{"cells":[{"metadata":{},"cell_type":"markdown","source":"You can get the same results using Tensorflow or PyTorch, don't take my word for it, see it for yourself.\n\nTo make sure each model has a same input pipeline, I used tf.data.dataset to fetch data for the Pytorch model as well, bit sacrilegious I admit.\nThe only different between to two implementations, is the external Efficientnet libary.\n\nForked from: @cdeotte's https://www.kaggle.com/cdeotte/triple-stratified-kfold-with-tfrecords","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"!pip install efficientnet_pytorch >> /dev/null","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","trusted":true},"cell_type":"code","source":"import os, random, re, math, time\nrandom.seed(a=42)\nos.environ[\"CUDA_VISIBLE_DEVICES\"] = \"0\"\nimport numpy as np\nimport pandas as pd\nfrom sklearn.model_selection import KFold\nfrom sklearn.metrics import roc_auc_score\n\nfrom glob import glob\nfrom tqdm import tqdm\n\nimport tensorflow as tf\nimport tensorflow.keras.backend as K\n\nfrom tensorflow.keras.mixed_precision import experimental as mixed_precision\n\ngpus = tf.config.experimental.list_physical_devices('GPU')\ntry:\n    for gpu in gpus:\n        tf.config.experimental.set_memory_growth(gpu, True)\nexcept RuntimeError as e:\n    print(e)\n\nCFG = dict(\n    batch_size        =  32,\n    read_size         = 256, \n    crop_size         = 235, \n    net_size          = 224, \n    \n    LR                =   1e-4,\n    epochs            =  30,\n    \n    rot               = 180.0,\n    shr               =   2.0,\n    hzoom             =   8.0,\n    wzoom             =   8.0,\n    hshift            =   8.0,\n    wshift            =   8.0,\n    tta_steps = 15,\n    \n    es_patience = 4,\n    \n)\n\n\nGCS_PATH = f'../input//melanoma-{CFG[\"read_size\"]}x{CFG[\"read_size\"]}'\nfiles_train = np.sort(np.array(tf.io.gfile.glob(GCS_PATH + '/train*.tfrec')))\nfiles_test  = np.sort(np.array(tf.io.gfile.glob(GCS_PATH + '/test*.tfrec')))\n\nmalig_files = sorted(glob(f'../input/malignant-v2-{CFG[\"read_size\"]}x{CFG[\"read_size\"]}/*.tfrec'))\nmalig_files = malig_files[15:]\n\nfiles_train = np.concatenate([files_train, malig_files])\n    ","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"def get_mat(rotation, shear, height_zoom, width_zoom, height_shift, width_shift):\n    # returns 3x3 transformmatrix which transforms indicies\n        \n    # CONVERT DEGREES TO RADIANS\n    rotation = math.pi * rotation / 180.\n    shear    = math.pi * shear    / 180.\n\n    def get_3x3_mat(lst):\n        return tf.reshape(tf.concat([lst],axis=0), [3,3])\n    \n    # ROTATION MATRIX\n    c1   = tf.math.cos(rotation)\n    s1   = tf.math.sin(rotation)\n    one  = tf.constant([1],dtype='float32')\n    zero = tf.constant([0],dtype='float32')\n    \n    rotation_matrix = get_3x3_mat([c1,   s1,   zero, \n                                   -s1,  c1,   zero, \n                                   zero, zero, one])    \n    # SHEAR MATRIX\n    c2 = tf.math.cos(shear)\n    s2 = tf.math.sin(shear)    \n    \n    shear_matrix = get_3x3_mat([one,  s2,   zero, \n                                zero, c2,   zero, \n                                zero, zero, one])        \n    # ZOOM MATRIX\n    zoom_matrix = get_3x3_mat([one/height_zoom, zero,           zero, \n                               zero,            one/width_zoom, zero, \n                               zero,            zero,           one])    \n    # SHIFT MATRIX\n    shift_matrix = get_3x3_mat([one,  zero, height_shift, \n                                zero, one,  width_shift, \n                                zero, zero, one])\n    \n    return K.dot(K.dot(rotation_matrix, shear_matrix), \n                 K.dot(zoom_matrix,     shift_matrix))\n\n\ndef transform(image, cfg):\n    DIM = cfg[\"read_size\"]\n    XDIM = DIM%2 #fix for size 331\n    \n    rot = cfg['rot'] * tf.random.normal([1], dtype='float32')\n    shr = cfg['shr'] * tf.random.normal([1], dtype='float32') \n    h_zoom = 1.0 + tf.random.normal([1], dtype='float32') / cfg['hzoom']\n    w_zoom = 1.0 + tf.random.normal([1], dtype='float32') / cfg['wzoom']\n    h_shift = cfg['hshift'] * tf.random.normal([1], dtype='float32') \n    w_shift = cfg['wshift'] * tf.random.normal([1], dtype='float32') \n\n    # GET TRANSFORMATION MATRIX\n    m = get_mat(rot,shr,h_zoom,w_zoom,h_shift,w_shift) \n\n    # LIST DESTINATION PIXEL INDICES\n    x   = tf.repeat(tf.range(DIM//2, -DIM//2,-1), DIM)\n    y   = tf.tile(tf.range(-DIM//2, DIM//2), [DIM])\n    z   = tf.ones([DIM*DIM], dtype='int32')\n    idx = tf.stack( [x,y,z] )\n    \n    # ROTATE DESTINATION PIXELS ONTO ORIGIN PIXELS\n    idx2 = K.dot(m, tf.cast(idx, dtype='float32'))\n    idx2 = K.cast(idx2, dtype='int32')\n    idx2 = K.clip(idx2, -DIM//2+XDIM+1, DIM//2)\n    \n    # FIND ORIGIN PIXEL VALUES           \n    idx3 = tf.stack([DIM//2-idx2[0,], DIM//2-1+idx2[1,]])\n    d    = tf.gather_nd(image, tf.transpose(idx3))\n        \n    return tf.reshape(d,[DIM, DIM,3])\n\ndef read_labeled_tfrecord(example):\n    tfrec_format = {\n        'image'                        : tf.io.FixedLenFeature([], tf.string),\n        'image_name'                   : tf.io.FixedLenFeature([], tf.string),\n        'patient_id'                   : tf.io.FixedLenFeature([], tf.int64),\n        'sex'                          : tf.io.FixedLenFeature([], tf.int64),\n        'age_approx'                   : tf.io.FixedLenFeature([], tf.int64),\n        'anatom_site_general_challenge': tf.io.FixedLenFeature([], tf.int64),\n        'diagnosis'                    : tf.io.FixedLenFeature([], tf.int64),\n        'target'                       : tf.io.FixedLenFeature([], tf.int64)\n    }           \n    example = tf.io.parse_single_example(example, tfrec_format)\n    return example['image'], example['target'], example['image_name']\n\n\ndef read_unlabeled_tfrecord(example):\n    tfrec_format = {\n        'image'                        : tf.io.FixedLenFeature([], tf.string),\n        'image_name'                   : tf.io.FixedLenFeature([], tf.string),\n    }\n    example = tf.io.parse_single_example(example, tfrec_format)\n    return example['image'], 0, example['image_name']\n\n \ndef prepare_image(img, cfg=None, augment=True):    \n    img = tf.image.decode_jpeg(img, channels=3)\n    img = tf.image.resize(img, [cfg['read_size'], cfg['read_size']])\n    img = tf.cast(img, tf.float32) / 255.0\n    \n    if augment:\n        img = transform(img, cfg)\n        img = tf.image.random_crop(img, [cfg['crop_size'], cfg['crop_size'], 3])\n        img = tf.image.random_flip_left_right(img)\n        img = tf.image.random_hue(img, 0.01)\n        img = tf.image.random_saturation(img, 0.7, 1.3)\n        img = tf.image.random_contrast(img, 0.8, 1.2)\n        img = tf.image.random_brightness(img, 0.1)\n\n    else:\n        img = tf.image.central_crop(img, cfg['crop_size'] / cfg['read_size'])\n                                   \n    img = tf.image.resize(img, [cfg['net_size'], cfg['net_size']])\n    img = tf.reshape(img, [cfg['net_size'], cfg['net_size'], 3])\n    return img\n\ndef get_dataset(files, cfg, augment = False, shuffle = False, \n                labeled=True):\n    AUTO     = tf.data.experimental.AUTOTUNE\n    ds = tf.data.TFRecordDataset(files, num_parallel_reads=AUTO)\n    ds = ds.cache()\n    if shuffle: \n        ds = ds.shuffle(1024*8)\n        opt = tf.data.Options()\n        opt.experimental_deterministic = False\n        ds = ds.with_options(opt)\n        \n    if labeled: \n        ds = ds.map(read_labeled_tfrecord, num_parallel_calls=AUTO)\n    else:\n        ds = ds.map(lambda example: read_unlabeled_tfrecord(example), \n                    num_parallel_calls=AUTO)      \n    \n    ds = ds.map(lambda img, label, fn: (prepare_image(img, augment=augment, cfg=cfg), \n                                               label, fn), \n                num_parallel_calls=AUTO)\n    \n    ds = ds.batch(cfg['batch_size'])\n    ds = ds.prefetch(AUTO)\n    return ds","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"#pytorch\nfrom efficientnet_pytorch import EfficientNet\nimport torch\n\ndef get_model():\n    model = EfficientNet.from_pretrained('efficientnet-b0', num_classes=1)\n    model.to(device)\n    return model\n\ndef get_opt_loss_fn(model):\n    optimizer = torch.optim.Adam(model.parameters(), lr=CFG[\"LR\"])\n    loss_object = torch.nn.BCEWithLogitsLoss().to(device)\n    return optimizer, loss_object\n\ndef get_train_fn():\n    def train_step(images, labels, model, optimizer, loss_object):\n        #sacrilegious conversion\n        images = torch.from_numpy(images.numpy()).permute(0,3,1,2).to(device)\n        labels = torch.from_numpy(labels.numpy()).to(device).float()\n        \n        if not model.training:\n            model.train()\n        predictions = model.forward(images)\n        loss = loss_object(predictions[:,0], labels)    \n        loss.backward()\n        optimizer.step()\n        optimizer.zero_grad() \n        return loss.cpu().detach()\n    \n    return train_step\n\n\ndef test_step(images, labels):\n    #sacrilegious conversion\n    images = torch.from_numpy(images.numpy()).permute(0,3,1,2).to(device)\n    labels = torch.from_numpy(labels.numpy()).to(device).float()\n    \n    if model.training:\n        model.eval()\n    with torch.no_grad():\n        predictions = model.forward(images)\n    loss = loss_object(predictions[:,0], labels)\n    return loss.cpu().detach()\n\n\ndef pred_step(images):\n    #sacrilegious conversion\n    images = torch.from_numpy(images.numpy()).permute(0,3,1,2).to(device)\n    \n    if model.training:\n        model.eval()\n    with torch.no_grad():\n        return torch.sigmoid(model(images)).cpu().numpy()\n\n\ndevice = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"fold_cv_scores = []\nsubmission_scores = []\n\nfolds = KFold(n_splits=5, shuffle = True, random_state = 42)\nfold_num = 0\nfor tr_idx, va_idx in folds.split(files_train):\n    print(f\"Starting fold: {fold_num}\")\n    no_imp = 0\n    CFG['batch_size'] = 32\n    checkpoint_filepath = f\"checkpoint_{fold_num}.h5\"\n    \n    files_train_tr = files_train[tr_idx]\n    files_train_va = files_train[va_idx]  \n    \n    ds_train     = get_dataset(files_train_tr, CFG, augment=True, shuffle=True)\n    ds_val     = get_dataset(files_train_va, CFG, augment=False, shuffle=False)\n    \n    model = get_model()\n    optimizer, loss_object = get_opt_loss_fn(model)\n    train_fn = get_train_fn()\n    torch.cuda.empty_cache()\n    \n    bestLoss = float(\"inf\")\n    for e in range(CFG[\"epochs\"]):\n        trainLoss = 0\n        tk0 = tqdm(ds_train)\n        for idx, (x,y,_) in enumerate(tk0):\n            loss = train_fn(x, y, model, optimizer, loss_object)\n            trainLoss += loss.numpy()\n            tk0.set_postfix(loss=trainLoss/(idx+1))\n            \n        testLoss = 0\n        tk0 = tqdm(ds_val)\n        for idx, (x,y,_) in enumerate(tk0):\n            loss = test_step(x,y)\n            testLoss += loss.numpy()\n            tk0.set_postfix(loss=testLoss/(idx+1))\n            \n        testLoss /= idx\n        if testLoss < bestLoss:\n            no_imp = 0\n            bestLoss = testLoss\n            torch.save(model.state_dict(), checkpoint_filepath)\n        else:\n            no_imp += 1\n            if no_imp > CFG[\"es_patience\"]:\n                print(\"Early stopping..\")\n                break\n            \n\n    model.load_state_dict(torch.load(checkpoint_filepath))\n    \n    ##############################################################\n    CFG['batch_size'] = 256\n    ds_valAug = get_dataset(files_train_va, CFG, augment=True)\n    ds_testAug = get_dataset(files_test, CFG, augment=True, labeled=False)\n    for t in range(CFG['tta_steps']):\n        ####EVAL FOLD\n        for idx, (x,y,fn) in enumerate(ds_valAug):\n            predictions = pred_step(x)\n            for j in range(predictions.shape[0]):\n                fold_cv_scores.append([fold_num, \n                                       fn[j].numpy().decode(\"utf-8\"),\n                                        predictions[j,0].item(),\n                                        y[j].numpy()])\n        \n        ####INFER ON TEST\n        for idx, (x,y,fn) in enumerate(ds_testAug):\n            predictions = pred_step(x)\n            for j in range(predictions.shape[0]):\n                submission_scores.append([fold_num, \n                                       fn[j].numpy().decode(\"utf-8\"),\n                                        predictions[j,0].item()])\n        \n    \n    fold_num += 1\n","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"df_fold = pd.DataFrame(fold_cv_scores, columns=[\"Fold\", \"Filename\", \"Pred\", \"Label\"])\ndf_sub = pd.DataFrame(submission_scores, columns=[\"Fold\", \"Filename\", \"Pred\"])","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"df_fold = df_fold.groupby([\"Filename\"]).mean().reset_index()\nprint(\"CV ROCAUC: \")\nprint(roc_auc_score(df_fold[\"Label\"], df_fold[\"Pred\"]))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"df_sub = df_sub.groupby([\"Filename\"]).mean().reset_index()\ndf_sub = df_sub[[\"Filename\", \"Pred\"]]\ndf_sub.columns=[\"image_name\", \"target\"]\ndf_sub.to_csv(\"submission.csv\", index=False)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat":4,"nbformat_minor":4}