = [lrfn(x) for x in rng]\n        plt.plot(rng, y)\n\n    return callbacks.LearningRateScheduler(lrfn, verbose=0)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-25T15:57:02.092421Z","iopub.execute_input":"2025-11-25T15:57:02.092584Z","iopub.status.idle":"2025-11-25T15:57:02.096704Z","shell.execute_reply.started":"2025-11-25T15:57:02.092568Z","shell.execute_reply":"2025-11-25T15:57:02.096032Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"def load_and_fit_model(print_summary=False):\n    with strategy.scope():\n        base_model = SwinTransformerLarge224(\n                include_top=False,\n                weights='imagenet')\n        \n        base_model.trainable = False\n        \n        inputs = tf.keras.Input(shape=(*IMAGE_SIZE, 3))\n        x = base_model(inputs, training=False)\n        x = L.GlobalAveragePooling2D()(x)\n        x = L.Dense(256)(x)\n        outputs = L.Dense(len(CLASSES), activation='softmax')(x)\n        \n        model = tf.keras.Model(inputs, outputs)\n        model.compile(\n            optimizer='adam',\n            loss = 'categorical_crossentropy',\n            metrics=[F1Score(average='macro')]\n        )\n\n    if print_summary:\n        model.summary()\n        \n    os.makedirs('checkpoints', exist_ok=True)\n    \n    lr_callback = get_lr_callback()\n    chk_callback = callbacks.ModelCheckpoint(f'checkpoints/swin_large_best.weights.h5',\n                     save_weights_only=True, monitor='val_f1_score',\n                     mode='max', save_best_only=True, verbose=2)\n    \n    wandb.init(project='flower-classification-tpu-public', \n               name='swin_large_224',\n               job_type='train', \n               reinit=True)\n    log_callback = WandbMetricsLogger()\n\n    _ = model.fit(get_training_dataset(do_onehot=True), \n                  steps_per_epoch=STEPS_PER_EPOCH,\n                  epochs=EPOCHS, \n                  validation_data=get_validation_dataset(do_onehot=True, repeat=True),\n                  validation_steps=VALIDATION_STEPS,\n                  callbacks=[lr_callback, chk_callback, log_callback],\n                  verbose=1)\n    model.load_weights(f'checkpoints/swin_large_best.weights.h5')\n    wandb.finish()\n    return model","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-25T15:57:02.097148Z","iopub.execute_input":"2025-11-25T15:57:02.097297Z","iopub.status.idle":"2025-11-25T15:57:02.113182Z","shell.execute_reply.started":"2025-11-25T15:57:02.097283Z","shell.execute_reply":"2025-11-25T15:57:02.112472Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"### Train model","metadata":{}},{"cell_type":"code","source":"model = load_and_fit_model()","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-11-25T15:57:02.113539Z","iopub.execute_input":"2025-11-25T15:57:02.113683Z","execution_failed":"2025-11-25T16:07:40.819Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## Plot confusion matrix and predict on test dataset","metadata":{}},{"cell_type":"code","source":"def predict(dataset, model):\n    print('Calculating predictions...')\n    images_ds = dataset.map(lambda image, idnum: image)\n    preds = model.predict(images_ds,verbose=0)\n    return preds","metadata":{"trusted":true,"execution":{"execution_failed":"2025-11-25T16:07:40.819Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"Let's plot confusion matrix to evaluate our model ensemble accuracy.","metadata":{}},{"cell_type":"code","source":"valid_ds = get_validation_dataset(ordered=True)\ncm_predictions = predict(valid_ds, model)\ncm_predictions = np.argmax(cm_predictions, axis=1)\n\nlabels_ds = valid_ds.map(lambda image, label: label).unbatch()\ncm_correct_labels = next(iter(labels_ds.batch(NUM_VALIDATION_IMAGES))).numpy() # get everything as one batch\n\ncmat = confusion_matrix(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)))\nscore = f1_score(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)), average='macro')\nprecision = precision_score(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)), average='macro')\nrecall = recall_score(cm_correct_labels, cm_predictions, labels=range(len(CLASSES)), average='macro')\n#cmat = (cmat.T / cmat.sum(axis=1)).T # normalized\ndisplay_confusion_matrix(cmat, score, precision, recall)","metadata":{"trusted":true,"execution":{"execution_failed":"2025-11-25T16:07:40.819Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"Now we predict on the test dataset and write results to the submission file (submission.csv)","metadata":{}},{"cell_type":"code","source":"test_ds_1 = get_test_dataset(ordered=True, augmented=False)\npreds_1 = predict(test_ds_1, model)\n\ntest_ds_2 = get_test_dataset(ordered=True, augmented=True)\npreds_2 = predict(test_ds_2, model)\n\npreds = (preds_1 + preds_2) / 2\npreds = np.argmax(preds, axis=1)\n\nprint('Generating submission file...')\ntest_ids_ds = test_ds_1.map(lambda image, idnum: idnum).unbatch()\ntest_ids = next(iter(test_ids_ds.batch(NUM_TEST_IMAGES))).numpy().astype('U') # all in one batch\n                 \nsub_df = pd.DataFrame({'id': test_ids, 'label': preds})\nsub_df.to_csv('submission.csv', index=False)","metadata":{"trusted":true,"execution":{"execution_failed":"2025-11-25T16:07:40.820Z"}},"outputs":[],"execution_count":null}]}