{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.11.13","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"nvidiaTeslaT4","dataSources":[],"dockerImageVersionId":31090,"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":true}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"from datetime import datetime\nimport math\nimport cv2\nimport glob\nimport numpy as np\nfrom numpy import zeros, ones, vstack, hstack\nfrom numpy.random import permutation\nimport os\nimport torch\nimport torch.nn as nn\nimport torch.optim as optim\nimport torch.nn.functional as F\nfrom torch.utils.data import Dataset, DataLoader\nimport torchvision.models as models\nimport torchvision.transforms as transforms\nfrom sklearn.utils import shuffle\nfrom skimage.metrics import structural_similarity as ssim","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true,"execution":{"iopub.status.busy":"2025-08-25T09:49:58.841967Z","iopub.execute_input":"2025-08-25T09:49:58.842174Z","iopub.status.idle":"2025-08-25T09:50:06.497326Z","shell.execute_reply.started":"2025-08-25T09:49:58.842157Z","shell.execute_reply":"2025-08-25T09:50:06.496552Z"}},"outputs":[],"execution_count":1},{"cell_type":"code","source":"!pip install opendatasets --quiet\nimport opendatasets as od\nod.download(\"https://www.kaggle.com/datasets/puneet6060/intel-image-classification\")","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-08-25T09:50:11.951011Z","iopub.execute_input":"2025-08-25T09:50:11.951622Z","iopub.status.idle":"2025-08-25T09:50:56.158470Z","shell.execute_reply.started":"2025-08-25T09:50:11.951597Z","shell.execute_reply":"2025-08-25T09:50:56.157815Z"}},"outputs":[{"name":"stdout","text":"Please provide your Kaggle credentials to download this dataset. Learn more: http://bit.ly/kaggle-creds\nYour Kaggle username:","output_type":"stream"},{"output_type":"stream","name":"stdin","text":"  fratzcan\n"},{"name":"stdout","text":"Your Kaggle Key:","output_type":"stream"},{"output_type":"stream","name":"stdin","text":"  ········\n"},{"name":"stdout","text":"Dataset URL: https://www.kaggle.com/datasets/puneet6060/intel-image-classification\nDownloading intel-image-classification.zip to ./intel-image-classification\n","output_type":"stream"},{"name":"stderr","text":"100%|██████████| 346M/346M [00:00<00:00, 1.07GB/s] \n","output_type":"stream"},{"name":"stdout","text":"\n","output_type":"stream"}],"execution_count":2},{"cell_type":"code","source":"# Set device\ndevice = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\nprint(f'Using device: {device}')","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-08-25T09:51:13.181271Z","iopub.execute_input":"2025-08-25T09:51:13.181678Z","iopub.status.idle":"2025-08-25T09:51:13.264644Z","shell.execute_reply.started":"2025-08-25T09:51:13.181654Z","shell.execute_reply":"2025-08-25T09:51:13.263807Z"}},"outputs":[{"name":"stdout","text":"Using device: cuda\n","output_type":"stream"}],"execution_count":3},{"cell_type":"code","source":"## convert RGB to the personal LAB (LAB2)\n# the input R,G,B,  must be 1D from 0 to 255\n# the outputs are 1D  L [0 1], a [-1 1] b [-1 1]\ndef RGB2LAB2(R0, G0, B0):\n    R = R0 / 255\n    G = G0 / 255\n    B = B0 / 255\n\n    Y = 0.299*R + 0.587*G + 0.114*B\n    X = 0.449*R + 0.353*G + 0.198*B\n    Z = 0.012*R + 0.089*G + 0.899*B\n\n    L = Y\n    a = (X - Y) / 0.234\n    b = (Y - Z) / 0.785\n\n    return L, a, b\n\n## convert the personal LAB (LAB2) to RGB\n# the input L,a,b,  must be 1D L [0 1], a [-1 1] b [-1 1]\n# the outputs are 1D  R G B [0 255]\ndef LAB22RGB(L, a, b):\n    a11 = 0.299\n    a12 = 0.587\n    a13 = 0.114\n    a21 = (0.15/0.234)\n    a22 = (-0.234/0.234)\n    a23 = (0.084/0.234)\n    a31 = (0.287/0.785)\n    a32 = (0.498/0.785)\n    a33 = (-0.785/0.785)\n\n    aa = np.array([[a11, a12, a13], [a21, a22, a23], [a31, a32, a33]])\n    C0 = np.zeros((L.shape[0], 3))\n    C0[:, 0] = L[:, 0]\n    C0[:, 1] = a[:, 0]\n    C0[:, 2] = b[:, 0]\n    C = np.transpose(C0)\n\n    X = np.linalg.inv(aa).dot(C)\n    X1D = np.reshape(X, (X.shape[0]*X.shape[1], 1))\n    p0 = np.where(X1D < 0)\n    X1D[p0[0]] = 0\n    p1 = np.where(X1D > 1)\n    X1D[p1[0]] = 1\n    Xr = np.reshape(X1D, (X.shape[0], X.shape[1]))\n\n    Rr = Xr[0][:]\n    Gr = Xr[1][:]\n    Br = Xr[2][:]\n\n    R = np.uint8(np.round(Rr * 255))\n    G = np.uint8(np.round(Gr * 255))\n    B = np.uint8(np.round(Br * 255))\n    return R, G, B","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-08-25T09:51:22.895293Z","iopub.execute_input":"2025-08-25T09:51:22.895571Z","iopub.status.idle":"2025-08-25T09:51:22.904146Z","shell.execute_reply.started":"2025-08-25T09:51:22.895549Z","shell.execute_reply":"2025-08-25T09:51:22.903257Z"}},"outputs":[],"execution_count":4},{"cell_type":"code","source":"def psnr(img1, img2):\n    mse = np.mean((img1.astype(\"float\") - img2.astype(\"float\")) ** 2)\n    if mse == 0:\n        return 100\n    PIXEL_MAX = 255.0\n    return 20 * math.log10(PIXEL_MAX / math.sqrt(mse))\n\ndef mse(imageA, imageB, bands):\n    err = np.sum((imageA.astype(\"float\") - imageB.astype(\"float\")) ** 2)\n    err /= float(imageA.shape[0] * imageA.shape[1] * bands)\n    return err\n\ndef mae(imageA, imageB, bands):\n    err = np.sum(np.abs((imageA.astype(\"float\") - imageB.astype(\"float\"))))\n    err /= float(imageA.shape[0] * imageA.shape[1] * bands)\n    return err\n\ndef rmse(imageA, imageB, bands):\n    err = np.sum((imageA.astype(\"float\") - imageB.astype(\"float\")) ** 2)\n    err /= float(imageA.shape[0] * imageA.shape[1] * bands)\n    err = np.sqrt(err)\n    return err","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-08-25T09:51:31.334980Z","iopub.execute_input":"2025-08-25T09:51:31.335541Z","iopub.status.idle":"2025-08-25T09:51:31.341634Z","shell.execute_reply.started":"2025-08-25T09:51:31.335514Z","shell.execute_reply":"2025-08-25T09:51:31.340831Z"}},"outputs":[],"execution_count":5},{"cell_type":"code","source":"class DoubleConv(nn.Module):\n    \"\"\"Double Convolution Block\"\"\"\n    def __init__(self, in_channels, out_channels):\n        super(DoubleConv, self).__init__()\n        self.double_conv = nn.Sequential(\n            nn.Conv2d(in_channels, out_channels, kernel_size=3, padding=1),\n            nn.ReLU(inplace=True),\n            nn.Conv2d(out_channels, out_channels, kernel_size=3, padding=1),\n            nn.ReLU(inplace=True)\n        )\n\n    def forward(self, x):\n        return self.double_conv(x)\n","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-08-25T09:51:39.392511Z","iopub.execute_input":"2025-08-25T09:51:39.393206Z","iopub.status.idle":"2025-08-25T09:51:39.397489Z","shell.execute_reply.started":"2025-08-25T09:51:39.393182Z","shell.execute_reply":"2025-08-25T09:51:39.396878Z"}},"outputs":[],"execution_count":6},{"cell_type":"code","source":"class TripleConv(nn.Module):\n    \"\"\"Triple Convolution Block\"\"\"\n    def __init__(self, in_channels, out_channels):\n        super(TripleConv, self).__init__()\n        self.triple_conv = nn.Sequential(\n            nn.Conv2d(in_channels, out_channels, kernel_size=3, padding=1),\n            nn.ReLU(inplace=True),\n            nn.Conv2d(out_channels, out_channels, kernel_size=3, padding=1),\n            nn.ReLU(inplace=True),\n            nn.Conv2d(out_channels, out_channels, kernel_size=3, padding=1),\n            nn.ReLU(inplace=True)\n        )\n\n    def forward(self, x):\n        return self.triple_conv(x)\n","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-08-25T09:51:45.973145Z","iopub.execute_input":"2025-08-25T09:51:45.973681Z","iopub.status.idle":"2025-08-25T09:51:45.978633Z","shell.execute_reply.started":"2025-08-25T09:51:45.973658Z","shell.execute_reply":"2025-08-25T09:51:45.977646Z"}},"outputs":[],"execution_count":7},{"cell_type":"code","source":"class UNet1(nn.Module):\n    def __init__(self, in_channels=1, out_channels=2):\n        super(UNet1, self).__init__()\n\n        # Encoder\n        self.conv1 = DoubleConv(in_channels, 64)\n        self.pool1 = nn.MaxPool2d(2)\n\n        self.conv2 = DoubleConv(64, 128)\n        self.pool2 = nn.MaxPool2d(2)\n\n        self.conv3 = TripleConv(128, 256)\n        self.pool3 = nn.MaxPool2d(2)\n\n        self.conv4 = TripleConv(256, 512)\n        self.pool4 = nn.MaxPool2d(2)\n\n        self.conv5 = TripleConv(512, 512)\n        self.pool5 = nn.MaxPool2d(2)\n\n        # Bottleneck\n        self.conv55 = TripleConv(512, 512)\n\n        # Decoder\n        self.up66 = nn.ConvTranspose2d(512, 512, kernel_size=2, stride=2)\n        self.conv66 = DoubleConv(1024, 512)  # 512 + 512 from skip connection\n\n        self.up6 = nn.ConvTranspose2d(512, 512, kernel_size=2, stride=2)\n        self.conv6 = DoubleConv(1024, 512)  # 512 + 512 from skip connection\n\n        self.up7 = nn.ConvTranspose2d(512, 256, kernel_size=2, stride=2)\n        self.conv7 = DoubleConv(512, 256)  # 256 + 256 from skip connection\n\n        self.up8 = nn.ConvTranspose2d(256, 128, kernel_size=2, stride=2)\n        self.conv8 = DoubleConv(256, 128)  # 128 + 128 from skip connection\n\n        self.up9 = nn.ConvTranspose2d(128, 64, kernel_size=2, stride=2)\n        self.conv9 = DoubleConv(128, 64)  # 64 + 64 from skip connection\n\n        # Multi-scale feature fusion\n        self.up_f02 = nn.Upsample(scale_factor=2, mode='bilinear', align_corners=True)\n        self.up_f12 = nn.Upsample(scale_factor=2, mode='bilinear', align_corners=True)\n\n        # Final layers\n        self.conv11 = nn.Conv2d(384, 128, kernel_size=3, padding=1)  # 64+64+128+128\n        self.relu11 = nn.ReLU(inplace=True)\n\n        self.conv12 = nn.Conv2d(128, 64, kernel_size=3, padding=1)\n        self.relu12 = nn.ReLU(inplace=True)\n\n        self.conv13 = nn.Conv2d(64, 64, kernel_size=3, padding=1)\n        self.relu13 = nn.ReLU(inplace=True)\n\n        self.conv14 = nn.Conv2d(64, out_channels, kernel_size=3, padding=1)\n        self.tanh = nn.Tanh()\n\n    def forward(self, x):\n        # Encoder\n        conv1 = self.conv1(x)\n        x1 = self.pool1(conv1)\n\n        conv2 = self.conv2(x1)\n        x2 = self.pool2(conv2)\n\n        conv3 = self.conv3(x2)\n        x3 = self.pool3(conv3)\n\n        conv4 = self.conv4(x3)\n        x4 = self.pool4(conv4)\n\n        conv5 = self.conv5(x4)\n        x5 = self.pool5(conv5)\n\n        # Bottleneck\n        conv55 = self.conv55(x5)\n\n        # Decoder\n        up66 = self.up66(conv55)\n        if up66.size()[2:] != conv5.size()[2:]:\n            up66 = F.interpolate(up66, size=conv5.size()[2:], mode=\"bilinear\", align_corners=True)\n        merge66 = torch.cat([conv5, up66], dim=1)\n        conv66 = self.conv66(merge66)\n\n        up6 = self.up6(conv66)\n        if up6.size()[2:] != conv4.size()[2:]:\n            up6 = F.interpolate(up6, size=conv4.size()[2:], mode=\"bilinear\", align_corners=True)\n        merge6 = torch.cat([conv4, up6], dim=1)\n        conv6 = self.conv6(merge6)\n\n        up7 = self.up7(conv6)\n        if up7.size()[2:] != conv3.size()[2:]:\n            up7 = F.interpolate(up7, size=conv3.size()[2:], mode=\"bilinear\", align_corners=True)\n        merge7 = torch.cat([conv3, up7], dim=1)\n        conv7 = self.conv7(merge7)\n\n        up8 = self.up8(conv7)\n        if up8.size()[2:] != conv2.size()[2:]:\n            up8 = F.interpolate(up8, size=conv2.size()[2:], mode=\"bilinear\", align_corners=True)\n        merge8 = torch.cat([conv2, up8], dim=1)\n        conv8 = self.conv8(merge8)\n\n        up9 = self.up9(conv8)\n        if up9.size()[2:] != conv1.size()[2:]:\n            up9 = F.interpolate(up9, size=conv1.size()[2:], mode=\"bilinear\", align_corners=True)\n        merge9 = torch.cat([conv1, up9], dim=1)\n        conv9 = self.conv9(merge9)\n\n\n        # Multi-scale feature fusion\n        up_f01 = conv1  # Original resolution\n        up_f11 = conv9  # Decoded features\n        up_f02 = self.up_f02(conv2)  # Upsampled encoder features\n        up_f12 = self.up_f12(conv8)  # Upsampled decoder features\n\n        # Concatenate multi-scale features\n        merge11 = torch.cat([up_f01, up_f11, up_f02, up_f12], dim=1)\n\n        # Final processing\n        conv11 = self.relu11(self.conv11(merge11))\n        conv12 = self.relu12(self.conv12(conv11))\n        conv13 = self.relu13(self.conv13(conv12))\n        output = self.tanh(self.conv14(conv13))\n\n        return output","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-08-25T09:51:56.250275Z","iopub.execute_input":"2025-08-25T09:51:56.250552Z","iopub.status.idle":"2025-08-25T09:51:56.265871Z","shell.execute_reply.started":"2025-08-25T09:51:56.250533Z","shell.execute_reply":"2025-08-25T09:51:56.265122Z"}},"outputs":[],"execution_count":8},{"cell_type":"code","source":"class ColorizationDataset(Dataset):\n    def __init__(self, file_list, dim=150):\n        self.file_list = file_list # image paths\n        self.dim = dim\n\n    def __len__(self):\n        return len(self.file_list)\n\n    def __getitem__(self, idx):\n        # Read image\n        img_path = self.file_list[idx]\n        img = cv2.imread(img_path)\n        img = cv2.resize(img, (self.dim, self.dim))\n\n        # Convert to LAB2\n        sz0, sz1 = img.shape[:2]\n\n        # R1 = img[:, :, 0].reshape(-1, 1)\n        # G1 = img[:, :, 1].reshape(-1, 1)\n        # B1 = img[:, :, 2].reshape(-1, 1)\n\n        R1 = img[:, :, 2].reshape(-1, 1)\n        G1 = img[:, :, 1].reshape(-1, 1)\n        B1 = img[:, :, 0].reshape(-1, 1)\n\n        L, A, B = RGB2LAB2(R1, G1, B1)\n\n        L = L.reshape(sz0, sz1, 1)\n        A = A.reshape(sz0, sz1)\n        B = B.reshape(sz0, sz1)\n\n        ab = np.stack([A, B], axis=2)\n\n        # Convert to tensors and normalize\n        L_tensor = torch.FloatTensor(L).permute(2, 0, 1)  # (1, H, W)\n        ab_tensor = torch.FloatTensor(ab).permute(2, 0, 1)  # (2, H, W)\n\n        return L_tensor, ab_tensor\n","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-08-25T09:52:04.420627Z","iopub.execute_input":"2025-08-25T09:52:04.420888Z","iopub.status.idle":"2025-08-25T09:52:04.427231Z","shell.execute_reply.started":"2025-08-25T09:52:04.420869Z","shell.execute_reply":"2025-08-25T09:52:04.426444Z"}},"outputs":[],"execution_count":9},{"cell_type":"code","source":"def load_vgg16_weights(model):\n    \"\"\"Load pretrained VGG16 weights to U-Net encoder\"\"\"\n    vgg16 = models.vgg16(pretrained=True).to(device)\n    vgg_features = vgg16.features\n\n    # Adapt first layer from RGB to grayscale\n    with torch.no_grad():\n        # Get original RGB weights\n        rgb_weights = vgg_features[0].weight  # Shape: (64, 3, 3, 3)\n        # Average across RGB channels\n        gray_weights = rgb_weights.mean(dim=1, keepdim=True)  # Shape: (64, 1, 3, 3)\n\n        # Set weights for first layer\n        model.conv1.double_conv[0].weight.data = gray_weights\n        model.conv1.double_conv[0].bias.data = vgg_features[0].bias.data\n\n        # Set weights for second conv in first block\n        model.conv1.double_conv[2].weight.data = vgg_features[2].weight.data\n        model.conv1.double_conv[2].bias.data = vgg_features[2].bias.data\n\n        # Second block\n        model.conv2.double_conv[0].weight.data = vgg_features[5].weight.data\n        model.conv2.double_conv[0].bias.data = vgg_features[5].bias.data\n        model.conv2.double_conv[2].weight.data = vgg_features[7].weight.data\n        model.conv2.double_conv[2].bias.data = vgg_features[7].bias.data\n\n        # Third block (first two convs)\n        model.conv3.triple_conv[0].weight.data = vgg_features[10].weight.data\n        model.conv3.triple_conv[0].bias.data = vgg_features[10].bias.data\n        model.conv3.triple_conv[2].weight.data = vgg_features[12].weight.data\n        model.conv3.triple_conv[2].bias.data = vgg_features[12].bias.data\n        model.conv3.triple_conv[4].weight.data = vgg_features[14].weight.data\n        model.conv3.triple_conv[4].bias.data = vgg_features[14].bias.data\n\n        # Fourth block\n        model.conv4.triple_conv[0].weight.data = vgg_features[17].weight.data\n        model.conv4.triple_conv[0].bias.data = vgg_features[17].bias.data\n        model.conv4.triple_conv[2].weight.data = vgg_features[19].weight.data\n        model.conv4.triple_conv[2].bias.data = vgg_features[19].bias.data\n        model.conv4.triple_conv[4].weight.data = vgg_features[21].weight.data\n        model.conv4.triple_conv[4].bias.data = vgg_features[21].bias.data\n\n        # Fifth block\n        model.conv5.triple_conv[0].weight.data = vgg_features[24].weight.data\n        model.conv5.triple_conv[0].bias.data = vgg_features[24].bias.data\n        model.conv5.triple_conv[2].weight.data = vgg_features[26].weight.data\n        model.conv5.triple_conv[2].bias.data = vgg_features[26].bias.data\n        model.conv5.triple_conv[4].weight.data = vgg_features[28].weight.data\n        model.conv5.triple_conv[4].bias.data = vgg_features[28].bias.data\n","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-08-25T09:52:13.308431Z","iopub.execute_input":"2025-08-25T09:52:13.308713Z","iopub.status.idle":"2025-08-25T09:52:13.317859Z","shell.execute_reply.started":"2025-08-25T09:52:13.308691Z","shell.execute_reply":"2025-08-25T09:52:13.317264Z"}},"outputs":[],"execution_count":10},{"cell_type":"code","source":"def train_model():\n    device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n\n    # Parameters\n    dim = 150\n    batch_size = 16\n    epochs_max = 30\n    max_nb_min = 3\n\n    cwd = os.getcwd()\n    # train_path = os.path.join(cwd, 'seg_train', '*.png')\n    # files_tr_list = glob.glob(train_path)\n\n    # N = len(files_tr_list)\n    # print(f'Number of training images: {N}')\n\n\n\n    # base_path = \"/content/intel-image-classification/seg_train/seg_train\"\n    base_path = \"/kaggle/working/intel-image-classification/seg_train/seg_train\"\n    classes = os.listdir(base_path)\n\n    image_paths = []\n\n    for cls in classes:\n        folder_path = os.path.join(base_path, cls, \"*.jpg\")\n        for file in glob.glob(folder_path):\n            image_paths.append(file)\n\n    print(\"Number of training images:\", len(image_paths))\n\n\n\n    # Create model\n    model = UNet1(in_channels=1, out_channels=2).to(device)\n\n    # Load VGG16 pretrained weights\n    print('Loading VGG16 pretrained weights...')\n    load_vgg16_weights(model)\n\n    # Loss function and optimizer\n    criterion = nn.L1Loss()  # Mean Absolute Error\n    optimizer = optim.Adam(model.parameters(), lr=1e-4)\n\n    # Create dataset and dataloader\n    dataset = ColorizationDataset(image_paths, dim=dim)\n    dataloader = DataLoader(dataset, batch_size=batch_size, shuffle=True, num_workers=2)\n\n    # wandb.init(project=\"image_colorization\", name=\"HyperUNet_experiment-Intel_Image\",\n    #            config={\n    #                \"batch_size\": batch_size,\n    #                \"epochs\": epochs_max,\n    #                \"learning_rate\": 1e-4,\n    #                \"model_name\": \"HyperUNet-Intel-Image\",\n    #                \"optimizer\": \"Adam\",\n    #                \"loss_function\": \"L1Loss\",\n    #                \"dataset\": \"Intel Image\"\n    #            })\n\n    # Training tracking\n    tr_acc = np.zeros((epochs_max, 2))\n    time_tr = np.zeros((epochs_max, 2))\n    mae_min = float('inf')\n    nb_min = 0\n    stop = 0\n\n    print('Starting training...')\n\n    for epoch in range(epochs_max):\n        if stop:\n            break\n\n        start_time = datetime.now()\n        model.train()\n\n        total_loss = 0.0\n        num_batches = 0\n\n        for batch_idx, (inputs, targets) in enumerate(dataloader):\n            inputs, targets = inputs.to(device), targets.to(device)\n\n            optimizer.zero_grad()\n            outputs = model(inputs)\n            loss = criterion(outputs, targets)\n            loss.backward()\n            optimizer.step()\n\n            total_loss += loss.item()\n            num_batches += 1\n\n            # wandb.log({\"batch_loss\": loss.item()})\n\n            if batch_idx % 10 == 0:\n                print(f'Epoch {epoch+1}, Batch {batch_idx}/{len(dataloader)}, Loss: {loss.item():.6f}')\n\n        avg_loss = total_loss / num_batches\n        # wandb.log({\"epoch\": epoch+1, \"avg_loss\": avg_loss})\n\n        # Update tracking arrays\n        tr_acc[epoch, 0] = epoch\n        tr_acc[epoch, 1] = avg_loss\n\n        end_time = datetime.now()\n        time_diff = end_time - start_time\n        time_tr[epoch, 0] = epoch\n        time_tr[epoch, 1] = time_diff.seconds\n\n        print(f'Epoch {epoch+1}/{epochs_max}, Average Loss: {avg_loss:.6f}, Time: {time_diff.seconds}s')\n\n        # Early stopping logic\n        if avg_loss > mae_min:\n            nb_min += 1\n        else:\n            mae_min = avg_loss\n            nb_min = 0\n            # Save best model\n            torch.save(model.state_dict(), os.path.join(cwd, 'Hyper_U_NET_pytorch.pth'))\n            print(f'New best model saved with loss: {mae_min:.6f}')\n\n        if nb_min > max_nb_min:\n            stop = 1\n            print('Early stopping triggered')\n\n        # Learning rate scheduling (same as original)\n        if epoch + 1 == 1:\n            for param_group in optimizer.param_groups:\n                param_group['lr'] = 5e-5\n        elif epoch + 1 == 2:\n            for param_group in optimizer.param_groups:\n                param_group['lr'] = 2e-5\n        elif epoch + 1 == 4:\n            for param_group in optimizer.param_groups:\n                param_group['lr'] = 1e-5\n        elif epoch + 1 == 8:\n            for param_group in optimizer.param_groups:\n                param_group['lr'] = 5e-6\n        elif epoch + 1 == 16:\n            for param_group in optimizer.param_groups:\n                param_group['lr'] = 2e-6\n        elif epoch + 1 == 32:\n            for param_group in optimizer.param_groups:\n                param_group['lr'] = 1e-6\n        elif epoch + 1 == 64:\n            for param_group in optimizer.param_groups:\n                param_group['lr'] = 5e-7\n        elif epoch + 1 == 128:\n            for param_group in optimizer.param_groups:\n                param_group['lr'] = 2e-7\n        elif epoch + 1 == 256:\n            for param_group in optimizer.param_groups:\n                param_group['lr'] = 1e-7\n\n        # Save progress\n        np.save(os.path.join(cwd, 'tr_Acc_Hyper_U_NET_pytorch.npy'), tr_acc)\n        np.save(os.path.join(cwd, 'Tr_runtime_Hyper_U_NET_pytorch.npy'), time_tr)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-08-25T09:54:02.890391Z","iopub.execute_input":"2025-08-25T09:54:02.890909Z","iopub.status.idle":"2025-08-25T09:54:02.903005Z","shell.execute_reply.started":"2025-08-25T09:54:02.890885Z","shell.execute_reply":"2025-08-25T09:54:02.902310Z"}},"outputs":[],"execution_count":13},{"cell_type":"code","source":"def load_model_for_inference(model_path, device):\n    \"\"\"Load trained model for inference\"\"\"\n    model = UNet1(in_channels=1, out_channels=2).to(device)\n    model.load_state_dict(torch.load(model_path, map_location=device))\n    model.eval()\n    return model\n\ndef inference(model, l_channel):\n    \"\"\"Perform inference on L channel to get ab channels\"\"\"\n    model.eval()\n    with torch.no_grad():\n        if len(l_channel.shape) == 3:\n            l_channel = l_channel.unsqueeze(0)  # Add batch dimension\n\n        l_tensor = torch.FloatTensor(l_channel).to(device)\n        ab_pred = model(l_tensor)\n\n        return ab_pred.cpu().numpy()\n\n","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-08-25T09:54:04.306877Z","iopub.execute_input":"2025-08-25T09:54:04.307696Z","iopub.status.idle":"2025-08-25T09:54:04.312571Z","shell.execute_reply.started":"2025-08-25T09:54:04.307669Z","shell.execute_reply":"2025-08-25T09:54:04.311816Z"}},"outputs":[],"execution_count":14},{"cell_type":"code","source":"if __name__ == \"__main__\":\n    # Start training\n    train_model()","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-08-25T09:54:05.741769Z","iopub.execute_input":"2025-08-25T09:54:05.741970Z","iopub.status.idle":"2025-08-25T13:46:01.526305Z","shell.execute_reply.started":"2025-08-25T09:54:05.741956Z","shell.execute_reply":"2025-08-25T13:46:01.525259Z"},"collapsed":true,"jupyter":{"outputs_hidden":true}},"outputs":[{"name":"stdout","text":"Number of training images: 14034\nLoading VGG16 pretrained weights...\n","output_type":"stream"},{"name":"stderr","text":"/usr/local/lib/python3.11/dist-packages/torchvision/models/_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead.\n  warnings.warn(\n/usr/local/lib/python3.11/dist-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=VGG16_Weights.IMAGENET1K_V1`. You can also use `weights=VGG16_Weights.DEFAULT` to get the most up-to-date weights.\n  warnings.warn(msg)\nDownloading: \"https://download.pytorch.org/models/vgg16-397923af.pth\" to /root/.cache/torch/hub/checkpoints/vgg16-397923af.pth\n100%|██████████| 528M/528M [00:03<00:00, 183MB/s]  \n","output_type":"stream"},{"name":"stdout","text":"Starting training...\nEpoch 1, Batch 0/878, Loss: 0.070598\nEpoch 1, Batch 10/878, Loss: 0.074728\nEpoch 1, Batch 20/878, Loss: 0.052602\nEpoch 1, Batch 30/878, Loss: 0.073228\nEpoch 1, Batch 40/878, Loss: 0.057651\nEpoch 1, Batch 50/878, Loss: 0.054156\nEpoch 1, Batch 60/878, Loss: 0.045435\nEpoch 1, Batch 70/878, Loss: 0.043975\nEpoch 1, Batch 80/878, Loss: 0.074865\nEpoch 1, Batch 90/878, Loss: 0.063429\nEpoch 1, Batch 100/878, Loss: 0.068418\nEpoch 1, Batch 110/878, Loss: 0.058281\nEpoch 1, Batch 120/878, Loss: 0.048126\nEpoch 1, Batch 130/878, Loss: 0.058792\nEpoch 1, Batch 140/878, Loss: 0.045423\nEpoch 1, Batch 150/878, Loss: 0.059007\nEpoch 1, Batch 160/878, Loss: 0.080718\nEpoch 1, Batch 170/878, Loss: 0.054839\nEpoch 1, Batch 180/878, Loss: 0.056738\nEpoch 1, Batch 190/878, Loss: 0.055126\nEpoch 1, Batch 200/878, Loss: 0.061762\nEpoch 1, Batch 210/878, Loss: 0.055855\nEpoch 1, Batch 220/878, Loss: 0.041474\nEpoch 1, Batch 230/878, Loss: 0.057545\nEpoch 1, Batch 240/878, Loss: 0.050002\nEpoch 1, Batch 250/878, Loss: 0.054547\nEpoch 1, Batch 260/878, Loss: 0.046063\nEpoch 1, Batch 270/878, Loss: 0.060570\nEpoch 1, Batch 280/878, Loss: 0.049382\nEpoch 1, Batch 290/878, Loss: 0.058270\nEpoch 1, Batch 300/878, Loss: 0.044802\nEpoch 1, Batch 310/878, Loss: 0.061397\nEpoch 1, Batch 320/878, Loss: 0.054343\nEpoch 1, Batch 330/878, Loss: 0.048667\nEpoch 1, Batch 340/878, Loss: 0.055078\nEpoch 1, Batch 350/878, Loss: 0.059521\nEpoch 1, Batch 360/878, Loss: 0.058220\nEpoch 1, Batch 370/878, Loss: 0.064795\nEpoch 1, Batch 380/878, Loss: 0.043378\nEpoch 1, Batch 390/878, Loss: 0.071293\nEpoch 1, Batch 400/878, Loss: 0.043741\nEpoch 1, Batch 410/878, Loss: 0.073510\nEpoch 1, Batch 420/878, Loss: 0.056576\nEpoch 1, Batch 430/878, Loss: 0.060951\nEpoch 1, Batch 440/878, Loss: 0.039490\nEpoch 1, Batch 450/878, Loss: 0.069587\nEpoch 1, Batch 460/878, Loss: 0.057864\nEpoch 1, Batch 470/878, Loss: 0.045966\nEpoch 1, Batch 480/878, Loss: 0.057269\nEpoch 1, Batch 490/878, Loss: 0.052792\nEpoch 1, Batch 500/878, Loss: 0.061484\nEpoch 1, Batch 510/878, Loss: 0.053010\nEpoch 1, Batch 520/878, Loss: 0.055168\nEpoch 1, Batch 530/878, Loss: 0.068453\nEpoch 1, Batch 540/878, Loss: 0.059330\nEpoch 1, Batch 550/878, Loss: 0.050106\nEpoch 1, Batch 560/878, Loss: 0.051833\nEpoch 1, Batch 570/878, Loss: 0.044979\nEpoch 1, Batch 580/878, Loss: 0.058946\nEpoch 1, Batch 590/878, Loss: 0.045736\nEpoch 1, Batch 600/878, Loss: 0.060092\nEpoch 1, Batch 610/878, Loss: 0.063554\nEpoch 1, Batch 620/878, Loss: 0.042710\nEpoch 1, Batch 630/878, Loss: 0.054122\nEpoch 1, Batch 640/878, Loss: 0.041295\nEpoch 1, Batch 650/878, Loss: 0.051371\nEpoch 1, Batch 660/878, Loss: 0.036994\nEpoch 1, Batch 670/878, Loss: 0.059373\nEpoch 1, Batch 680/878, Loss: 0.041692\nEpoch 1, Batch 690/878, Loss: 0.059191\nEpoch 1, Batch 700/878, Loss: 0.048007\nEpoch 1, Batch 710/878, Loss: 0.057815\nEpoch 1, Batch 720/878, Loss: 0.062983\nEpoch 1, Batch 730/878, Loss: 0.049455\nEpoch 1, Batch 740/878, Loss: 0.047544\nEpoch 1, Batch 750/878, Loss: 0.059348\nEpoch 1, Batch 760/878, Loss: 0.042198\nEpoch 1, Batch 770/878, Loss: 0.049146\nEpoch 1, Batch 780/878, Loss: 0.044231\nEpoch 1, Batch 790/878, Loss: 0.051397\nEpoch 1, Batch 800/878, Loss: 0.053957\nEpoch 1, Batch 810/878, Loss: 0.045101\nEpoch 1, Batch 820/878, Loss: 0.056312\nEpoch 1, Batch 830/878, Loss: 0.057079\nEpoch 1, Batch 840/878, Loss: 0.055172\nEpoch 1, Batch 850/878, Loss: 0.051037\nEpoch 1, Batch 860/878, Loss: 0.052642\nEpoch 1, Batch 870/878, Loss: 0.058240\nEpoch 1/30, Average Loss: 0.054363, Time: 457s\nNew best model saved with loss: 0.054363\nEpoch 2, Batch 0/878, Loss: 0.071568\nEpoch 2, Batch 10/878, Loss: 0.059555\nEpoch 2, Batch 20/878, Loss: 0.053691\nEpoch 2, Batch 30/878, Loss: 0.044958\nEpoch 2, Batch 40/878, Loss: 0.048508\nEpoch 2, Batch 50/878, Loss: 0.047893\nEpoch 2, Batch 60/878, Loss: 0.057299\nEpoch 2, Batch 70/878, Loss: 0.053921\nEpoch 2, Batch 80/878, Loss: 0.058914\nEpoch 2, Batch 90/878, Loss: 0.051266\nEpoch 2, Batch 100/878, Loss: 0.050840\nEpoch 2, Batch 110/878, Loss: 0.049592\nEpoch 2, Batch 120/878, Loss: 0.037395\nEpoch 2, Batch 130/878, Loss: 0.062847\nEpoch 2, Batch 140/878, Loss: 0.042508\nEpoch 2, Batch 150/878, Loss: 0.050072\nEpoch 2, Batch 160/878, Loss: 0.063595\nEpoch 2, Batch 170/878, Loss: 0.054622\nEpoch 2, Batch 180/878, Loss: 0.039407\nEpoch 2, Batch 190/878, Loss: 0.047617\nEpoch 2, Batch 200/878, Loss: 0.039252\nEpoch 2, Batch 210/878, Loss: 0.063942\nEpoch 2, Batch 220/878, Loss: 0.052812\nEpoch 2, Batch 230/878, Loss: 0.043910\nEpoch 2, Batch 240/878, Loss: 0.049917\nEpoch 2, Batch 250/878, Loss: 0.043204\nEpoch 2, Batch 260/878, Loss: 0.060882\nEpoch 2, Batch 270/878, Loss: 0.060898\nEpoch 2, Batch 280/878, Loss: 0.048666\nEpoch 2, Batch 290/878, Loss: 0.051851\nEpoch 2, Batch 300/878, Loss: 0.076389\nEpoch 2, Batch 310/878, Loss: 0.046085\nEpoch 2, Batch 320/878, Loss: 0.062659\nEpoch 2, Batch 330/878, Loss: 0.046638\nEpoch 2, Batch 340/878, Loss: 0.043200\nEpoch 2, Batch 350/878, Loss: 0.062322\nEpoch 2, Batch 360/878, Loss: 0.054610\nEpoch 2, Batch 370/878, Loss: 0.050591\nEpoch 2, Batch 380/878, Loss: 0.044376\nEpoch 2, Batch 390/878, Loss: 0.043806\nEpoch 2, Batch 400/878, Loss: 0.051319\nEpoch 2, Batch 410/878, Loss: 0.051424\nEpoch 2, Batch 420/878, Loss: 0.057038\nEpoch 2, Batch 430/878, Loss: 0.040838\nEpoch 2, Batch 440/878, Loss: 0.049949\nEpoch 2, Batch 450/878, Loss: 0.043789\nEpoch 2, Batch 460/878, Loss: 0.048297\nEpoch 2, Batch 470/878, Loss: 0.052252\nEpoch 2, Batch 480/878, Loss: 0.052721\nEpoch 2, Batch 490/878, Loss: 0.048202\nEpoch 2, Batch 500/878, Loss: 0.043251\nEpoch 2, Batch 510/878, Loss: 0.056809\nEpoch 2, Batch 520/878, Loss: 0.044395\nEpoch 2, Batch 530/878, Loss: 0.050978\nEpoch 2, Batch 540/878, Loss: 0.045413\nEpoch 2, Batch 550/878, Loss: 0.041497\nEpoch 2, Batch 560/878, Loss: 0.036197\nEpoch 2, Batch 570/878, Loss: 0.055714\nEpoch 2, Batch 580/878, Loss: 0.047975\nEpoch 2, Batch 590/878, Loss: 0.054610\nEpoch 2, Batch 600/878, Loss: 0.044910\nEpoch 2, Batch 610/878, Loss: 0.057875\nEpoch 2, Batch 620/878, Loss: 0.039661\nEpoch 2, Batch 630/878, Loss: 0.051859\nEpoch 2, Batch 640/878, Loss: 0.049669\nEpoch 2, Batch 650/878, Loss: 0.036901\nEpoch 2, Batch 660/878, Loss: 0.051472\nEpoch 2, Batch 670/878, Loss: 0.043509\nEpoch 2, Batch 680/878, Loss: 0.068689\nEpoch 2, Batch 690/878, Loss: 0.053631\nEpoch 2, Batch 700/878, Loss: 0.029907\nEpoch 2, Batch 710/878, Loss: 0.043998\nEpoch 2, Batch 720/878, Loss: 0.047702\nEpoch 2, Batch 730/878, Loss: 0.050392\nEpoch 2, Batch 740/878, Loss: 0.052660\nEpoch 2, Batch 750/878, Loss: 0.042531\nEpoch 2, Batch 760/878, Loss: 0.046869\nEpoch 2, Batch 770/878, Loss: 0.038063\nEpoch 2, Batch 780/878, Loss: 0.045096\nEpoch 2, Batch 790/878, Loss: 0.054642\nEpoch 2, Batch 800/878, Loss: 0.062119\nEpoch 2, Batch 810/878, Loss: 0.066549\nEpoch 2, Batch 820/878, Loss: 0.052133\nEpoch 2, Batch 830/878, Loss: 0.042099\nEpoch 2, Batch 840/878, Loss: 0.052105\nEpoch 2, Batch 850/878, Loss: 0.041941\nEpoch 2, Batch 860/878, Loss: 0.050775\nEpoch 2, Batch 870/878, Loss: 0.046307\nEpoch 2/30, Average Loss: 0.049806, Time: 461s\nNew best model saved with loss: 0.049806\nEpoch 3, Batch 0/878, Loss: 0.052765\nEpoch 3, Batch 10/878, Loss: 0.069585\nEpoch 3, Batch 20/878, Loss: 0.048401\nEpoch 3, Batch 30/878, Loss: 0.043466\nEpoch 3, Batch 40/878, Loss: 0.050202\nEpoch 3, Batch 50/878, Loss: 0.060546\nEpoch 3, Batch 60/878, Loss: 0.055058\nEpoch 3, Batch 70/878, Loss: 0.046843\nEpoch 3, Batch 80/878, Loss: 0.040515\nEpoch 3, Batch 90/878, Loss: 0.044123\nEpoch 3, Batch 100/878, Loss: 0.050261\nEpoch 3, Batch 110/878, Loss: 0.038937\nEpoch 3, Batch 120/878, Loss: 0.031558\nEpoch 3, Batch 130/878, Loss: 0.053051\nEpoch 3, Batch 140/878, Loss: 0.051564\nEpoch 3, Batch 150/878, Loss: 0.050868\nEpoch 3, Batch 160/878, Loss: 0.040696\nEpoch 3, Batch 170/878, Loss: 0.046894\nEpoch 3, Batch 180/878, Loss: 0.039976\nEpoch 3, Batch 190/878, Loss: 0.048726\nEpoch 3, Batch 200/878, Loss: 0.043009\nEpoch 3, Batch 210/878, Loss: 0.039617\nEpoch 3, Batch 220/878, Loss: 0.041144\nEpoch 3, Batch 230/878, Loss: 0.049190\nEpoch 3, Batch 240/878, Loss: 0.049252\nEpoch 3, Batch 250/878, Loss: 0.050212\nEpoch 3, Batch 260/878, Loss: 0.053306\nEpoch 3, Batch 270/878, Loss: 0.039754\nEpoch 3, Batch 280/878, Loss: 0.054272\nEpoch 3, Batch 290/878, Loss: 0.054362\nEpoch 3, Batch 300/878, Loss: 0.056466\nEpoch 3, Batch 310/878, Loss: 0.042145\nEpoch 3, Batch 320/878, Loss: 0.051250\nEpoch 3, Batch 330/878, Loss: 0.052626\nEpoch 3, Batch 340/878, Loss: 0.041365\nEpoch 3, Batch 350/878, Loss: 0.036393\nEpoch 3, Batch 360/878, Loss: 0.048801\nEpoch 3, Batch 370/878, Loss: 0.058053\nEpoch 3, Batch 380/878, Loss: 0.040044\nEpoch 3, Batch 390/878, Loss: 0.046888\nEpoch 3, Batch 400/878, Loss: 0.039026\nEpoch 3, Batch 410/878, Loss: 0.038753\nEpoch 3, Batch 420/878, Loss: 0.054615\nEpoch 3, Batch 430/878, Loss: 0.044604\nEpoch 3, Batch 440/878, Loss: 0.046048\nEpoch 3, Batch 450/878, Loss: 0.045340\nEpoch 3, Batch 460/878, Loss: 0.048129\nEpoch 3, Batch 470/878, Loss: 0.050276\nEpoch 3, Batch 480/878, Loss: 0.041532\nEpoch 3, Batch 490/878, Loss: 0.045329\nEpoch 3, Batch 500/878, Loss: 0.049545\nEpoch 3, Batch 510/878, Loss: 0.047993\nEpoch 3, Batch 520/878, Loss: 0.048774\nEpoch 3, Batch 530/878, Loss: 0.047250\nEpoch 3, Batch 540/878, Loss: 0.047184\nEpoch 3, Batch 550/878, Loss: 0.047079\nEpoch 3, Batch 560/878, Loss: 0.055289\nEpoch 3, Batch 570/878, Loss: 0.038126\nEpoch 3, Batch 580/878, Loss: 0.033921\nEpoch 3, Batch 590/878, Loss: 0.033749\nEpoch 3, Batch 600/878, Loss: 0.043829\nEpoch 3, Batch 610/878, Loss: 0.058561\nEpoch 3, Batch 620/878, Loss: 0.037415\nEpoch 3, Batch 630/878, Loss: 0.042864\nEpoch 3, Batch 640/878, Loss: 0.043144\nEpoch 3, Batch 650/878, Loss: 0.053602\nEpoch 3, Batch 660/878, Loss: 0.058954\nEpoch 3, Batch 670/878, Loss: 0.070420\nEpoch 3, Batch 680/878, Loss: 0.045527\nEpoch 3, Batch 690/878, Loss: 0.044716\nEpoch 3, Batch 700/878, Loss: 0.044711\nEpoch 3, Batch 710/878, Loss: 0.036155\nEpoch 3, Batch 720/878, Loss: 0.045670\nEpoch 3, Batch 730/878, Loss: 0.037572\nEpoch 3, Batch 740/878, Loss: 0.038102\nEpoch 3, Batch 750/878, Loss: 0.038394\nEpoch 3, Batch 760/878, Loss: 0.049026\nEpoch 3, Batch 770/878, Loss: 0.047490\nEpoch 3, Batch 780/878, Loss: 0.039325\nEpoch 3, Batch 790/878, Loss: 0.036053\nEpoch 3, Batch 800/878, Loss: 0.052019\nEpoch 3, Batch 810/878, Loss: 0.036998\nEpoch 3, Batch 820/878, Loss: 0.044766\nEpoch 3, Batch 830/878, Loss: 0.052727\nEpoch 3, Batch 840/878, Loss: 0.055253\nEpoch 3, Batch 850/878, Loss: 0.047490\nEpoch 3, Batch 860/878, Loss: 0.042636\nEpoch 3, Batch 870/878, Loss: 0.061354\nEpoch 3/30, Average Loss: 0.047331, Time: 461s\nNew best model saved with loss: 0.047331\nEpoch 4, Batch 0/878, Loss: 0.055860\nEpoch 4, Batch 10/878, Loss: 0.049240\nEpoch 4, Batch 20/878, Loss: 0.040405\nEpoch 4, Batch 30/878, Loss: 0.044179\nEpoch 4, Batch 40/878, Loss: 0.042628\nEpoch 4, Batch 50/878, Loss: 0.049119\nEpoch 4, Batch 60/878, Loss: 0.041779\nEpoch 4, Batch 70/878, Loss: 0.050498\nEpoch 4, Batch 80/878, Loss: 0.042009\nEpoch 4, Batch 90/878, Loss: 0.045578\nEpoch 4, Batch 100/878, Loss: 0.044359\nEpoch 4, Batch 110/878, Loss: 0.048325\nEpoch 4, Batch 120/878, Loss: 0.044986\nEpoch 4, Batch 130/878, Loss: 0.046007\nEpoch 4, Batch 140/878, Loss: 0.052767\nEpoch 4, Batch 150/878, Loss: 0.045557\nEpoch 4, Batch 160/878, Loss: 0.044040\nEpoch 4, Batch 170/878, Loss: 0.043935\nEpoch 4, Batch 180/878, Loss: 0.048985\nEpoch 4, Batch 190/878, Loss: 0.047897\nEpoch 4, Batch 200/878, Loss: 0.041311\nEpoch 4, Batch 210/878, Loss: 0.044108\nEpoch 4, Batch 220/878, Loss: 0.051012\nEpoch 4, Batch 230/878, Loss: 0.046954\nEpoch 4, Batch 240/878, Loss: 0.051218\nEpoch 4, Batch 250/878, Loss: 0.047014\nEpoch 4, Batch 260/878, Loss: 0.048769\nEpoch 4, Batch 270/878, Loss: 0.060430\nEpoch 4, Batch 280/878, Loss: 0.059824\nEpoch 4, Batch 290/878, Loss: 0.041187\nEpoch 4, Batch 300/878, Loss: 0.042067\nEpoch 4, Batch 310/878, Loss: 0.055813\nEpoch 4, Batch 320/878, Loss: 0.034237\nEpoch 4, Batch 330/878, Loss: 0.055192\nEpoch 4, Batch 340/878, Loss: 0.047964\nEpoch 4, Batch 350/878, Loss: 0.030646\nEpoch 4, Batch 360/878, Loss: 0.055772\nEpoch 4, Batch 370/878, Loss: 0.045173\nEpoch 4, Batch 380/878, Loss: 0.048787\nEpoch 4, Batch 390/878, Loss: 0.053105\nEpoch 4, Batch 400/878, Loss: 0.049802\nEpoch 4, Batch 410/878, Loss: 0.036131\nEpoch 4, Batch 420/878, Loss: 0.045681\nEpoch 4, Batch 430/878, Loss: 0.040919\nEpoch 4, Batch 440/878, Loss: 0.040403\nEpoch 4, Batch 450/878, Loss: 0.042989\nEpoch 4, Batch 460/878, Loss: 0.055533\nEpoch 4, Batch 470/878, Loss: 0.043093\nEpoch 4, Batch 480/878, Loss: 0.043840\nEpoch 4, Batch 490/878, Loss: 0.045499\nEpoch 4, Batch 500/878, Loss: 0.049092\nEpoch 4, Batch 510/878, Loss: 0.057652\nEpoch 4, Batch 520/878, Loss: 0.046527\nEpoch 4, Batch 530/878, Loss: 0.046018\nEpoch 4, Batch 540/878, Loss: 0.038073\nEpoch 4, Batch 550/878, Loss: 0.042849\nEpoch 4, Batch 560/878, Loss: 0.042045\nEpoch 4, Batch 570/878, Loss: 0.054457\nEpoch 4, Batch 580/878, Loss: 0.045177\nEpoch 4, Batch 590/878, Loss: 0.033750\nEpoch 4, Batch 600/878, Loss: 0.057273\nEpoch 4, Batch 610/878, Loss: 0.037829\nEpoch 4, Batch 620/878, Loss: 0.041168\nEpoch 4, Batch 630/878, Loss: 0.041973\nEpoch 4, Batch 640/878, Loss: 0.051733\nEpoch 4, Batch 650/878, Loss: 0.050881\nEpoch 4, Batch 660/878, Loss: 0.040133\nEpoch 4, Batch 670/878, Loss: 0.044444\nEpoch 4, Batch 680/878, Loss: 0.038169\nEpoch 4, Batch 690/878, Loss: 0.040817\nEpoch 4, Batch 700/878, Loss: 0.050496\nEpoch 4, Batch 710/878, Loss: 0.039509\nEpoch 4, Batch 720/878, Loss: 0.051653\nEpoch 4, Batch 730/878, Loss: 0.047015\nEpoch 4, Batch 740/878, Loss: 0.050026\nEpoch 4, Batch 750/878, Loss: 0.054173\nEpoch 4, Batch 760/878, Loss: 0.042993\nEpoch 4, Batch 770/878, Loss: 0.051888\nEpoch 4, Batch 780/878, Loss: 0.041211\nEpoch 4, Batch 790/878, Loss: 0.039406\nEpoch 4, Batch 800/878, Loss: 0.049240\nEpoch 4, Batch 810/878, Loss: 0.061593\nEpoch 4, Batch 820/878, Loss: 0.045911\nEpoch 4, Batch 830/878, Loss: 0.038176\nEpoch 4, Batch 840/878, Loss: 0.041044\nEpoch 4, Batch 850/878, Loss: 0.044816\nEpoch 4, Batch 860/878, Loss: 0.052082\nEpoch 4, Batch 870/878, Loss: 0.048485\nEpoch 4/30, Average Loss: 0.046249, Time: 461s\nNew best model saved with loss: 0.046249\nEpoch 5, Batch 0/878, Loss: 0.045768\nEpoch 5, Batch 10/878, Loss: 0.053946\nEpoch 5, Batch 20/878, Loss: 0.055208\nEpoch 5, Batch 30/878, Loss: 0.050316\nEpoch 5, Batch 40/878, Loss: 0.042092\nEpoch 5, Batch 50/878, Loss: 0.050396\nEpoch 5, Batch 60/878, Loss: 0.045910\nEpoch 5, Batch 70/878, Loss: 0.039405\nEpoch 5, Batch 80/878, Loss: 0.042416\nEpoch 5, Batch 90/878, Loss: 0.052463\nEpoch 5, Batch 100/878, Loss: 0.041739\nEpoch 5, Batch 110/878, Loss: 0.045521\nEpoch 5, Batch 120/878, Loss: 0.037623\nEpoch 5, Batch 130/878, Loss: 0.038559\nEpoch 5, Batch 140/878, Loss: 0.042473\nEpoch 5, Batch 150/878, Loss: 0.037497\nEpoch 5, Batch 160/878, Loss: 0.034093\nEpoch 5, Batch 170/878, Loss: 0.050291\nEpoch 5, Batch 180/878, Loss: 0.044300\nEpoch 5, Batch 190/878, Loss: 0.040353\nEpoch 5, Batch 200/878, Loss: 0.043065\nEpoch 5, Batch 210/878, Loss: 0.033334\nEpoch 5, Batch 220/878, Loss: 0.048109\nEpoch 5, Batch 230/878, Loss: 0.040695\nEpoch 5, Batch 240/878, Loss: 0.056595\nEpoch 5, Batch 250/878, Loss: 0.043172\nEpoch 5, Batch 260/878, Loss: 0.030951\nEpoch 5, Batch 270/878, Loss: 0.048341\nEpoch 5, Batch 280/878, Loss: 0.048290\nEpoch 5, Batch 290/878, Loss: 0.041386\nEpoch 5, Batch 300/878, Loss: 0.054062\nEpoch 5, Batch 310/878, Loss: 0.055092\nEpoch 5, Batch 320/878, Loss: 0.050416\nEpoch 5, Batch 330/878, Loss: 0.039103\nEpoch 5, Batch 340/878, Loss: 0.039394\nEpoch 5, Batch 350/878, Loss: 0.050525\nEpoch 5, Batch 360/878, Loss: 0.035177\nEpoch 5, Batch 370/878, Loss: 0.040239\nEpoch 5, Batch 380/878, Loss: 0.047072\nEpoch 5, Batch 390/878, Loss: 0.039619\nEpoch 5, Batch 400/878, Loss: 0.041742\nEpoch 5, Batch 410/878, Loss: 0.047552\nEpoch 5, Batch 420/878, Loss: 0.050467\nEpoch 5, Batch 430/878, Loss: 0.042817\nEpoch 5, Batch 440/878, Loss: 0.043964\nEpoch 5, Batch 450/878, Loss: 0.048002\nEpoch 5, Batch 460/878, Loss: 0.047955\nEpoch 5, Batch 470/878, Loss: 0.037793\nEpoch 5, Batch 480/878, Loss: 0.064665\nEpoch 5, Batch 490/878, Loss: 0.044016\nEpoch 5, Batch 500/878, Loss: 0.049968\nEpoch 5, Batch 510/878, Loss: 0.031705\nEpoch 5, Batch 520/878, Loss: 0.044819\nEpoch 5, Batch 530/878, Loss: 0.044217\nEpoch 5, Batch 540/878, Loss: 0.046165\nEpoch 5, Batch 550/878, Loss: 0.038515\nEpoch 5, Batch 560/878, Loss: 0.059196\nEpoch 5, Batch 570/878, Loss: 0.047771\nEpoch 5, Batch 580/878, Loss: 0.047914\nEpoch 5, Batch 590/878, Loss: 0.038009\nEpoch 5, Batch 600/878, Loss: 0.044480\nEpoch 5, Batch 610/878, Loss: 0.045081\nEpoch 5, Batch 620/878, Loss: 0.035401\nEpoch 5, Batch 630/878, Loss: 0.034913\nEpoch 5, Batch 640/878, Loss: 0.037041\nEpoch 5, Batch 650/878, Loss: 0.052033\nEpoch 5, Batch 660/878, Loss: 0.042832\nEpoch 5, Batch 670/878, Loss: 0.040532\nEpoch 5, Batch 680/878, Loss: 0.038777\nEpoch 5, Batch 690/878, Loss: 0.045181\nEpoch 5, Batch 700/878, Loss: 0.041395\nEpoch 5, Batch 710/878, Loss: 0.045609\nEpoch 5, Batch 720/878, Loss: 0.044131\nEpoch 5, Batch 730/878, Loss: 0.048632\nEpoch 5, Batch 740/878, Loss: 0.036298\nEpoch 5, Batch 750/878, Loss: 0.042052\nEpoch 5, Batch 760/878, Loss: 0.036192\nEpoch 5, Batch 770/878, Loss: 0.034358\nEpoch 5, Batch 780/878, Loss: 0.061695\nEpoch 5, Batch 790/878, Loss: 0.044108\nEpoch 5, Batch 800/878, Loss: 0.052947\nEpoch 5, Batch 810/878, Loss: 0.043954\nEpoch 5, Batch 820/878, Loss: 0.047575\nEpoch 5, Batch 830/878, Loss: 0.040757\nEpoch 5, Batch 840/878, Loss: 0.049200\nEpoch 5, Batch 850/878, Loss: 0.041278\nEpoch 5, Batch 860/878, Loss: 0.053713\nEpoch 5, Batch 870/878, Loss: 0.049537\nEpoch 5/30, Average Loss: 0.044641, Time: 461s\nNew best model saved with loss: 0.044641\nEpoch 6, Batch 0/878, Loss: 0.037291\nEpoch 6, Batch 10/878, Loss: 0.036678\nEpoch 6, Batch 20/878, Loss: 0.034163\nEpoch 6, Batch 30/878, Loss: 0.042419\nEpoch 6, Batch 40/878, Loss: 0.057812\nEpoch 6, Batch 50/878, Loss: 0.046892\nEpoch 6, Batch 60/878, Loss: 0.043964\nEpoch 6, Batch 70/878, Loss: 0.039872\nEpoch 6, Batch 80/878, Loss: 0.044813\nEpoch 6, Batch 90/878, Loss: 0.033083\nEpoch 6, Batch 100/878, Loss: 0.039750\nEpoch 6, Batch 110/878, Loss: 0.037272\nEpoch 6, Batch 120/878, Loss: 0.046021\nEpoch 6, Batch 130/878, Loss: 0.037607\nEpoch 6, Batch 140/878, Loss: 0.043162\nEpoch 6, Batch 150/878, Loss: 0.036668\nEpoch 6, Batch 160/878, Loss: 0.038376\nEpoch 6, Batch 170/878, Loss: 0.041290\nEpoch 6, Batch 180/878, Loss: 0.043137\nEpoch 6, Batch 190/878, Loss: 0.039967\nEpoch 6, Batch 200/878, Loss: 0.038798\nEpoch 6, Batch 210/878, Loss: 0.034690\nEpoch 6, Batch 220/878, Loss: 0.037469\nEpoch 6, Batch 230/878, Loss: 0.040020\nEpoch 6, Batch 240/878, Loss: 0.042599\nEpoch 6, Batch 250/878, Loss: 0.030670\nEpoch 6, Batch 260/878, Loss: 0.047975\nEpoch 6, Batch 270/878, Loss: 0.055594\nEpoch 6, Batch 280/878, Loss: 0.046539\nEpoch 6, Batch 290/878, Loss: 0.038890\nEpoch 6, Batch 300/878, Loss: 0.038482\nEpoch 6, Batch 310/878, Loss: 0.048920\nEpoch 6, Batch 320/878, Loss: 0.045662\nEpoch 6, Batch 330/878, Loss: 0.038745\nEpoch 6, Batch 340/878, Loss: 0.034428\nEpoch 6, Batch 350/878, Loss: 0.038969\nEpoch 6, Batch 360/878, Loss: 0.039788\nEpoch 6, Batch 370/878, Loss: 0.062731\nEpoch 6, Batch 380/878, Loss: 0.034976\nEpoch 6, Batch 390/878, Loss: 0.043021\nEpoch 6, Batch 400/878, Loss: 0.035754\nEpoch 6, Batch 410/878, Loss: 0.046908\nEpoch 6, Batch 420/878, Loss: 0.044035\nEpoch 6, Batch 430/878, Loss: 0.048777\nEpoch 6, Batch 440/878, Loss: 0.033400\nEpoch 6, Batch 450/878, Loss: 0.046317\nEpoch 6, Batch 460/878, Loss: 0.038379\nEpoch 6, Batch 470/878, Loss: 0.050846\nEpoch 6, Batch 480/878, Loss: 0.043111\nEpoch 6, Batch 490/878, Loss: 0.048373\nEpoch 6, Batch 500/878, Loss: 0.061394\nEpoch 6, Batch 510/878, Loss: 0.052201\nEpoch 6, Batch 520/878, Loss: 0.051114\nEpoch 6, Batch 530/878, Loss: 0.045644\nEpoch 6, Batch 540/878, Loss: 0.043355\nEpoch 6, Batch 550/878, Loss: 0.043661\nEpoch 6, Batch 560/878, Loss: 0.036637\nEpoch 6, Batch 570/878, Loss: 0.046828\nEpoch 6, Batch 580/878, Loss: 0.046015\nEpoch 6, Batch 590/878, Loss: 0.035015\nEpoch 6, Batch 600/878, Loss: 0.048292\nEpoch 6, Batch 610/878, Loss: 0.030099\nEpoch 6, Batch 620/878, Loss: 0.049895\nEpoch 6, Batch 630/878, Loss: 0.036451\nEpoch 6, Batch 640/878, Loss: 0.042926\nEpoch 6, Batch 650/878, Loss: 0.035089\nEpoch 6, Batch 660/878, Loss: 0.040608\nEpoch 6, Batch 670/878, Loss: 0.044994\nEpoch 6, Batch 680/878, Loss: 0.047994\nEpoch 6, Batch 690/878, Loss: 0.047483\nEpoch 6, Batch 700/878, Loss: 0.048608\nEpoch 6, Batch 710/878, Loss: 0.037543\nEpoch 6, Batch 720/878, Loss: 0.037660\nEpoch 6, Batch 730/878, Loss: 0.038851\nEpoch 6, Batch 740/878, Loss: 0.045982\nEpoch 6, Batch 750/878, Loss: 0.039674\nEpoch 6, Batch 760/878, Loss: 0.050045\nEpoch 6, Batch 770/878, Loss: 0.038957\nEpoch 6, Batch 780/878, Loss: 0.045345\nEpoch 6, Batch 790/878, Loss: 0.040861\nEpoch 6, Batch 800/878, Loss: 0.045601\nEpoch 6, Batch 810/878, Loss: 0.046522\nEpoch 6, Batch 820/878, Loss: 0.046864\nEpoch 6, Batch 830/878, Loss: 0.051534\nEpoch 6, Batch 840/878, Loss: 0.036025\nEpoch 6, Batch 850/878, Loss: 0.057749\nEpoch 6, Batch 860/878, Loss: 0.047674\nEpoch 6, Batch 870/878, Loss: 0.052815\nEpoch 6/30, Average Loss: 0.043781, Time: 461s\nNew best model saved with loss: 0.043781\nEpoch 7, Batch 0/878, Loss: 0.041113\nEpoch 7, Batch 10/878, Loss: 0.046168\nEpoch 7, Batch 20/878, Loss: 0.036352\nEpoch 7, Batch 30/878, Loss: 0.039435\nEpoch 7, Batch 40/878, Loss: 0.032642\nEpoch 7, Batch 50/878, Loss: 0.032971\nEpoch 7, Batch 60/878, Loss: 0.060046\nEpoch 7, Batch 70/878, Loss: 0.042847\nEpoch 7, Batch 80/878, Loss: 0.035899\nEpoch 7, Batch 90/878, Loss: 0.040849\nEpoch 7, Batch 100/878, Loss: 0.040085\nEpoch 7, Batch 110/878, Loss: 0.048868\nEpoch 7, Batch 120/878, Loss: 0.041797\nEpoch 7, Batch 130/878, Loss: 0.035070\nEpoch 7, Batch 140/878, Loss: 0.041525\nEpoch 7, Batch 150/878, Loss: 0.037403\nEpoch 7, Batch 160/878, Loss: 0.036495\nEpoch 7, Batch 170/878, Loss: 0.060531\nEpoch 7, Batch 180/878, Loss: 0.043812\nEpoch 7, Batch 190/878, Loss: 0.036202\nEpoch 7, Batch 200/878, Loss: 0.048706\nEpoch 7, Batch 210/878, Loss: 0.046665\nEpoch 7, Batch 220/878, Loss: 0.049332\nEpoch 7, Batch 230/878, Loss: 0.046313\nEpoch 7, Batch 240/878, Loss: 0.037920\nEpoch 7, Batch 250/878, Loss: 0.046351\nEpoch 7, Batch 260/878, Loss: 0.057756\nEpoch 7, Batch 270/878, Loss: 0.039357\nEpoch 7, Batch 280/878, Loss: 0.044888\nEpoch 7, Batch 290/878, Loss: 0.040309\nEpoch 7, Batch 300/878, Loss: 0.049102\nEpoch 7, Batch 310/878, Loss: 0.034968\nEpoch 7, Batch 320/878, Loss: 0.039628\nEpoch 7, Batch 330/878, Loss: 0.032026\nEpoch 7, Batch 340/878, Loss: 0.036734\nEpoch 7, Batch 350/878, Loss: 0.046350\nEpoch 7, Batch 360/878, Loss: 0.038368\nEpoch 7, Batch 370/878, Loss: 0.035014\nEpoch 7, Batch 380/878, Loss: 0.038089\nEpoch 7, Batch 390/878, Loss: 0.043261\nEpoch 7, Batch 400/878, Loss: 0.044773\nEpoch 7, Batch 410/878, Loss: 0.049988\nEpoch 7, Batch 420/878, Loss: 0.030187\nEpoch 7, Batch 430/878, Loss: 0.038871\nEpoch 7, Batch 440/878, Loss: 0.046570\nEpoch 7, Batch 450/878, Loss: 0.042052\nEpoch 7, Batch 460/878, Loss: 0.039644\nEpoch 7, Batch 470/878, Loss: 0.043293\nEpoch 7, Batch 480/878, Loss: 0.048425\nEpoch 7, Batch 490/878, Loss: 0.038876\nEpoch 7, Batch 500/878, Loss: 0.040976\nEpoch 7, Batch 510/878, Loss: 0.041356\nEpoch 7, Batch 520/878, Loss: 0.038955\nEpoch 7, Batch 530/878, Loss: 0.049977\nEpoch 7, Batch 540/878, Loss: 0.046146\nEpoch 7, Batch 550/878, Loss: 0.055792\nEpoch 7, Batch 560/878, Loss: 0.032927\nEpoch 7, Batch 570/878, Loss: 0.039965\nEpoch 7, Batch 580/878, Loss: 0.042503\nEpoch 7, Batch 590/878, Loss: 0.038737\nEpoch 7, Batch 600/878, Loss: 0.044814\nEpoch 7, Batch 610/878, Loss: 0.054069\nEpoch 7, Batch 620/878, Loss: 0.048734\nEpoch 7, Batch 630/878, Loss: 0.043369\nEpoch 7, Batch 640/878, Loss: 0.031061\nEpoch 7, Batch 650/878, Loss: 0.051754\nEpoch 7, Batch 660/878, Loss: 0.052809\nEpoch 7, Batch 670/878, Loss: 0.042398\nEpoch 7, Batch 680/878, Loss: 0.039116\nEpoch 7, Batch 690/878, Loss: 0.036944\nEpoch 7, Batch 700/878, Loss: 0.034155\nEpoch 7, Batch 710/878, Loss: 0.046754\nEpoch 7, Batch 720/878, Loss: 0.041534\nEpoch 7, Batch 730/878, Loss: 0.030952\nEpoch 7, Batch 740/878, Loss: 0.045826\nEpoch 7, Batch 750/878, Loss: 0.042700\nEpoch 7, Batch 760/878, Loss: 0.040835\nEpoch 7, Batch 770/878, Loss: 0.044574\nEpoch 7, Batch 780/878, Loss: 0.051542\nEpoch 7, Batch 790/878, Loss: 0.041562\nEpoch 7, Batch 800/878, Loss: 0.054353\nEpoch 7, Batch 810/878, Loss: 0.046240\nEpoch 7, Batch 820/878, Loss: 0.055220\nEpoch 7, Batch 830/878, Loss: 0.045925\nEpoch 7, Batch 840/878, Loss: 0.035952\nEpoch 7, Batch 850/878, Loss: 0.031880\nEpoch 7, Batch 860/878, Loss: 0.066938\nEpoch 7, Batch 870/878, Loss: 0.045178\nEpoch 7/30, Average Loss: 0.042864, Time: 462s\nNew best model saved with loss: 0.042864\nEpoch 8, Batch 0/878, Loss: 0.036173\nEpoch 8, Batch 10/878, Loss: 0.042611\nEpoch 8, Batch 20/878, Loss: 0.046787\nEpoch 8, Batch 30/878, Loss: 0.044715\nEpoch 8, Batch 40/878, Loss: 0.046863\nEpoch 8, Batch 50/878, Loss: 0.033984\nEpoch 8, Batch 60/878, Loss: 0.042699\nEpoch 8, Batch 70/878, Loss: 0.039098\nEpoch 8, Batch 80/878, Loss: 0.033000\nEpoch 8, Batch 90/878, Loss: 0.033910\nEpoch 8, Batch 100/878, Loss: 0.049833\nEpoch 8, Batch 110/878, Loss: 0.039506\nEpoch 8, Batch 120/878, Loss: 0.034383\nEpoch 8, Batch 130/878, Loss: 0.039930\nEpoch 8, Batch 140/878, Loss: 0.047971\nEpoch 8, Batch 150/878, Loss: 0.035048\nEpoch 8, Batch 160/878, Loss: 0.055595\nEpoch 8, Batch 170/878, Loss: 0.040235\nEpoch 8, Batch 180/878, Loss: 0.044020\nEpoch 8, Batch 190/878, Loss: 0.037249\nEpoch 8, Batch 200/878, Loss: 0.038411\nEpoch 8, Batch 210/878, Loss: 0.041122\nEpoch 8, Batch 220/878, Loss: 0.039481\nEpoch 8, Batch 230/878, Loss: 0.039005\nEpoch 8, Batch 240/878, Loss: 0.038384\nEpoch 8, Batch 250/878, Loss: 0.040399\nEpoch 8, Batch 260/878, Loss: 0.035444\nEpoch 8, Batch 270/878, Loss: 0.048495\nEpoch 8, Batch 280/878, Loss: 0.046002\nEpoch 8, Batch 290/878, Loss: 0.041506\nEpoch 8, Batch 300/878, Loss: 0.031138\nEpoch 8, Batch 310/878, Loss: 0.051246\nEpoch 8, Batch 320/878, Loss: 0.035439\nEpoch 8, Batch 330/878, Loss: 0.038448\nEpoch 8, Batch 340/878, Loss: 0.038957\nEpoch 8, Batch 350/878, Loss: 0.043997\nEpoch 8, Batch 360/878, Loss: 0.042384\nEpoch 8, Batch 370/878, Loss: 0.043066\nEpoch 8, Batch 380/878, Loss: 0.037389\nEpoch 8, Batch 390/878, Loss: 0.040676\nEpoch 8, Batch 400/878, Loss: 0.043460\nEpoch 8, Batch 410/878, Loss: 0.032236\nEpoch 8, Batch 420/878, Loss: 0.035181\nEpoch 8, Batch 430/878, Loss: 0.055995\nEpoch 8, Batch 440/878, Loss: 0.050590\nEpoch 8, Batch 450/878, Loss: 0.041152\nEpoch 8, Batch 460/878, Loss: 0.053856\nEpoch 8, Batch 470/878, Loss: 0.063210\nEpoch 8, Batch 480/878, Loss: 0.044223\nEpoch 8, Batch 490/878, Loss: 0.055899\nEpoch 8, Batch 500/878, Loss: 0.049081\nEpoch 8, Batch 510/878, Loss: 0.040666\nEpoch 8, Batch 520/878, Loss: 0.043054\nEpoch 8, Batch 530/878, Loss: 0.036684\nEpoch 8, Batch 540/878, Loss: 0.047454\nEpoch 8, Batch 550/878, Loss: 0.039849\nEpoch 8, Batch 560/878, Loss: 0.041567\nEpoch 8, Batch 570/878, Loss: 0.039204\nEpoch 8, Batch 580/878, Loss: 0.034704\nEpoch 8, Batch 590/878, Loss: 0.036418\nEpoch 8, Batch 600/878, Loss: 0.041367\nEpoch 8, Batch 610/878, Loss: 0.055559\nEpoch 8, Batch 620/878, Loss: 0.037828\nEpoch 8, Batch 630/878, Loss: 0.039906\nEpoch 8, Batch 640/878, Loss: 0.047461\nEpoch 8, Batch 650/878, Loss: 0.042271\nEpoch 8, Batch 660/878, Loss: 0.040550\nEpoch 8, Batch 670/878, Loss: 0.041411\nEpoch 8, Batch 680/878, Loss: 0.052560\nEpoch 8, Batch 690/878, Loss: 0.050591\nEpoch 8, Batch 700/878, Loss: 0.040480\nEpoch 8, Batch 710/878, Loss: 0.047519\nEpoch 8, Batch 720/878, Loss: 0.042119\nEpoch 8, Batch 730/878, Loss: 0.046662\nEpoch 8, Batch 740/878, Loss: 0.044302\nEpoch 8, Batch 750/878, Loss: 0.040941\nEpoch 8, Batch 760/878, Loss: 0.031503\nEpoch 8, Batch 770/878, Loss: 0.043035\nEpoch 8, Batch 780/878, Loss: 0.045472\nEpoch 8, Batch 790/878, Loss: 0.057181\nEpoch 8, Batch 800/878, Loss: 0.047791\nEpoch 8, Batch 810/878, Loss: 0.031640\nEpoch 8, Batch 820/878, Loss: 0.037883\nEpoch 8, Batch 830/878, Loss: 0.040188\nEpoch 8, Batch 840/878, Loss: 0.041939\nEpoch 8, Batch 850/878, Loss: 0.039702\nEpoch 8, Batch 860/878, Loss: 0.037165\nEpoch 8, Batch 870/878, Loss: 0.036605\nEpoch 8/30, Average Loss: 0.041883, Time: 462s\nNew best model saved with loss: 0.041883\nEpoch 9, Batch 0/878, Loss: 0.046447\nEpoch 9, Batch 10/878, Loss: 0.037989\nEpoch 9, Batch 20/878, Loss: 0.036111\nEpoch 9, Batch 30/878, Loss: 0.049499\nEpoch 9, Batch 40/878, Loss: 0.042732\nEpoch 9, Batch 50/878, Loss: 0.038709\nEpoch 9, Batch 60/878, Loss: 0.027709\nEpoch 9, Batch 70/878, Loss: 0.043843\nEpoch 9, Batch 80/878, Loss: 0.039262\nEpoch 9, Batch 90/878, Loss: 0.038639\nEpoch 9, Batch 100/878, Loss: 0.048651\nEpoch 9, Batch 110/878, Loss: 0.038482\nEpoch 9, Batch 120/878, Loss: 0.040439\nEpoch 9, Batch 130/878, Loss: 0.040985\nEpoch 9, Batch 140/878, Loss: 0.037291\nEpoch 9, Batch 150/878, Loss: 0.042777\nEpoch 9, Batch 160/878, Loss: 0.037966\nEpoch 9, Batch 170/878, Loss: 0.030566\nEpoch 9, Batch 180/878, Loss: 0.031572\nEpoch 9, Batch 190/878, Loss: 0.052738\nEpoch 9, Batch 200/878, Loss: 0.051417\nEpoch 9, Batch 210/878, Loss: 0.040366\nEpoch 9, Batch 220/878, Loss: 0.044445\nEpoch 9, Batch 230/878, Loss: 0.043062\nEpoch 9, Batch 240/878, Loss: 0.037806\nEpoch 9, Batch 250/878, Loss: 0.050052\nEpoch 9, Batch 260/878, Loss: 0.044729\nEpoch 9, Batch 270/878, Loss: 0.027100\nEpoch 9, Batch 280/878, Loss: 0.038032\nEpoch 9, Batch 290/878, Loss: 0.049742\nEpoch 9, Batch 300/878, Loss: 0.033770\nEpoch 9, Batch 310/878, Loss: 0.040829\nEpoch 9, Batch 320/878, Loss: 0.028524\nEpoch 9, Batch 330/878, Loss: 0.042588\nEpoch 9, Batch 340/878, Loss: 0.035049\nEpoch 9, Batch 350/878, Loss: 0.040418\nEpoch 9, Batch 360/878, Loss: 0.038265\nEpoch 9, Batch 370/878, Loss: 0.041238\nEpoch 9, Batch 380/878, Loss: 0.034463\nEpoch 9, Batch 390/878, Loss: 0.035436\nEpoch 9, Batch 400/878, Loss: 0.039747\nEpoch 9, Batch 410/878, Loss: 0.062850\nEpoch 9, Batch 420/878, Loss: 0.042611\nEpoch 9, Batch 430/878, Loss: 0.049373\nEpoch 9, Batch 440/878, Loss: 0.039677\nEpoch 9, Batch 450/878, Loss: 0.035729\nEpoch 9, Batch 460/878, Loss: 0.059242\nEpoch 9, Batch 470/878, Loss: 0.049465\nEpoch 9, Batch 480/878, Loss: 0.043076\nEpoch 9, Batch 490/878, Loss: 0.030292\nEpoch 9, Batch 500/878, Loss: 0.050850\nEpoch 9, Batch 510/878, Loss: 0.052682\nEpoch 9, Batch 520/878, Loss: 0.036512\nEpoch 9, Batch 530/878, Loss: 0.039479\nEpoch 9, Batch 540/878, Loss: 0.041677\nEpoch 9, Batch 550/878, Loss: 0.040012\nEpoch 9, Batch 560/878, Loss: 0.046782\nEpoch 9, Batch 570/878, Loss: 0.039332\nEpoch 9, Batch 580/878, Loss: 0.045663\nEpoch 9, Batch 590/878, Loss: 0.042167\nEpoch 9, Batch 600/878, Loss: 0.039280\nEpoch 9, Batch 610/878, Loss: 0.043486\nEpoch 9, Batch 620/878, Loss: 0.054275\nEpoch 9, Batch 630/878, Loss: 0.040109\nEpoch 9, Batch 640/878, Loss: 0.038940\nEpoch 9, Batch 650/878, Loss: 0.037567\nEpoch 9, Batch 660/878, Loss: 0.049905\nEpoch 9, Batch 670/878, Loss: 0.039658\nEpoch 9, Batch 680/878, Loss: 0.041397\nEpoch 9, Batch 690/878, Loss: 0.036251\nEpoch 9, Batch 700/878, Loss: 0.032988\nEpoch 9, Batch 710/878, Loss: 0.036523\nEpoch 9, Batch 720/878, Loss: 0.037645\nEpoch 9, Batch 730/878, Loss: 0.041959\nEpoch 9, Batch 740/878, Loss: 0.048103\nEpoch 9, Batch 750/878, Loss: 0.042160\nEpoch 9, Batch 760/878, Loss: 0.049024\nEpoch 9, Batch 770/878, Loss: 0.046785\nEpoch 9, Batch 780/878, Loss: 0.038837\nEpoch 9, Batch 790/878, Loss: 0.031423\nEpoch 9, Batch 800/878, Loss: 0.033105\nEpoch 9, Batch 810/878, Loss: 0.043547\nEpoch 9, Batch 820/878, Loss: 0.037407\nEpoch 9, Batch 830/878, Loss: 0.040705\nEpoch 9, Batch 840/878, Loss: 0.049135\nEpoch 9, Batch 850/878, Loss: 0.041442\nEpoch 9, Batch 860/878, Loss: 0.042839\nEpoch 9, Batch 870/878, Loss: 0.053555\nEpoch 9/30, Average Loss: 0.040358, Time: 462s\nNew best model saved with loss: 0.040358\nEpoch 10, Batch 0/878, Loss: 0.036324\nEpoch 10, Batch 10/878, Loss: 0.042439\nEpoch 10, Batch 20/878, Loss: 0.031097\nEpoch 10, Batch 30/878, Loss: 0.042996\nEpoch 10, Batch 40/878, Loss: 0.031672\nEpoch 10, Batch 50/878, Loss: 0.029738\nEpoch 10, Batch 60/878, Loss: 0.043435\nEpoch 10, Batch 70/878, Loss: 0.040346\nEpoch 10, Batch 80/878, Loss: 0.040834\nEpoch 10, Batch 90/878, Loss: 0.044875\nEpoch 10, Batch 100/878, Loss: 0.038158\nEpoch 10, Batch 110/878, Loss: 0.033947\nEpoch 10, Batch 120/878, Loss: 0.032994\nEpoch 10, Batch 130/878, Loss: 0.037429\nEpoch 10, Batch 140/878, Loss: 0.035561\nEpoch 10, Batch 150/878, Loss: 0.038484\nEpoch 10, Batch 160/878, Loss: 0.047069\nEpoch 10, Batch 170/878, Loss: 0.042732\nEpoch 10, Batch 180/878, Loss: 0.039406\nEpoch 10, Batch 190/878, Loss: 0.043483\nEpoch 10, Batch 200/878, Loss: 0.036512\nEpoch 10, Batch 210/878, Loss: 0.039469\nEpoch 10, Batch 220/878, Loss: 0.033919\nEpoch 10, Batch 230/878, Loss: 0.039014\nEpoch 10, Batch 240/878, Loss: 0.037108\nEpoch 10, Batch 250/878, Loss: 0.032700\nEpoch 10, Batch 260/878, Loss: 0.040449\nEpoch 10, Batch 270/878, Loss: 0.042594\nEpoch 10, Batch 280/878, Loss: 0.036688\nEpoch 10, Batch 290/878, Loss: 0.031245\nEpoch 10, Batch 300/878, Loss: 0.038569\nEpoch 10, Batch 310/878, Loss: 0.037924\nEpoch 10, Batch 320/878, Loss: 0.035827\nEpoch 10, Batch 330/878, Loss: 0.035488\nEpoch 10, Batch 340/878, Loss: 0.043241\nEpoch 10, Batch 350/878, Loss: 0.047641\nEpoch 10, Batch 360/878, Loss: 0.036955\nEpoch 10, Batch 370/878, Loss: 0.033954\nEpoch 10, Batch 380/878, Loss: 0.046676\nEpoch 10, Batch 390/878, Loss: 0.043035\nEpoch 10, Batch 400/878, Loss: 0.040871\nEpoch 10, Batch 410/878, Loss: 0.040395\nEpoch 10, Batch 420/878, Loss: 0.039682\nEpoch 10, Batch 430/878, Loss: 0.048010\nEpoch 10, Batch 440/878, Loss: 0.043435\nEpoch 10, Batch 450/878, Loss: 0.039255\nEpoch 10, Batch 460/878, Loss: 0.026518\nEpoch 10, Batch 470/878, Loss: 0.034053\nEpoch 10, Batch 480/878, Loss: 0.039156\nEpoch 10, Batch 490/878, Loss: 0.037612\nEpoch 10, Batch 500/878, Loss: 0.032632\nEpoch 10, Batch 510/878, Loss: 0.045059\nEpoch 10, Batch 520/878, Loss: 0.037084\nEpoch 10, Batch 530/878, Loss: 0.047693\nEpoch 10, Batch 540/878, Loss: 0.032443\nEpoch 10, Batch 550/878, Loss: 0.037341\nEpoch 10, Batch 560/878, Loss: 0.040491\nEpoch 10, Batch 570/878, Loss: 0.052289\nEpoch 10, Batch 580/878, Loss: 0.042037\nEpoch 10, Batch 590/878, Loss: 0.035654\nEpoch 10, Batch 600/878, Loss: 0.047242\nEpoch 10, Batch 610/878, Loss: 0.041438\nEpoch 10, Batch 620/878, Loss: 0.044229\nEpoch 10, Batch 630/878, Loss: 0.044767\nEpoch 10, Batch 640/878, Loss: 0.054447\nEpoch 10, Batch 650/878, Loss: 0.042536\nEpoch 10, Batch 660/878, Loss: 0.046414\nEpoch 10, Batch 670/878, Loss: 0.039035\nEpoch 10, Batch 680/878, Loss: 0.035110\nEpoch 10, Batch 690/878, Loss: 0.035108\nEpoch 10, Batch 700/878, Loss: 0.033733\nEpoch 10, Batch 710/878, Loss: 0.040647\nEpoch 10, Batch 720/878, Loss: 0.043432\nEpoch 10, Batch 730/878, Loss: 0.030601\nEpoch 10, Batch 740/878, Loss: 0.040383\nEpoch 10, Batch 750/878, Loss: 0.042173\nEpoch 10, Batch 760/878, Loss: 0.036123\nEpoch 10, Batch 770/878, Loss: 0.034376\nEpoch 10, Batch 780/878, Loss: 0.045673\nEpoch 10, Batch 790/878, Loss: 0.045359\nEpoch 10, Batch 800/878, Loss: 0.038922\nEpoch 10, Batch 810/878, Loss: 0.032799\nEpoch 10, Batch 820/878, Loss: 0.046748\nEpoch 10, Batch 830/878, Loss: 0.053457\nEpoch 10, Batch 840/878, Loss: 0.038126\nEpoch 10, Batch 850/878, Loss: 0.052958\nEpoch 10, Batch 860/878, Loss: 0.051131\nEpoch 10, Batch 870/878, Loss: 0.034676\nEpoch 10/30, Average Loss: 0.039521, Time: 462s\nNew best model saved with loss: 0.039521\nEpoch 11, Batch 0/878, Loss: 0.039265\nEpoch 11, Batch 10/878, Loss: 0.033231\nEpoch 11, Batch 20/878, Loss: 0.032856\nEpoch 11, Batch 30/878, Loss: 0.054382\nEpoch 11, Batch 40/878, Loss: 0.034261\nEpoch 11, Batch 50/878, Loss: 0.035267\nEpoch 11, Batch 60/878, Loss: 0.044764\nEpoch 11, Batch 70/878, Loss: 0.039883\nEpoch 11, Batch 80/878, Loss: 0.043643\nEpoch 11, Batch 90/878, Loss: 0.053943\nEpoch 11, Batch 100/878, Loss: 0.035342\nEpoch 11, Batch 110/878, Loss: 0.038694\nEpoch 11, Batch 120/878, Loss: 0.037135\nEpoch 11, Batch 130/878, Loss: 0.033225\nEpoch 11, Batch 140/878, Loss: 0.038917\nEpoch 11, Batch 150/878, Loss: 0.038378\nEpoch 11, Batch 160/878, Loss: 0.038543\nEpoch 11, Batch 170/878, Loss: 0.040409\nEpoch 11, Batch 180/878, Loss: 0.036219\nEpoch 11, Batch 190/878, Loss: 0.035411\nEpoch 11, Batch 200/878, Loss: 0.044342\nEpoch 11, Batch 210/878, Loss: 0.034784\nEpoch 11, Batch 220/878, Loss: 0.031724\nEpoch 11, Batch 230/878, Loss: 0.038646\nEpoch 11, Batch 240/878, Loss: 0.038228\nEpoch 11, Batch 250/878, Loss: 0.051679\nEpoch 11, Batch 260/878, Loss: 0.045850\nEpoch 11, Batch 270/878, Loss: 0.046360\nEpoch 11, Batch 280/878, Loss: 0.041447\nEpoch 11, Batch 290/878, Loss: 0.044341\nEpoch 11, Batch 300/878, Loss: 0.034936\nEpoch 11, Batch 310/878, Loss: 0.033923\nEpoch 11, Batch 320/878, Loss: 0.037302\nEpoch 11, Batch 330/878, Loss: 0.042827\nEpoch 11, Batch 340/878, Loss: 0.037664\nEpoch 11, Batch 350/878, Loss: 0.039984\nEpoch 11, Batch 360/878, Loss: 0.043082\nEpoch 11, Batch 370/878, Loss: 0.038793\nEpoch 11, Batch 380/878, Loss: 0.033677\nEpoch 11, Batch 390/878, Loss: 0.037246\nEpoch 11, Batch 400/878, Loss: 0.042424\nEpoch 11, Batch 410/878, Loss: 0.039515\nEpoch 11, Batch 420/878, Loss: 0.041342\nEpoch 11, Batch 430/878, Loss: 0.050711\nEpoch 11, Batch 440/878, Loss: 0.044588\nEpoch 11, Batch 450/878, Loss: 0.046786\nEpoch 11, Batch 460/878, Loss: 0.043541\nEpoch 11, Batch 470/878, Loss: 0.035037\nEpoch 11, Batch 480/878, Loss: 0.034414\nEpoch 11, Batch 490/878, Loss: 0.031803\nEpoch 11, Batch 500/878, Loss: 0.035658\nEpoch 11, Batch 510/878, Loss: 0.037195\nEpoch 11, Batch 520/878, Loss: 0.042551\nEpoch 11, Batch 530/878, Loss: 0.043026\nEpoch 11, Batch 540/878, Loss: 0.038772\nEpoch 11, Batch 550/878, Loss: 0.033421\nEpoch 11, Batch 560/878, Loss: 0.043359\nEpoch 11, Batch 570/878, Loss: 0.035802\nEpoch 11, Batch 580/878, Loss: 0.037373\nEpoch 11, Batch 590/878, Loss: 0.038370\nEpoch 11, Batch 600/878, Loss: 0.038939\nEpoch 11, Batch 610/878, Loss: 0.042661\nEpoch 11, Batch 620/878, Loss: 0.038101\nEpoch 11, Batch 630/878, Loss: 0.051158\nEpoch 11, Batch 640/878, Loss: 0.041976\nEpoch 11, Batch 650/878, Loss: 0.038153\nEpoch 11, Batch 660/878, Loss: 0.039191\nEpoch 11, Batch 670/878, Loss: 0.048839\nEpoch 11, Batch 680/878, Loss: 0.042404\nEpoch 11, Batch 690/878, Loss: 0.038141\nEpoch 11, Batch 700/878, Loss: 0.032289\nEpoch 11, Batch 710/878, Loss: 0.036508\nEpoch 11, Batch 720/878, Loss: 0.028144\nEpoch 11, Batch 730/878, Loss: 0.044761\nEpoch 11, Batch 740/878, Loss: 0.029625\nEpoch 11, Batch 750/878, Loss: 0.051345\nEpoch 11, Batch 760/878, Loss: 0.037240\nEpoch 11, Batch 770/878, Loss: 0.036203\nEpoch 11, Batch 780/878, Loss: 0.034767\nEpoch 11, Batch 790/878, Loss: 0.047739\nEpoch 11, Batch 800/878, Loss: 0.043872\nEpoch 11, Batch 810/878, Loss: 0.035446\nEpoch 11, Batch 820/878, Loss: 0.038246\nEpoch 11, Batch 830/878, Loss: 0.036225\nEpoch 11, Batch 840/878, Loss: 0.040581\nEpoch 11, Batch 850/878, Loss: 0.032896\nEpoch 11, Batch 860/878, Loss: 0.036062\nEpoch 11, Batch 870/878, Loss: 0.038742\nEpoch 11/30, Average Loss: 0.038712, Time: 462s\nNew best model saved with loss: 0.038712\nEpoch 12, Batch 0/878, Loss: 0.038440\nEpoch 12, Batch 10/878, Loss: 0.033232\nEpoch 12, Batch 20/878, Loss: 0.034629\nEpoch 12, Batch 30/878, Loss: 0.051914\nEpoch 12, Batch 40/878, Loss: 0.036157\nEpoch 12, Batch 50/878, Loss: 0.028676\nEpoch 12, Batch 60/878, Loss: 0.055104\nEpoch 12, Batch 70/878, Loss: 0.033121\nEpoch 12, Batch 80/878, Loss: 0.030809\nEpoch 12, Batch 90/878, Loss: 0.040820\nEpoch 12, Batch 100/878, Loss: 0.041263\nEpoch 12, Batch 110/878, Loss: 0.037975\nEpoch 12, Batch 120/878, Loss: 0.036465\nEpoch 12, Batch 130/878, Loss: 0.035709\nEpoch 12, Batch 140/878, Loss: 0.033589\nEpoch 12, Batch 150/878, Loss: 0.035063\nEpoch 12, Batch 160/878, Loss: 0.046702\nEpoch 12, Batch 170/878, Loss: 0.032790\nEpoch 12, Batch 180/878, Loss: 0.030952\nEpoch 12, Batch 190/878, Loss: 0.034892\nEpoch 12, Batch 200/878, Loss: 0.029184\nEpoch 12, Batch 210/878, Loss: 0.031876\nEpoch 12, Batch 220/878, Loss: 0.028865\nEpoch 12, Batch 230/878, Loss: 0.040739\nEpoch 12, Batch 240/878, Loss: 0.030085\nEpoch 12, Batch 250/878, Loss: 0.037796\nEpoch 12, Batch 260/878, Loss: 0.054257\nEpoch 12, Batch 270/878, Loss: 0.047977\nEpoch 12, Batch 280/878, Loss: 0.034325\nEpoch 12, Batch 290/878, Loss: 0.054247\nEpoch 12, Batch 300/878, Loss: 0.041098\nEpoch 12, Batch 310/878, Loss: 0.035976\nEpoch 12, Batch 320/878, Loss: 0.028159\nEpoch 12, Batch 330/878, Loss: 0.043077\nEpoch 12, Batch 340/878, Loss: 0.045153\nEpoch 12, Batch 350/878, Loss: 0.040407\nEpoch 12, Batch 360/878, Loss: 0.040482\nEpoch 12, Batch 370/878, Loss: 0.034014\nEpoch 12, Batch 380/878, Loss: 0.036316\nEpoch 12, Batch 390/878, Loss: 0.033324\nEpoch 12, Batch 400/878, Loss: 0.040951\nEpoch 12, Batch 410/878, Loss: 0.041600\nEpoch 12, Batch 420/878, Loss: 0.039386\nEpoch 12, Batch 430/878, Loss: 0.037145\nEpoch 12, Batch 440/878, Loss: 0.040830\nEpoch 12, Batch 450/878, Loss: 0.038097\nEpoch 12, Batch 460/878, Loss: 0.035803\nEpoch 12, Batch 470/878, Loss: 0.029374\nEpoch 12, Batch 480/878, Loss: 0.033440\nEpoch 12, Batch 490/878, Loss: 0.038677\nEpoch 12, Batch 500/878, Loss: 0.036833\nEpoch 12, Batch 510/878, Loss: 0.033129\nEpoch 12, Batch 520/878, Loss: 0.031070\nEpoch 12, Batch 530/878, Loss: 0.043298\nEpoch 12, Batch 540/878, Loss: 0.039363\nEpoch 12, Batch 550/878, Loss: 0.041522\nEpoch 12, Batch 560/878, Loss: 0.039666\nEpoch 12, Batch 570/878, Loss: 0.033337\nEpoch 12, Batch 580/878, Loss: 0.046225\nEpoch 12, Batch 590/878, Loss: 0.028807\nEpoch 12, Batch 600/878, Loss: 0.037726\nEpoch 12, Batch 610/878, Loss: 0.031331\nEpoch 12, Batch 620/878, Loss: 0.032516\nEpoch 12, Batch 630/878, Loss: 0.054854\nEpoch 12, Batch 640/878, Loss: 0.034308\nEpoch 12, Batch 650/878, Loss: 0.036590\nEpoch 12, Batch 660/878, Loss: 0.039162\nEpoch 12, Batch 670/878, Loss: 0.032192\nEpoch 12, Batch 680/878, Loss: 0.038797\nEpoch 12, Batch 690/878, Loss: 0.041486\nEpoch 12, Batch 700/878, Loss: 0.036349\nEpoch 12, Batch 710/878, Loss: 0.037286\nEpoch 12, Batch 720/878, Loss: 0.034302\nEpoch 12, Batch 730/878, Loss: 0.039218\nEpoch 12, Batch 740/878, Loss: 0.030389\nEpoch 12, Batch 750/878, Loss: 0.034994\nEpoch 12, Batch 760/878, Loss: 0.038711\nEpoch 12, Batch 770/878, Loss: 0.046181\nEpoch 12, Batch 780/878, Loss: 0.040555\nEpoch 12, Batch 790/878, Loss: 0.041314\nEpoch 12, Batch 800/878, Loss: 0.037106\nEpoch 12, Batch 810/878, Loss: 0.034830\nEpoch 12, Batch 820/878, Loss: 0.030426\nEpoch 12, Batch 830/878, Loss: 0.034196\nEpoch 12, Batch 840/878, Loss: 0.035678\nEpoch 12, Batch 850/878, Loss: 0.033063\nEpoch 12, Batch 860/878, Loss: 0.033839\nEpoch 12, Batch 870/878, Loss: 0.035089\nEpoch 12/30, Average Loss: 0.038013, Time: 462s\nNew best model saved with loss: 0.038013\nEpoch 13, Batch 0/878, Loss: 0.040194\nEpoch 13, Batch 10/878, Loss: 0.030584\nEpoch 13, Batch 20/878, Loss: 0.028351\nEpoch 13, Batch 30/878, Loss: 0.026567\nEpoch 13, Batch 40/878, Loss: 0.042024\nEpoch 13, Batch 50/878, Loss: 0.030012\nEpoch 13, Batch 60/878, Loss: 0.031451\nEpoch 13, Batch 70/878, Loss: 0.042331\nEpoch 13, Batch 80/878, Loss: 0.037968\nEpoch 13, Batch 90/878, Loss: 0.029963\nEpoch 13, Batch 100/878, Loss: 0.033772\nEpoch 13, Batch 110/878, Loss: 0.041648\nEpoch 13, Batch 120/878, Loss: 0.036985\nEpoch 13, Batch 130/878, Loss: 0.034366\nEpoch 13, Batch 140/878, Loss: 0.030770\nEpoch 13, Batch 150/878, Loss: 0.043325\nEpoch 13, Batch 160/878, Loss: 0.043838\nEpoch 13, Batch 170/878, Loss: 0.030829\nEpoch 13, Batch 180/878, Loss: 0.045710\nEpoch 13, Batch 190/878, Loss: 0.033379\nEpoch 13, Batch 200/878, Loss: 0.031364\nEpoch 13, Batch 210/878, Loss: 0.035069\nEpoch 13, Batch 220/878, Loss: 0.037878\nEpoch 13, Batch 230/878, Loss: 0.031028\nEpoch 13, Batch 240/878, Loss: 0.043102\nEpoch 13, Batch 250/878, Loss: 0.025593\nEpoch 13, Batch 260/878, Loss: 0.042553\nEpoch 13, Batch 270/878, Loss: 0.041301\nEpoch 13, Batch 280/878, Loss: 0.028562\nEpoch 13, Batch 290/878, Loss: 0.034953\nEpoch 13, Batch 300/878, Loss: 0.032674\nEpoch 13, Batch 310/878, Loss: 0.033583\nEpoch 13, Batch 320/878, Loss: 0.051902\nEpoch 13, Batch 330/878, Loss: 0.045545\nEpoch 13, Batch 340/878, Loss: 0.037857\nEpoch 13, Batch 350/878, Loss: 0.027130\nEpoch 13, Batch 360/878, Loss: 0.033054\nEpoch 13, Batch 370/878, Loss: 0.030700\nEpoch 13, Batch 380/878, Loss: 0.030426\nEpoch 13, Batch 390/878, Loss: 0.037240\nEpoch 13, Batch 400/878, Loss: 0.030657\nEpoch 13, Batch 410/878, Loss: 0.039890\nEpoch 13, Batch 420/878, Loss: 0.050374\nEpoch 13, Batch 430/878, Loss: 0.040566\nEpoch 13, Batch 440/878, Loss: 0.032558\nEpoch 13, Batch 450/878, Loss: 0.035092\nEpoch 13, Batch 460/878, Loss: 0.034037\nEpoch 13, Batch 470/878, Loss: 0.030363\nEpoch 13, Batch 480/878, Loss: 0.034757\nEpoch 13, Batch 490/878, Loss: 0.038420\nEpoch 13, Batch 500/878, Loss: 0.048212\nEpoch 13, Batch 510/878, Loss: 0.036144\nEpoch 13, Batch 520/878, Loss: 0.034682\nEpoch 13, Batch 530/878, Loss: 0.038316\nEpoch 13, Batch 540/878, Loss: 0.038973\nEpoch 13, Batch 550/878, Loss: 0.039053\nEpoch 13, Batch 560/878, Loss: 0.030288\nEpoch 13, Batch 570/878, Loss: 0.050752\nEpoch 13, Batch 580/878, Loss: 0.039999\nEpoch 13, Batch 590/878, Loss: 0.035040\nEpoch 13, Batch 600/878, Loss: 0.035749\nEpoch 13, Batch 610/878, Loss: 0.035055\nEpoch 13, Batch 620/878, Loss: 0.039557\nEpoch 13, Batch 630/878, Loss: 0.038800\nEpoch 13, Batch 640/878, Loss: 0.043424\nEpoch 13, Batch 650/878, Loss: 0.034307\nEpoch 13, Batch 660/878, Loss: 0.036428\nEpoch 13, Batch 670/878, Loss: 0.030204\nEpoch 13, Batch 680/878, Loss: 0.037306\nEpoch 13, Batch 690/878, Loss: 0.032872\nEpoch 13, Batch 700/878, Loss: 0.036235\nEpoch 13, Batch 710/878, Loss: 0.029115\nEpoch 13, Batch 720/878, Loss: 0.030763\nEpoch 13, Batch 730/878, Loss: 0.029592\nEpoch 13, Batch 740/878, Loss: 0.036261\nEpoch 13, Batch 750/878, Loss: 0.036401\nEpoch 13, Batch 760/878, Loss: 0.034583\nEpoch 13, Batch 770/878, Loss: 0.035340\nEpoch 13, Batch 780/878, Loss: 0.032030\nEpoch 13, Batch 790/878, Loss: 0.043549\nEpoch 13, Batch 800/878, Loss: 0.037685\nEpoch 13, Batch 810/878, Loss: 0.036289\nEpoch 13, Batch 820/878, Loss: 0.039532\nEpoch 13, Batch 830/878, Loss: 0.044256\nEpoch 13, Batch 840/878, Loss: 0.040567\nEpoch 13, Batch 850/878, Loss: 0.033086\nEpoch 13, Batch 860/878, Loss: 0.042483\nEpoch 13, Batch 870/878, Loss: 0.032995\nEpoch 13/30, Average Loss: 0.037272, Time: 462s\nNew best model saved with loss: 0.037272\nEpoch 14, Batch 0/878, Loss: 0.031426\nEpoch 14, Batch 10/878, Loss: 0.036861\nEpoch 14, Batch 20/878, Loss: 0.034101\nEpoch 14, Batch 30/878, Loss: 0.033847\nEpoch 14, Batch 40/878, Loss: 0.032067\nEpoch 14, Batch 50/878, Loss: 0.042558\nEpoch 14, Batch 60/878, Loss: 0.046704\nEpoch 14, Batch 70/878, Loss: 0.040089\nEpoch 14, Batch 80/878, Loss: 0.040417\nEpoch 14, Batch 90/878, Loss: 0.036034\nEpoch 14, Batch 100/878, Loss: 0.042966\nEpoch 14, Batch 110/878, Loss: 0.035668\nEpoch 14, Batch 120/878, Loss: 0.033929\nEpoch 14, Batch 130/878, Loss: 0.036744\nEpoch 14, Batch 140/878, Loss: 0.036577\nEpoch 14, Batch 150/878, Loss: 0.038804\nEpoch 14, Batch 160/878, Loss: 0.033952\nEpoch 14, Batch 170/878, Loss: 0.039097\nEpoch 14, Batch 180/878, Loss: 0.044662\nEpoch 14, Batch 190/878, Loss: 0.039089\nEpoch 14, Batch 200/878, Loss: 0.042031\nEpoch 14, Batch 210/878, Loss: 0.041909\nEpoch 14, Batch 220/878, Loss: 0.030956\nEpoch 14, Batch 230/878, Loss: 0.034185\nEpoch 14, Batch 240/878, Loss: 0.038755\nEpoch 14, Batch 250/878, Loss: 0.031971\nEpoch 14, Batch 260/878, Loss: 0.033502\nEpoch 14, Batch 270/878, Loss: 0.039684\nEpoch 14, Batch 280/878, Loss: 0.042161\nEpoch 14, Batch 290/878, Loss: 0.029279\nEpoch 14, Batch 300/878, Loss: 0.027340\nEpoch 14, Batch 310/878, Loss: 0.043136\nEpoch 14, Batch 320/878, Loss: 0.034358\nEpoch 14, Batch 330/878, Loss: 0.033781\nEpoch 14, Batch 340/878, Loss: 0.037565\nEpoch 14, Batch 350/878, Loss: 0.033345\nEpoch 14, Batch 360/878, Loss: 0.040141\nEpoch 14, Batch 370/878, Loss: 0.034964\nEpoch 14, Batch 380/878, Loss: 0.049500\nEpoch 14, Batch 390/878, Loss: 0.032023\nEpoch 14, Batch 400/878, Loss: 0.031673\nEpoch 14, Batch 410/878, Loss: 0.047735\nEpoch 14, Batch 420/878, Loss: 0.045904\nEpoch 14, Batch 430/878, Loss: 0.042552\nEpoch 14, Batch 440/878, Loss: 0.040670\nEpoch 14, Batch 450/878, Loss: 0.034832\nEpoch 14, Batch 460/878, Loss: 0.049154\nEpoch 14, Batch 470/878, Loss: 0.041126\nEpoch 14, Batch 480/878, Loss: 0.038286\nEpoch 14, Batch 490/878, Loss: 0.036046\nEpoch 14, Batch 500/878, Loss: 0.037912\nEpoch 14, Batch 510/878, Loss: 0.029338\nEpoch 14, Batch 520/878, Loss: 0.030937\nEpoch 14, Batch 530/878, Loss: 0.036575\nEpoch 14, Batch 540/878, Loss: 0.036004\nEpoch 14, Batch 550/878, Loss: 0.032210\nEpoch 14, Batch 560/878, Loss: 0.031188\nEpoch 14, Batch 570/878, Loss: 0.035498\nEpoch 14, Batch 580/878, Loss: 0.028886\nEpoch 14, Batch 590/878, Loss: 0.034982\nEpoch 14, Batch 600/878, Loss: 0.029250\nEpoch 14, Batch 610/878, Loss: 0.026299\nEpoch 14, Batch 620/878, Loss: 0.036032\nEpoch 14, Batch 630/878, Loss: 0.042458\nEpoch 14, Batch 640/878, Loss: 0.032323\nEpoch 14, Batch 650/878, Loss: 0.031471\nEpoch 14, Batch 660/878, Loss: 0.031322\nEpoch 14, Batch 670/878, Loss: 0.039810\nEpoch 14, Batch 680/878, Loss: 0.031362\nEpoch 14, Batch 690/878, Loss: 0.029216\nEpoch 14, Batch 700/878, Loss: 0.042618\nEpoch 14, Batch 710/878, Loss: 0.037068\nEpoch 14, Batch 720/878, Loss: 0.033505\nEpoch 14, Batch 730/878, Loss: 0.033585\nEpoch 14, Batch 740/878, Loss: 0.033040\nEpoch 14, Batch 750/878, Loss: 0.034021\nEpoch 14, Batch 760/878, Loss: 0.045591\nEpoch 14, Batch 770/878, Loss: 0.038526\nEpoch 14, Batch 780/878, Loss: 0.031192\nEpoch 14, Batch 790/878, Loss: 0.036500\nEpoch 14, Batch 800/878, Loss: 0.036690\nEpoch 14, Batch 810/878, Loss: 0.039728\nEpoch 14, Batch 820/878, Loss: 0.039848\nEpoch 14, Batch 830/878, Loss: 0.043739\nEpoch 14, Batch 840/878, Loss: 0.043779\nEpoch 14, Batch 850/878, Loss: 0.028020\nEpoch 14, Batch 860/878, Loss: 0.040768\nEpoch 14, Batch 870/878, Loss: 0.037376\nEpoch 14/30, Average Loss: 0.036574, Time: 463s\nNew best model saved with loss: 0.036574\nEpoch 15, Batch 0/878, Loss: 0.029528\nEpoch 15, Batch 10/878, Loss: 0.026207\nEpoch 15, Batch 20/878, Loss: 0.037951\nEpoch 15, Batch 30/878, Loss: 0.036372\nEpoch 15, Batch 40/878, Loss: 0.038182\nEpoch 15, Batch 50/878, Loss: 0.028781\nEpoch 15, Batch 60/878, Loss: 0.049371\nEpoch 15, Batch 70/878, Loss: 0.027118\nEpoch 15, Batch 80/878, Loss: 0.026653\nEpoch 15, Batch 90/878, Loss: 0.041152\nEpoch 15, Batch 100/878, Loss: 0.039979\nEpoch 15, Batch 110/878, Loss: 0.026366\nEpoch 15, Batch 120/878, Loss: 0.036153\nEpoch 15, Batch 130/878, Loss: 0.039860\nEpoch 15, Batch 140/878, Loss: 0.028186\nEpoch 15, Batch 150/878, Loss: 0.031665\nEpoch 15, Batch 160/878, Loss: 0.028528\nEpoch 15, Batch 170/878, Loss: 0.040304\nEpoch 15, Batch 180/878, Loss: 0.036968\nEpoch 15, Batch 190/878, Loss: 0.038464\nEpoch 15, Batch 200/878, Loss: 0.036176\nEpoch 15, Batch 210/878, Loss: 0.032873\nEpoch 15, Batch 220/878, Loss: 0.030711\nEpoch 15, Batch 230/878, Loss: 0.026135\nEpoch 15, Batch 460/878, Loss: 0.041372\nEpoch 15, Batch 470/878, Loss: 0.031680\nEpoch 15, Batch 480/878, Loss: 0.035471\nEpoch 15, Batch 490/878, Loss: 0.032223\nEpoch 15, Batch 500/878, Loss: 0.032733\nEpoch 15, Batch 510/878, Loss: 0.037250\nEpoch 15, Batch 520/878, Loss: 0.032359\nEpoch 15, Batch 530/878, Loss: 0.043256\nEpoch 15, Batch 540/878, Loss: 0.041330\nEpoch 15, Batch 550/878, Loss: 0.044034\nEpoch 15, Batch 560/878, Loss: 0.040863\nEpoch 15, Batch 570/878, Loss: 0.042176\nEpoch 15, Batch 580/878, Loss: 0.033421\nEpoch 15, Batch 590/878, Loss: 0.029638\nEpoch 15, Batch 600/878, Loss: 0.025038\nEpoch 15, Batch 610/878, Loss: 0.040134\nEpoch 15, Batch 620/878, Loss: 0.035377\nEpoch 15, Batch 630/878, Loss: 0.044890\nEpoch 15, Batch 640/878, Loss: 0.043180\nEpoch 15, Batch 650/878, Loss: 0.038645\nEpoch 15, Batch 660/878, Loss: 0.033556\nEpoch 15, Batch 670/878, Loss: 0.044397\nEpoch 15, Batch 680/878, Loss: 0.046426\nEpoch 15, Batch 690/878, Loss: 0.038715\nEpoch 15, Batch 700/878, Loss: 0.033488\nEpoch 15, Batch 710/878, Loss: 0.037547\nEpoch 15, Batch 720/878, Loss: 0.047071\nEpoch 15, Batch 730/878, Loss: 0.032027\nEpoch 15, Batch 740/878, Loss: 0.040332\nEpoch 15, Batch 750/878, Loss: 0.033303\nEpoch 15, Batch 760/878, Loss: 0.035823\nEpoch 15, Batch 770/878, Loss: 0.025192\nEpoch 15, Batch 780/878, Loss: 0.028148\nEpoch 15, Batch 790/878, Loss: 0.027227\nEpoch 15, Batch 800/878, Loss: 0.033068\nEpoch 15, Batch 810/878, Loss: 0.043319\nEpoch 15, Batch 820/878, Loss: 0.030685\nEpoch 15, Batch 830/878, Loss: 0.031514\nEpoch 15, Batch 840/878, Loss: 0.025333\nEpoch 15, Batch 850/878, Loss: 0.029483\nEpoch 15, Batch 860/878, Loss: 0.044723\nEpoch 15, Batch 870/878, Loss: 0.038644\nEpoch 15/30, Average Loss: 0.035876, Time: 463s\nNew best model saved with loss: 0.035876\nEpoch 16, Batch 0/878, Loss: 0.046675\nEpoch 16, Batch 10/878, Loss: 0.035250\nEpoch 16, Batch 20/878, Loss: 0.035899\nEpoch 16, Batch 30/878, Loss: 0.023071\nEpoch 16, Batch 40/878, Loss: 0.041931\nEpoch 16, Batch 50/878, Loss: 0.035705\nEpoch 16, Batch 60/878, Loss: 0.031242\nEpoch 16, Batch 70/878, Loss: 0.036145\nEpoch 16, Batch 80/878, Loss: 0.035455\nEpoch 16, Batch 90/878, Loss: 0.039340\nEpoch 16, Batch 100/878, Loss: 0.037792\nEpoch 16, Batch 110/878, Loss: 0.034939\nEpoch 16, Batch 120/878, Loss: 0.031554\nEpoch 16, Batch 130/878, Loss: 0.032538\nEpoch 16, Batch 140/878, Loss: 0.034324\nEpoch 16, Batch 150/878, Loss: 0.029296\nEpoch 16, Batch 160/878, Loss: 0.035988\nEpoch 16, Batch 170/878, Loss: 0.032032\nEpoch 16, Batch 180/878, Loss: 0.039744\nEpoch 16, Batch 190/878, Loss: 0.035057\nEpoch 16, Batch 200/878, Loss: 0.040689\nEpoch 16, Batch 210/878, Loss: 0.035753\nEpoch 16, Batch 220/878, Loss: 0.036166\nEpoch 16, Batch 230/878, Loss: 0.037785\nEpoch 16, Batch 240/878, Loss: 0.024338\nEpoch 16, Batch 250/878, Loss: 0.034315\nEpoch 16, Batch 260/878, Loss: 0.032425\nEpoch 16, Batch 270/878, Loss: 0.033033\nEpoch 16, Batch 280/878, Loss: 0.029896\nEpoch 16, Batch 290/878, Loss: 0.037974\nEpoch 16, Batch 300/878, Loss: 0.037582\nEpoch 16, Batch 310/878, Loss: 0.040198\nEpoch 16, Batch 320/878, Loss: 0.037025\nEpoch 16, Batch 330/878, Loss: 0.033114\nEpoch 16, Batch 340/878, Loss: 0.042226\nEpoch 16, Batch 350/878, Loss: 0.047811\nEpoch 16, Batch 360/878, Loss: 0.041418\nEpoch 16, Batch 370/878, Loss: 0.040638\nEpoch 16, Batch 380/878, Loss: 0.039588\nEpoch 16, Batch 390/878, Loss: 0.035137\nEpoch 16, Batch 400/878, Loss: 0.041646\nEpoch 16, Batch 410/878, Loss: 0.031086\nEpoch 16, Batch 420/878, Loss: 0.030382\nEpoch 16, Batch 430/878, Loss: 0.033508\nEpoch 16, Batch 440/878, Loss: 0.027809\nEpoch 16, Batch 450/878, Loss: 0.036930\nEpoch 16, Batch 460/878, Loss: 0.032553\nEpoch 16, Batch 470/878, Loss: 0.044152\nEpoch 16, Batch 480/878, Loss: 0.036858\nEpoch 16, Batch 490/878, Loss: 0.028403\nEpoch 16, Batch 500/878, Loss: 0.034570\nEpoch 16, Batch 510/878, Loss: 0.039729\nEpoch 16, Batch 520/878, Loss: 0.040193\nEpoch 16, Batch 530/878, Loss: 0.029724\nEpoch 16, Batch 540/878, Loss: 0.031791\nEpoch 16, Batch 550/878, Loss: 0.042375\nEpoch 16, Batch 560/878, Loss: 0.031370\nEpoch 16, Batch 570/878, Loss: 0.034140\nEpoch 16, Batch 580/878, Loss: 0.025177\nEpoch 16, Batch 590/878, Loss: 0.039284\nEpoch 16, Batch 600/878, Loss: 0.035601\nEpoch 16, Batch 610/878, Loss: 0.041899\nEpoch 16, Batch 620/878, Loss: 0.040089\nEpoch 16, Batch 630/878, Loss: 0.037379\nEpoch 16, Batch 640/878, Loss: 0.039123\nEpoch 16, Batch 650/878, Loss: 0.039364\nEpoch 16, Batch 660/878, Loss: 0.038781\nEpoch 16, Batch 670/878, Loss: 0.037526\nEpoch 16, Batch 680/878, Loss: 0.048064\nEpoch 16, Batch 690/878, Loss: 0.030264\nEpoch 16, Batch 700/878, Loss: 0.027297\nEpoch 16, Batch 710/878, Loss: 0.032794\nEpoch 16, Batch 720/878, Loss: 0.033581\nEpoch 16, Batch 730/878, Loss: 0.042366\nEpoch 16, Batch 740/878, Loss: 0.046975\nEpoch 16, Batch 750/878, Loss: 0.029918\nEpoch 16, Batch 760/878, Loss: 0.037526\nEpoch 16, Batch 770/878, Loss: 0.023544\nEpoch 16, Batch 780/878, Loss: 0.035092\nEpoch 16, Batch 790/878, Loss: 0.039103\nEpoch 16, Batch 800/878, Loss: 0.032423\nEpoch 16, Batch 810/878, Loss: 0.031593\nEpoch 16, Batch 820/878, Loss: 0.036198\nEpoch 16, Batch 830/878, Loss: 0.037924\nEpoch 16, Batch 840/878, Loss: 0.034587\nEpoch 16, Batch 850/878, Loss: 0.029212\nEpoch 16, Batch 860/878, Loss: 0.040161\nEpoch 16, Batch 870/878, Loss: 0.031619\nEpoch 16/30, Average Loss: 0.035206, Time: 462s\nNew best model saved with loss: 0.035206\nEpoch 17, Batch 0/878, Loss: 0.033420\nEpoch 17, Batch 10/878, Loss: 0.029327\nEpoch 17, Batch 20/878, Loss: 0.030192\nEpoch 17, Batch 30/878, Loss: 0.037371\nEpoch 17, Batch 40/878, Loss: 0.039384\nEpoch 17, Batch 50/878, Loss: 0.046635\nEpoch 17, Batch 60/878, Loss: 0.037769\nEpoch 17, Batch 70/878, Loss: 0.033091\nEpoch 17, Batch 80/878, Loss: 0.042599\nEpoch 17, Batch 90/878, Loss: 0.029065\nEpoch 17, Batch 100/878, Loss: 0.032402\nEpoch 17, Batch 110/878, Loss: 0.030669\nEpoch 17, Batch 120/878, Loss: 0.034868\nEpoch 17, Batch 130/878, Loss: 0.044144\nEpoch 17, Batch 140/878, Loss: 0.043398\nEpoch 17, Batch 150/878, Loss: 0.033501\nEpoch 17, Batch 160/878, Loss: 0.037778\nEpoch 17, Batch 170/878, Loss: 0.030526\nEpoch 17, Batch 180/878, Loss: 0.028030\nEpoch 17, Batch 190/878, Loss: 0.034275\nEpoch 17, Batch 200/878, Loss: 0.026985\nEpoch 17, Batch 210/878, Loss: 0.028501\nEpoch 17, Batch 220/878, Loss: 0.036680\nEpoch 17, Batch 230/878, Loss: 0.028599\nEpoch 17, Batch 240/878, Loss: 0.034423\nEpoch 17, Batch 250/878, Loss: 0.034986\nEpoch 17, Batch 260/878, Loss: 0.026854\nEpoch 17, Batch 270/878, Loss: 0.038317\nEpoch 17, Batch 280/878, Loss: 0.028772\nEpoch 17, Batch 290/878, Loss: 0.039777\nEpoch 17, Batch 300/878, Loss: 0.033390\nEpoch 17, Batch 310/878, Loss: 0.033600\nEpoch 17, Batch 320/878, Loss: 0.032045\nEpoch 17, Batch 330/878, Loss: 0.035895\nEpoch 17, Batch 340/878, Loss: 0.033306\nEpoch 17, Batch 350/878, Loss: 0.037463\nEpoch 17, Batch 360/878, Loss: 0.038143\nEpoch 17, Batch 370/878, Loss: 0.036242\nEpoch 17, Batch 380/878, Loss: 0.031339\nEpoch 17, Batch 390/878, Loss: 0.028838\nEpoch 17, Batch 400/878, Loss: 0.036052\nEpoch 17, Batch 410/878, Loss: 0.034120\nEpoch 17, Batch 420/878, Loss: 0.039684\nEpoch 17, Batch 430/878, Loss: 0.040226\nEpoch 17, Batch 440/878, Loss: 0.046581\nEpoch 17, Batch 450/878, Loss: 0.031890\nEpoch 17, Batch 460/878, Loss: 0.027482\nEpoch 17, Batch 470/878, Loss: 0.029583\nEpoch 17, Batch 480/878, Loss: 0.038120\nEpoch 17, Batch 490/878, Loss: 0.033112\nEpoch 17, Batch 500/878, Loss: 0.034001\nEpoch 17, Batch 510/878, Loss: 0.040010\nEpoch 17, Batch 520/878, Loss: 0.031753\nEpoch 17, Batch 530/878, Loss: 0.028034\nEpoch 17, Batch 540/878, Loss: 0.034547\nEpoch 17, Batch 550/878, Loss: 0.038796\nEpoch 17, Batch 560/878, Loss: 0.031001\nEpoch 17, Batch 570/878, Loss: 0.043222\nEpoch 17, Batch 580/878, Loss: 0.029294\nEpoch 17, Batch 590/878, Loss: 0.023114\nEpoch 17, Batch 600/878, Loss: 0.030954\nEpoch 17, Batch 610/878, Loss: 0.033508\nEpoch 17, Batch 620/878, Loss: 0.035548\nEpoch 17, Batch 630/878, Loss: 0.037567\nEpoch 17, Batch 640/878, Loss: 0.036011\nEpoch 17, Batch 650/878, Loss: 0.037591\nEpoch 17, Batch 660/878, Loss: 0.035543\nEpoch 17, Batch 670/878, Loss: 0.035405\nEpoch 17, Batch 680/878, Loss: 0.043995\nEpoch 17, Batch 690/878, Loss: 0.035678\nEpoch 17, Batch 700/878, Loss: 0.039689\nEpoch 17, Batch 710/878, Loss: 0.037543\nEpoch 17, Batch 720/878, Loss: 0.032202\nEpoch 17, Batch 730/878, Loss: 0.027493\nEpoch 17, Batch 740/878, Loss: 0.028420\nEpoch 17, Batch 750/878, Loss: 0.025929\nEpoch 17, Batch 760/878, Loss: 0.032497\nEpoch 17, Batch 770/878, Loss: 0.034832\nEpoch 17, Batch 780/878, Loss: 0.034211\nEpoch 17, Batch 790/878, Loss: 0.032880\nEpoch 17, Batch 800/878, Loss: 0.033891\nEpoch 17, Batch 810/878, Loss: 0.034165\nEpoch 17, Batch 820/878, Loss: 0.034582\nEpoch 17, Batch 830/878, Loss: 0.032282\nEpoch 17, Batch 840/878, Loss: 0.036680\nEpoch 17, Batch 850/878, Loss: 0.029906\nEpoch 17, Batch 860/878, Loss: 0.037877\nEpoch 17, Batch 870/878, Loss: 0.032711\nEpoch 17/30, Average Loss: 0.034323, Time: 464s\nNew best model saved with loss: 0.034323\nEpoch 18, Batch 0/878, Loss: 0.034531\nEpoch 18, Batch 10/878, Loss: 0.029158\nEpoch 18, Batch 20/878, Loss: 0.032852\nEpoch 18, Batch 30/878, Loss: 0.033096\nEpoch 18, Batch 40/878, Loss: 0.036247\nEpoch 18, Batch 50/878, Loss: 0.043119\nEpoch 18, Batch 60/878, Loss: 0.039461\nEpoch 18, Batch 70/878, Loss: 0.027102\nEpoch 18, Batch 80/878, Loss: 0.043446\nEpoch 18, Batch 90/878, Loss: 0.033375\nEpoch 18, Batch 100/878, Loss: 0.032814\nEpoch 18, Batch 110/878, Loss: 0.026363\nEpoch 18, Batch 120/878, Loss: 0.032134\nEpoch 18, Batch 130/878, Loss: 0.039441\nEpoch 18, Batch 140/878, Loss: 0.032713\nEpoch 18, Batch 150/878, Loss: 0.034471\nEpoch 18, Batch 160/878, Loss: 0.046306\nEpoch 18, Batch 170/878, Loss: 0.034419\nEpoch 18, Batch 180/878, Loss: 0.033688\nEpoch 18, Batch 190/878, Loss: 0.036153\nEpoch 18, Batch 200/878, Loss: 0.038361\nEpoch 18, Batch 210/878, Loss: 0.028961\nEpoch 18, Batch 220/878, Loss: 0.034912\nEpoch 18, Batch 230/878, Loss: 0.039375\nEpoch 18, Batch 240/878, Loss: 0.029414\nEpoch 18, Batch 250/878, Loss: 0.031038\nEpoch 18, Batch 260/878, Loss: 0.032688\nEpoch 18, Batch 270/878, Loss: 0.045320\nEpoch 18, Batch 280/878, Loss: 0.036659\nEpoch 18, Batch 290/878, Loss: 0.038487\nEpoch 18, Batch 300/878, Loss: 0.035561\nEpoch 18, Batch 310/878, Loss: 0.034282\nEpoch 18, Batch 320/878, Loss: 0.025982\nEpoch 18, Batch 330/878, Loss: 0.029509\nEpoch 18, Batch 340/878, Loss: 0.035304\nEpoch 18, Batch 350/878, Loss: 0.039790\nEpoch 18, Batch 360/878, Loss: 0.043850\nEpoch 18, Batch 370/878, Loss: 0.030056\nEpoch 18, Batch 380/878, Loss: 0.032494\nEpoch 18, Batch 390/878, Loss: 0.030538\nEpoch 18, Batch 400/878, Loss: 0.028588\nEpoch 18, Batch 410/878, Loss: 0.027003\nEpoch 18, Batch 420/878, Loss: 0.028579\nEpoch 18, Batch 430/878, Loss: 0.037632\nEpoch 18, Batch 440/878, Loss: 0.027552\nEpoch 18, Batch 450/878, Loss: 0.039293\nEpoch 18, Batch 460/878, Loss: 0.029103\nEpoch 18, Batch 470/878, Loss: 0.030360\nEpoch 18, Batch 480/878, Loss: 0.039930\nEpoch 18, Batch 490/878, Loss: 0.031315\nEpoch 18, Batch 500/878, Loss: 0.032802\nEpoch 18, Batch 510/878, Loss: 0.034895\nEpoch 18, Batch 520/878, Loss: 0.033238\nEpoch 18, Batch 530/878, Loss: 0.033450\nEpoch 18, Batch 540/878, Loss: 0.039421\nEpoch 18, Batch 550/878, Loss: 0.041247\nEpoch 18, Batch 560/878, Loss: 0.039030\nEpoch 18, Batch 570/878, Loss: 0.029978\nEpoch 18, Batch 580/878, Loss: 0.031941\nEpoch 18, Batch 590/878, Loss: 0.032577\nEpoch 18, Batch 600/878, Loss: 0.029482\nEpoch 18, Batch 610/878, Loss: 0.032777\nEpoch 18, Batch 620/878, Loss: 0.030183\nEpoch 18, Batch 630/878, Loss: 0.033395\nEpoch 18, Batch 640/878, Loss: 0.032006\nEpoch 18, Batch 650/878, Loss: 0.027336\nEpoch 18, Batch 660/878, Loss: 0.034601\nEpoch 18, Batch 670/878, Loss: 0.024127\nEpoch 18, Batch 680/878, Loss: 0.038311\nEpoch 18, Batch 690/878, Loss: 0.033230\nEpoch 18, Batch 700/878, Loss: 0.031057\nEpoch 18, Batch 710/878, Loss: 0.033517\nEpoch 18, Batch 720/878, Loss: 0.030636\nEpoch 18, Batch 730/878, Loss: 0.031099\nEpoch 18, Batch 740/878, Loss: 0.032329\nEpoch 18, Batch 750/878, Loss: 0.035559\nEpoch 18, Batch 760/878, Loss: 0.046967\nEpoch 18, Batch 770/878, Loss: 0.030325\nEpoch 18, Batch 780/878, Loss: 0.024782\nEpoch 18, Batch 790/878, Loss: 0.041272\nEpoch 18, Batch 800/878, Loss: 0.028733\nEpoch 18, Batch 810/878, Loss: 0.029766\nEpoch 18, Batch 820/878, Loss: 0.034885\nEpoch 18, Batch 830/878, Loss: 0.027281\nEpoch 18, Batch 840/878, Loss: 0.028052\nEpoch 18, Batch 850/878, Loss: 0.030902\nEpoch 18, Batch 860/878, Loss: 0.041805\nEpoch 18, Batch 870/878, Loss: 0.029412\nEpoch 18/30, Average Loss: 0.033897, Time: 464s\nNew best model saved with loss: 0.033897\nEpoch 19, Batch 0/878, Loss: 0.041419\nEpoch 19, Batch 10/878, Loss: 0.029387\nEpoch 19, Batch 20/878, Loss: 0.029587\nEpoch 19, Batch 30/878, Loss: 0.033703\nEpoch 19, Batch 40/878, Loss: 0.027851\nEpoch 19, Batch 50/878, Loss: 0.041449\nEpoch 19, Batch 60/878, Loss: 0.038318\nEpoch 19, Batch 70/878, Loss: 0.043541\nEpoch 19, Batch 80/878, Loss: 0.029564\nEpoch 19, Batch 90/878, Loss: 0.039680\nEpoch 19, Batch 100/878, Loss: 0.027280\nEpoch 19, Batch 110/878, Loss: 0.032447\nEpoch 19, Batch 120/878, Loss: 0.031142\nEpoch 19, Batch 130/878, Loss: 0.045421\nEpoch 19, Batch 140/878, Loss: 0.032661\nEpoch 19, Batch 150/878, Loss: 0.034983\nEpoch 19, Batch 160/878, Loss: 0.034684\nEpoch 19, Batch 170/878, Loss: 0.031751\nEpoch 19, Batch 180/878, Loss: 0.034070\nEpoch 19, Batch 190/878, Loss: 0.033831\nEpoch 19, Batch 200/878, Loss: 0.035289\nEpoch 19, Batch 210/878, Loss: 0.027515\nEpoch 19, Batch 220/878, Loss: 0.035429\nEpoch 19, Batch 230/878, Loss: 0.029889\nEpoch 19, Batch 240/878, Loss: 0.032898\nEpoch 19, Batch 250/878, Loss: 0.036011\nEpoch 19, Batch 260/878, Loss: 0.031624\nEpoch 19, Batch 270/878, Loss: 0.031362\nEpoch 19, Batch 280/878, Loss: 0.027098\nEpoch 19, Batch 290/878, Loss: 0.036269\nEpoch 19, Batch 300/878, Loss: 0.034284\nEpoch 19, Batch 310/878, Loss: 0.024581\nEpoch 19, Batch 320/878, Loss: 0.034611\nEpoch 19, Batch 330/878, Loss: 0.043143\nEpoch 19, Batch 340/878, Loss: 0.034014\nEpoch 19, Batch 350/878, Loss: 0.042148\nEpoch 19, Batch 360/878, Loss: 0.032986\nEpoch 19, Batch 370/878, Loss: 0.030849\nEpoch 19, Batch 380/878, Loss: 0.032120\nEpoch 19, Batch 390/878, Loss: 0.027187\nEpoch 19, Batch 400/878, Loss: 0.033677\nEpoch 19, Batch 410/878, Loss: 0.034795\nEpoch 19, Batch 420/878, Loss: 0.033301\nEpoch 19, Batch 430/878, Loss: 0.038804\nEpoch 19, Batch 440/878, Loss: 0.028932\nEpoch 19, Batch 450/878, Loss: 0.038336\nEpoch 19, Batch 460/878, Loss: 0.035146\nEpoch 19, Batch 470/878, Loss: 0.042813\nEpoch 19, Batch 480/878, Loss: 0.030634\nEpoch 19, Batch 490/878, Loss: 0.035973\nEpoch 19, Batch 500/878, Loss: 0.036155\nEpoch 19, Batch 510/878, Loss: 0.034124\nEpoch 19, Batch 520/878, Loss: 0.056868\nEpoch 19, Batch 530/878, Loss: 0.036647\nEpoch 19, Batch 540/878, Loss: 0.032060\nEpoch 19, Batch 550/878, Loss: 0.033050\nEpoch 19, Batch 560/878, Loss: 0.028839\nEpoch 19, Batch 570/878, Loss: 0.033302\nEpoch 19, Batch 580/878, Loss: 0.024518\nEpoch 19, Batch 590/878, Loss: 0.030718\nEpoch 19, Batch 600/878, Loss: 0.034072\nEpoch 19, Batch 610/878, Loss: 0.032753\nEpoch 19, Batch 620/878, Loss: 0.030489\nEpoch 19, Batch 630/878, Loss: 0.034267\nEpoch 19, Batch 640/878, Loss: 0.030600\nEpoch 19, Batch 650/878, Loss: 0.025540\nEpoch 19, Batch 660/878, Loss: 0.033645\nEpoch 19, Batch 670/878, Loss: 0.034906\nEpoch 19, Batch 680/878, Loss: 0.033630\nEpoch 19, Batch 690/878, Loss: 0.035938\nEpoch 19, Batch 700/878, Loss: 0.032800\nEpoch 19, Batch 710/878, Loss: 0.030861\nEpoch 19, Batch 720/878, Loss: 0.040795\nEpoch 19, Batch 730/878, Loss: 0.024748\nEpoch 19, Batch 740/878, Loss: 0.029564\nEpoch 19, Batch 750/878, Loss: 0.032081\nEpoch 19, Batch 760/878, Loss: 0.028453\nEpoch 19, Batch 770/878, Loss: 0.032850\nEpoch 19, Batch 780/878, Loss: 0.024445\nEpoch 19, Batch 790/878, Loss: 0.029202\nEpoch 19, Batch 800/878, Loss: 0.029465\nEpoch 19, Batch 810/878, Loss: 0.038966\nEpoch 19, Batch 820/878, Loss: 0.031593\nEpoch 19, Batch 830/878, Loss: 0.036649\nEpoch 19, Batch 840/878, Loss: 0.025028\nEpoch 19, Batch 850/878, Loss: 0.034152\nEpoch 19, Batch 860/878, Loss: 0.037107\nEpoch 19, Batch 870/878, Loss: 0.035484\nEpoch 19/30, Average Loss: 0.033610, Time: 465s\nNew best model saved with loss: 0.033610\nEpoch 20, Batch 0/878, Loss: 0.033651\nEpoch 20, Batch 10/878, Loss: 0.041934\nEpoch 20, Batch 20/878, Loss: 0.040450\nEpoch 20, Batch 30/878, Loss: 0.038965\nEpoch 20, Batch 40/878, Loss: 0.033899\nEpoch 20, Batch 50/878, Loss: 0.059110\nEpoch 20, Batch 60/878, Loss: 0.037103\nEpoch 20, Batch 70/878, Loss: 0.036771\nEpoch 20, Batch 80/878, Loss: 0.040998\nEpoch 20, Batch 90/878, Loss: 0.037496\nEpoch 20, Batch 100/878, Loss: 0.033721\nEpoch 20, Batch 110/878, Loss: 0.031476\nEpoch 20, Batch 120/878, Loss: 0.033243\nEpoch 20, Batch 130/878, Loss: 0.029649\nEpoch 20, Batch 140/878, Loss: 0.044916\nEpoch 20, Batch 150/878, Loss: 0.032823\nEpoch 20, Batch 160/878, Loss: 0.040719\nEpoch 20, Batch 170/878, Loss: 0.026985\nEpoch 20, Batch 180/878, Loss: 0.037010\nEpoch 20, Batch 190/878, Loss: 0.036591\nEpoch 20, Batch 200/878, Loss: 0.036063\nEpoch 20, Batch 210/878, Loss: 0.036887\nEpoch 20, Batch 220/878, Loss: 0.033851\nEpoch 20, Batch 230/878, Loss: 0.038501\nEpoch 20, Batch 240/878, Loss: 0.035437\nEpoch 20, Batch 250/878, Loss: 0.038012\nEpoch 20, Batch 260/878, Loss: 0.027972\nEpoch 20, Batch 270/878, Loss: 0.027551\nEpoch 20, Batch 280/878, Loss: 0.030831\nEpoch 20, Batch 290/878, Loss: 0.034560\nEpoch 20, Batch 300/878, Loss: 0.032579\nEpoch 20, Batch 310/878, Loss: 0.038576\nEpoch 20, Batch 320/878, Loss: 0.032831\nEpoch 20, Batch 330/878, Loss: 0.039595\nEpoch 20, Batch 340/878, Loss: 0.035287\nEpoch 20, Batch 350/878, Loss: 0.034603\nEpoch 20, Batch 360/878, Loss: 0.034893\nEpoch 20, Batch 370/878, Loss: 0.029995\nEpoch 20, Batch 380/878, Loss: 0.033962\nEpoch 20, Batch 390/878, Loss: 0.028682\nEpoch 20, Batch 400/878, Loss: 0.024932\nEpoch 20, Batch 410/878, Loss: 0.030738\nEpoch 20, Batch 420/878, Loss: 0.030593\nEpoch 20, Batch 430/878, Loss: 0.028123\nEpoch 20, Batch 440/878, Loss: 0.024921\nEpoch 20, Batch 450/878, Loss: 0.030946\nEpoch 20, Batch 460/878, Loss: 0.033546\nEpoch 20, Batch 470/878, Loss: 0.028636\nEpoch 20, Batch 480/878, Loss: 0.032059\nEpoch 20, Batch 490/878, Loss: 0.026657\nEpoch 20, Batch 500/878, Loss: 0.033211\nEpoch 20, Batch 510/878, Loss: 0.028916\nEpoch 20, Batch 520/878, Loss: 0.029526\nEpoch 20, Batch 530/878, Loss: 0.032340\nEpoch 20, Batch 540/878, Loss: 0.029775\nEpoch 20, Batch 550/878, Loss: 0.027188\nEpoch 20, Batch 560/878, Loss: 0.033010\nEpoch 20, Batch 570/878, Loss: 0.029322\nEpoch 20, Batch 580/878, Loss: 0.033444\nEpoch 20, Batch 590/878, Loss: 0.026885\nEpoch 20, Batch 600/878, Loss: 0.030120\nEpoch 20, Batch 610/878, Loss: 0.036126\nEpoch 20, Batch 620/878, Loss: 0.027873\nEpoch 20, Batch 630/878, Loss: 0.039979\nEpoch 20, Batch 640/878, Loss: 0.030269\nEpoch 20, Batch 650/878, Loss: 0.025645\nEpoch 20, Batch 660/878, Loss: 0.031322\nEpoch 20, Batch 670/878, Loss: 0.031535\nEpoch 20, Batch 680/878, Loss: 0.040756\nEpoch 20, Batch 690/878, Loss: 0.029271\nEpoch 20, Batch 700/878, Loss: 0.032629\nEpoch 20, Batch 710/878, Loss: 0.034832\nEpoch 20, Batch 720/878, Loss: 0.035444\nEpoch 20, Batch 730/878, Loss: 0.030205\nEpoch 20, Batch 740/878, Loss: 0.033039\nEpoch 20, Batch 750/878, Loss: 0.028163\nEpoch 20, Batch 760/878, Loss: 0.032092\nEpoch 20, Batch 770/878, Loss: 0.031317\nEpoch 20, Batch 780/878, Loss: 0.030805\nEpoch 20, Batch 790/878, Loss: 0.033004\nEpoch 20, Batch 800/878, Loss: 0.031773\nEpoch 20, Batch 810/878, Loss: 0.035987\nEpoch 20, Batch 820/878, Loss: 0.033901\nEpoch 20, Batch 830/878, Loss: 0.037367\nEpoch 20, Batch 840/878, Loss: 0.036147\nEpoch 20, Batch 850/878, Loss: 0.028257\nEpoch 20, Batch 860/878, Loss: 0.034773\nEpoch 20, Batch 870/878, Loss: 0.035637\nEpoch 20/30, Average Loss: 0.033306, Time: 464s\nNew best model saved with loss: 0.033306\nEpoch 21, Batch 0/878, Loss: 0.041499\nEpoch 21, Batch 10/878, Loss: 0.032828\nEpoch 21, Batch 20/878, Loss: 0.029260\nEpoch 21, Batch 30/878, Loss: 0.035027\nEpoch 21, Batch 40/878, Loss: 0.036493\nEpoch 21, Batch 50/878, Loss: 0.034501\nEpoch 21, Batch 60/878, Loss: 0.026797\nEpoch 21, Batch 70/878, Loss: 0.028314\nEpoch 21, Batch 80/878, Loss: 0.037315\nEpoch 21, Batch 90/878, Loss: 0.032203\nEpoch 21, Batch 100/878, Loss: 0.035086\nEpoch 21, Batch 110/878, Loss: 0.030293\nEpoch 21, Batch 120/878, Loss: 0.044313\nEpoch 21, Batch 130/878, Loss: 0.034185\nEpoch 21, Batch 140/878, Loss: 0.034522\nEpoch 21, Batch 150/878, Loss: 0.027574\nEpoch 21, Batch 160/878, Loss: 0.033001\nEpoch 21, Batch 170/878, Loss: 0.038076\nEpoch 21, Batch 180/878, Loss: 0.030747\nEpoch 21, Batch 190/878, Loss: 0.024303\nEpoch 21, Batch 200/878, Loss: 0.044650\nEpoch 21, Batch 210/878, Loss: 0.033999\nEpoch 21, Batch 220/878, Loss: 0.029722\nEpoch 21, Batch 230/878, Loss: 0.028334\nEpoch 21, Batch 240/878, Loss: 0.036337\nEpoch 21, Batch 250/878, Loss: 0.040608\nEpoch 21, Batch 260/878, Loss: 0.041067\nEpoch 21, Batch 270/878, Loss: 0.029350\nEpoch 21, Batch 280/878, Loss: 0.035036\nEpoch 21, Batch 290/878, Loss: 0.027344\nEpoch 21, Batch 300/878, Loss: 0.028774\nEpoch 21, Batch 310/878, Loss: 0.033860\nEpoch 21, Batch 320/878, Loss: 0.040504\nEpoch 21, Batch 330/878, Loss: 0.027628\nEpoch 21, Batch 340/878, Loss: 0.034693\nEpoch 21, Batch 350/878, Loss: 0.036134\nEpoch 21, Batch 360/878, Loss: 0.036565\nEpoch 21, Batch 370/878, Loss: 0.035101\nEpoch 21, Batch 380/878, Loss: 0.031658\nEpoch 21, Batch 390/878, Loss: 0.032607\nEpoch 21, Batch 400/878, Loss: 0.023034\nEpoch 21, Batch 410/878, Loss: 0.030182\nEpoch 21, Batch 420/878, Loss: 0.041719\nEpoch 21, Batch 430/878, Loss: 0.028649\nEpoch 21, Batch 440/878, Loss: 0.032456\nEpoch 21, Batch 450/878, Loss: 0.026504\nEpoch 21, Batch 460/878, Loss: 0.029967\nEpoch 21, Batch 470/878, Loss: 0.026656\nEpoch 21, Batch 480/878, Loss: 0.034158\nEpoch 21, Batch 490/878, Loss: 0.036647\nEpoch 21, Batch 500/878, Loss: 0.036011\nEpoch 21, Batch 510/878, Loss: 0.030102\nEpoch 21, Batch 520/878, Loss: 0.034129\nEpoch 21, Batch 530/878, Loss: 0.036923\nEpoch 21, Batch 540/878, Loss: 0.025737\nEpoch 21, Batch 550/878, Loss: 0.044890\nEpoch 21, Batch 560/878, Loss: 0.044500\nEpoch 21, Batch 570/878, Loss: 0.034757\nEpoch 21, Batch 580/878, Loss: 0.029079\nEpoch 21, Batch 590/878, Loss: 0.030051\nEpoch 21, Batch 600/878, Loss: 0.036815\nEpoch 21, Batch 610/878, Loss: 0.031234\nEpoch 21, Batch 620/878, Loss: 0.031882\nEpoch 21, Batch 630/878, Loss: 0.031782\nEpoch 21, Batch 640/878, Loss: 0.032765\nEpoch 21, Batch 650/878, Loss: 0.027503\nEpoch 21, Batch 660/878, Loss: 0.031291\nEpoch 21, Batch 670/878, Loss: 0.032625\nEpoch 21, Batch 680/878, Loss: 0.032278\nEpoch 21, Batch 690/878, Loss: 0.034626\nEpoch 21, Batch 700/878, Loss: 0.045226\nEpoch 21, Batch 710/878, Loss: 0.039255\nEpoch 21, Batch 720/878, Loss: 0.032425\nEpoch 21, Batch 730/878, Loss: 0.023428\nEpoch 21, Batch 740/878, Loss: 0.031604\nEpoch 21, Batch 750/878, Loss: 0.029648\nEpoch 21, Batch 760/878, Loss: 0.030227\nEpoch 21, Batch 770/878, Loss: 0.036010\nEpoch 21, Batch 780/878, Loss: 0.029471\nEpoch 21, Batch 790/878, Loss: 0.035464\nEpoch 21, Batch 800/878, Loss: 0.031860\nEpoch 21, Batch 810/878, Loss: 0.035371\nEpoch 21, Batch 820/878, Loss: 0.032967\nEpoch 21, Batch 830/878, Loss: 0.037825\nEpoch 21, Batch 840/878, Loss: 0.031175\nEpoch 21, Batch 850/878, Loss: 0.032640\nEpoch 21, Batch 860/878, Loss: 0.039368\nEpoch 21, Batch 870/878, Loss: 0.032134\nEpoch 21/30, Average Loss: 0.033045, Time: 465s\nNew best model saved with loss: 0.033045\nEpoch 22, Batch 0/878, Loss: 0.032106\nEpoch 22, Batch 10/878, Loss: 0.033280\nEpoch 22, Batch 20/878, Loss: 0.030079\nEpoch 22, Batch 30/878, Loss: 0.032218\nEpoch 22, Batch 40/878, Loss: 0.034164\nEpoch 22, Batch 50/878, Loss: 0.041059\nEpoch 22, Batch 60/878, Loss: 0.032625\nEpoch 22, Batch 70/878, Loss: 0.034579\nEpoch 22, Batch 80/878, Loss: 0.042718\nEpoch 22, Batch 90/878, Loss: 0.031536\nEpoch 22, Batch 100/878, Loss: 0.023187\nEpoch 22, Batch 110/878, Loss: 0.027640\nEpoch 22, Batch 120/878, Loss: 0.033185\nEpoch 22, Batch 130/878, Loss: 0.029924\nEpoch 22, Batch 140/878, Loss: 0.030524\nEpoch 22, Batch 150/878, Loss: 0.034227\nEpoch 22, Batch 160/878, Loss: 0.035667\nEpoch 22, Batch 170/878, Loss: 0.032015\nEpoch 22, Batch 180/878, Loss: 0.038765\nEpoch 22, Batch 190/878, Loss: 0.026918\nEpoch 22, Batch 200/878, Loss: 0.027880\nEpoch 22, Batch 210/878, Loss: 0.038348\nEpoch 22, Batch 220/878, Loss: 0.038541\nEpoch 22, Batch 230/878, Loss: 0.029069\nEpoch 22, Batch 240/878, Loss: 0.036299\nEpoch 22, Batch 250/878, Loss: 0.033410\nEpoch 22, Batch 260/878, Loss: 0.028487\nEpoch 22, Batch 270/878, Loss: 0.033977\nEpoch 22, Batch 280/878, Loss: 0.034711\nEpoch 22, Batch 290/878, Loss: 0.032267\nEpoch 22, Batch 300/878, Loss: 0.033650\nEpoch 22, Batch 310/878, Loss: 0.028532\nEpoch 22, Batch 320/878, Loss: 0.033723\nEpoch 22, Batch 330/878, Loss: 0.031158\nEpoch 22, Batch 340/878, Loss: 0.034592\nEpoch 22, Batch 350/878, Loss: 0.042498\nEpoch 22, Batch 360/878, Loss: 0.033676\nEpoch 22, Batch 370/878, Loss: 0.031085\nEpoch 22, Batch 380/878, Loss: 0.032454\nEpoch 22, Batch 390/878, Loss: 0.036867\nEpoch 22, Batch 400/878, Loss: 0.024566\nEpoch 22, Batch 410/878, Loss: 0.027729\nEpoch 22, Batch 420/878, Loss: 0.025657\nEpoch 22, Batch 430/878, Loss: 0.032944\nEpoch 22, Batch 440/878, Loss: 0.033195\nEpoch 22, Batch 450/878, Loss: 0.041154\nEpoch 22, Batch 460/878, Loss: 0.042655\nEpoch 22, Batch 470/878, Loss: 0.036421\nEpoch 22, Batch 480/878, Loss: 0.028948\nEpoch 22, Batch 490/878, Loss: 0.032691\nEpoch 22, Batch 500/878, Loss: 0.032490\nEpoch 22, Batch 510/878, Loss: 0.032584\nEpoch 22, Batch 520/878, Loss: 0.029831\nEpoch 22, Batch 530/878, Loss: 0.033732\nEpoch 22, Batch 540/878, Loss: 0.036074\nEpoch 22, Batch 550/878, Loss: 0.029613\nEpoch 22, Batch 560/878, Loss: 0.029844\nEpoch 22, Batch 570/878, Loss: 0.037907\nEpoch 22, Batch 580/878, Loss: 0.026468\nEpoch 22, Batch 590/878, Loss: 0.032716\nEpoch 22, Batch 600/878, Loss: 0.033030\nEpoch 22, Batch 610/878, Loss: 0.035498\nEpoch 22, Batch 620/878, Loss: 0.030193\nEpoch 22, Batch 630/878, Loss: 0.032544\nEpoch 22, Batch 640/878, Loss: 0.035773\nEpoch 22, Batch 650/878, Loss: 0.027533\nEpoch 22, Batch 660/878, Loss: 0.040513\nEpoch 22, Batch 670/878, Loss: 0.037823\nEpoch 22, Batch 680/878, Loss: 0.027899\nEpoch 22, Batch 690/878, Loss: 0.026167\nEpoch 22, Batch 700/878, Loss: 0.027465\nEpoch 22, Batch 710/878, Loss: 0.037255\nEpoch 22, Batch 720/878, Loss: 0.029858\nEpoch 22, Batch 730/878, Loss: 0.031450\nEpoch 22, Batch 740/878, Loss: 0.031932\nEpoch 22, Batch 750/878, Loss: 0.049706\nEpoch 22, Batch 760/878, Loss: 0.029151\nEpoch 22, Batch 770/878, Loss: 0.024862\nEpoch 22, Batch 780/878, Loss: 0.030356\nEpoch 22, Batch 790/878, Loss: 0.030331\nEpoch 22, Batch 800/878, Loss: 0.030501\nEpoch 22, Batch 810/878, Loss: 0.042849\nEpoch 22, Batch 820/878, Loss: 0.039884\nEpoch 22, Batch 830/878, Loss: 0.029247\nEpoch 22, Batch 840/878, Loss: 0.036639\nEpoch 22, Batch 850/878, Loss: 0.031079\nEpoch 22, Batch 860/878, Loss: 0.035458\nEpoch 22, Batch 870/878, Loss: 0.032591\nEpoch 22/30, Average Loss: 0.032728, Time: 465s\nNew best model saved with loss: 0.032728\nEpoch 23, Batch 0/878, Loss: 0.030417\nEpoch 23, Batch 10/878, Loss: 0.026843\nEpoch 23, Batch 20/878, Loss: 0.030070\nEpoch 23, Batch 30/878, Loss: 0.031103\nEpoch 23, Batch 40/878, Loss: 0.033515\nEpoch 23, Batch 50/878, Loss: 0.041844\nEpoch 23, Batch 60/878, Loss: 0.029964\nEpoch 23, Batch 70/878, Loss: 0.032331\nEpoch 23, Batch 80/878, Loss: 0.030391\nEpoch 23, Batch 90/878, Loss: 0.028466\nEpoch 23, Batch 100/878, Loss: 0.037752\nEpoch 23, Batch 110/878, Loss: 0.030674\nEpoch 23, Batch 120/878, Loss: 0.027208\nEpoch 23, Batch 130/878, Loss: 0.028687\nEpoch 23, Batch 140/878, Loss: 0.031888\nEpoch 23, Batch 150/878, Loss: 0.034233\nEpoch 23, Batch 160/878, Loss: 0.036070\nEpoch 23, Batch 170/878, Loss: 0.044027\nEpoch 23, Batch 180/878, Loss: 0.033915\nEpoch 23, Batch 190/878, Loss: 0.034296\nEpoch 23, Batch 200/878, Loss: 0.033888\nEpoch 23, Batch 210/878, Loss: 0.039146\nEpoch 23, Batch 220/878, Loss: 0.028040\nEpoch 23, Batch 230/878, Loss: 0.030003\nEpoch 23, Batch 240/878, Loss: 0.037309\nEpoch 23, Batch 250/878, Loss: 0.032465\nEpoch 23, Batch 260/878, Loss: 0.037438\nEpoch 23, Batch 270/878, Loss: 0.036771\nEpoch 23, Batch 280/878, Loss: 0.032386\nEpoch 23, Batch 290/878, Loss: 0.028885\nEpoch 23, Batch 300/878, Loss: 0.033794\nEpoch 23, Batch 310/878, Loss: 0.027726\nEpoch 23, Batch 320/878, Loss: 0.027210\nEpoch 23, Batch 330/878, Loss: 0.031038\nEpoch 23, Batch 340/878, Loss: 0.038072\nEpoch 23, Batch 350/878, Loss: 0.031732\nEpoch 23, Batch 360/878, Loss: 0.032307\nEpoch 23, Batch 370/878, Loss: 0.028702\nEpoch 23, Batch 380/878, Loss: 0.034187\nEpoch 23, Batch 390/878, Loss: 0.044767\nEpoch 23, Batch 400/878, Loss: 0.024814\nEpoch 23, Batch 410/878, Loss: 0.031842\nEpoch 23, Batch 420/878, Loss: 0.037934\nEpoch 23, Batch 430/878, Loss: 0.034314\nEpoch 23, Batch 440/878, Loss: 0.031334\nEpoch 23, Batch 450/878, Loss: 0.025458\nEpoch 23, Batch 460/878, Loss: 0.030035\nEpoch 23, Batch 470/878, Loss: 0.034972\nEpoch 23, Batch 480/878, Loss: 0.025941\nEpoch 23, Batch 490/878, Loss: 0.033658\nEpoch 23, Batch 500/878, Loss: 0.034894\nEpoch 23, Batch 510/878, Loss: 0.028859\nEpoch 23, Batch 520/878, Loss: 0.029987\nEpoch 23, Batch 530/878, Loss: 0.027067\nEpoch 23, Batch 540/878, Loss: 0.035086\nEpoch 23, Batch 550/878, Loss: 0.042812\nEpoch 23, Batch 560/878, Loss: 0.036345\nEpoch 23, Batch 570/878, Loss: 0.041218\nEpoch 23, Batch 580/878, Loss: 0.029362\nEpoch 23, Batch 590/878, Loss: 0.026962\nEpoch 23, Batch 600/878, Loss: 0.044954\nEpoch 23, Batch 610/878, Loss: 0.036173\nEpoch 23, Batch 620/878, Loss: 0.039107\nEpoch 23, Batch 630/878, Loss: 0.035750\nEpoch 23, Batch 640/878, Loss: 0.043580\nEpoch 23, Batch 650/878, Loss: 0.026352\nEpoch 23, Batch 660/878, Loss: 0.032226\nEpoch 23, Batch 670/878, Loss: 0.026322\nEpoch 23, Batch 680/878, Loss: 0.030992\nEpoch 23, Batch 690/878, Loss: 0.026451\nEpoch 23, Batch 700/878, Loss: 0.025758\nEpoch 23, Batch 710/878, Loss: 0.037500\nEpoch 23, Batch 720/878, Loss: 0.031042\nEpoch 23, Batch 730/878, Loss: 0.027425\nEpoch 23, Batch 740/878, Loss: 0.026550\nEpoch 23, Batch 750/878, Loss: 0.035037\nEpoch 23, Batch 760/878, Loss: 0.030915\nEpoch 23, Batch 770/878, Loss: 0.030023\nEpoch 23, Batch 780/878, Loss: 0.036977\nEpoch 23, Batch 790/878, Loss: 0.026711\nEpoch 23, Batch 800/878, Loss: 0.036213\nEpoch 23, Batch 810/878, Loss: 0.025903\nEpoch 23, Batch 820/878, Loss: 0.031127\nEpoch 23, Batch 830/878, Loss: 0.037576\nEpoch 23, Batch 840/878, Loss: 0.030463\nEpoch 23, Batch 850/878, Loss: 0.029161\nEpoch 23, Batch 860/878, Loss: 0.035724\nEpoch 23, Batch 870/878, Loss: 0.037924\nEpoch 23/30, Average Loss: 0.032468, Time: 464s\nNew best model saved with loss: 0.032468\nEpoch 24, Batch 0/878, Loss: 0.039937\nEpoch 24, Batch 10/878, Loss: 0.029280\nEpoch 24, Batch 20/878, Loss: 0.031147\nEpoch 24, Batch 30/878, Loss: 0.023681\nEpoch 24, Batch 40/878, Loss: 0.022121\nEpoch 24, Batch 50/878, Loss: 0.042391\nEpoch 24, Batch 60/878, Loss: 0.029269\nEpoch 24, Batch 70/878, Loss: 0.029178\nEpoch 24, Batch 80/878, Loss: 0.036256\nEpoch 24, Batch 90/878, Loss: 0.026606\nEpoch 24, Batch 100/878, Loss: 0.025343\nEpoch 24, Batch 110/878, Loss: 0.030262\nEpoch 24, Batch 120/878, Loss: 0.027940\nEpoch 24, Batch 130/878, Loss: 0.031095\nEpoch 24, Batch 140/878, Loss: 0.040072\nEpoch 24, Batch 150/878, Loss: 0.038234\nEpoch 24, Batch 160/878, Loss: 0.026480\nEpoch 24, Batch 170/878, Loss: 0.043388\nEpoch 24, Batch 180/878, Loss: 0.027356\nEpoch 24, Batch 190/878, Loss: 0.035582\nEpoch 24, Batch 200/878, Loss: 0.028235\nEpoch 24, Batch 210/878, Loss: 0.028141\nEpoch 24, Batch 220/878, Loss: 0.027685\nEpoch 24, Batch 230/878, Loss: 0.028874\nEpoch 24, Batch 240/878, Loss: 0.024894\nEpoch 24, Batch 250/878, Loss: 0.035537\nEpoch 24, Batch 260/878, Loss: 0.036467\nEpoch 24, Batch 270/878, Loss: 0.033276\nEpoch 24, Batch 280/878, Loss: 0.022753\nEpoch 24, Batch 290/878, Loss: 0.027622\nEpoch 24, Batch 300/878, Loss: 0.033234\nEpoch 24, Batch 310/878, Loss: 0.029097\nEpoch 24, Batch 320/878, Loss: 0.028480\nEpoch 24, Batch 330/878, Loss: 0.030721\nEpoch 24, Batch 340/878, Loss: 0.033261\nEpoch 24, Batch 350/878, Loss: 0.038646\nEpoch 24, Batch 360/878, Loss: 0.032440\nEpoch 24, Batch 370/878, Loss: 0.035199\nEpoch 24, Batch 380/878, Loss: 0.027058\nEpoch 24, Batch 390/878, Loss: 0.030244\nEpoch 24, Batch 400/878, Loss: 0.034519\nEpoch 24, Batch 410/878, Loss: 0.035129\nEpoch 24, Batch 420/878, Loss: 0.035925\nEpoch 24, Batch 430/878, Loss: 0.036982\nEpoch 24, Batch 440/878, Loss: 0.031117\nEpoch 24, Batch 450/878, Loss: 0.043085\nEpoch 24, Batch 460/878, Loss: 0.029477\nEpoch 24, Batch 470/878, Loss: 0.030365\nEpoch 24, Batch 480/878, Loss: 0.023318\nEpoch 24, Batch 490/878, Loss: 0.026184\nEpoch 24, Batch 500/878, Loss: 0.032407\nEpoch 24, Batch 510/878, Loss: 0.023052\nEpoch 24, Batch 520/878, Loss: 0.024120\nEpoch 24, Batch 530/878, Loss: 0.036014\nEpoch 24, Batch 540/878, Loss: 0.036350\nEpoch 24, Batch 550/878, Loss: 0.025549\nEpoch 24, Batch 560/878, Loss: 0.031990\nEpoch 24, Batch 570/878, Loss: 0.034823\nEpoch 24, Batch 580/878, Loss: 0.035209\nEpoch 24, Batch 590/878, Loss: 0.031064\nEpoch 24, Batch 600/878, Loss: 0.037719\nEpoch 24, Batch 610/878, Loss: 0.035861\nEpoch 24, Batch 620/878, Loss: 0.035260\nEpoch 24, Batch 630/878, Loss: 0.033208\nEpoch 24, Batch 640/878, Loss: 0.030459\nEpoch 24, Batch 650/878, Loss: 0.033769\nEpoch 24, Batch 660/878, Loss: 0.028872\nEpoch 24, Batch 670/878, Loss: 0.031167\nEpoch 24, Batch 680/878, Loss: 0.032238\nEpoch 24, Batch 690/878, Loss: 0.026380\nEpoch 24, Batch 700/878, Loss: 0.033073\nEpoch 24, Batch 710/878, Loss: 0.030671\nEpoch 24, Batch 720/878, Loss: 0.030736\nEpoch 24, Batch 730/878, Loss: 0.034171\nEpoch 24, Batch 740/878, Loss: 0.029630\nEpoch 24, Batch 750/878, Loss: 0.043179\nEpoch 24, Batch 760/878, Loss: 0.031353\nEpoch 24, Batch 770/878, Loss: 0.033400\nEpoch 24, Batch 780/878, Loss: 0.041667\nEpoch 24, Batch 790/878, Loss: 0.032940\nEpoch 24, Batch 800/878, Loss: 0.032708\nEpoch 24, Batch 810/878, Loss: 0.037877\nEpoch 24, Batch 820/878, Loss: 0.035161\nEpoch 24, Batch 830/878, Loss: 0.034142\nEpoch 24, Batch 840/878, Loss: 0.033452\nEpoch 24, Batch 850/878, Loss: 0.031055\nEpoch 24, Batch 860/878, Loss: 0.025993\nEpoch 24, Batch 870/878, Loss: 0.028012\nEpoch 24/30, Average Loss: 0.032226, Time: 463s\nNew best model saved with loss: 0.032226\nEpoch 25, Batch 0/878, Loss: 0.033435\nEpoch 25, Batch 10/878, Loss: 0.028684\nEpoch 25, Batch 20/878, Loss: 0.035314\nEpoch 25, Batch 30/878, Loss: 0.036440\nEpoch 25, Batch 40/878, Loss: 0.031395\nEpoch 25, Batch 50/878, Loss: 0.031857\nEpoch 25, Batch 60/878, Loss: 0.034125\nEpoch 25, Batch 70/878, Loss: 0.042550\nEpoch 25, Batch 80/878, Loss: 0.033060\nEpoch 25, Batch 90/878, Loss: 0.030805\nEpoch 25, Batch 100/878, Loss: 0.041202\nEpoch 25, Batch 110/878, Loss: 0.036351\nEpoch 25, Batch 120/878, Loss: 0.031903\nEpoch 25, Batch 130/878, Loss: 0.020613\nEpoch 25, Batch 140/878, Loss: 0.030313\nEpoch 25, Batch 150/878, Loss: 0.030404\nEpoch 25, Batch 160/878, Loss: 0.029721\nEpoch 25, Batch 170/878, Loss: 0.033528\nEpoch 25, Batch 180/878, Loss: 0.024145\nEpoch 25, Batch 190/878, Loss: 0.022800\nEpoch 25, Batch 200/878, Loss: 0.027091\nEpoch 25, Batch 210/878, Loss: 0.033971\nEpoch 25, Batch 220/878, Loss: 0.037041\nEpoch 25, Batch 230/878, Loss: 0.027835\nEpoch 25, Batch 240/878, Loss: 0.033308\nEpoch 25, Batch 250/878, Loss: 0.035779\nEpoch 25, Batch 260/878, Loss: 0.034754\nEpoch 25, Batch 270/878, Loss: 0.031098\nEpoch 25, Batch 280/878, Loss: 0.030238\nEpoch 25, Batch 290/878, Loss: 0.028022\nEpoch 25, Batch 300/878, Loss: 0.035841\nEpoch 25, Batch 310/878, Loss: 0.023235\nEpoch 25, Batch 320/878, Loss: 0.035608\nEpoch 25, Batch 330/878, Loss: 0.032628\nEpoch 25, Batch 340/878, Loss: 0.027595\nEpoch 25, Batch 350/878, Loss: 0.031370\nEpoch 25, Batch 360/878, Loss: 0.032381\nEpoch 25, Batch 370/878, Loss: 0.028119\nEpoch 25, Batch 380/878, Loss: 0.033651\nEpoch 25, Batch 390/878, Loss: 0.036139\nEpoch 25, Batch 400/878, Loss: 0.035291\nEpoch 25, Batch 410/878, Loss: 0.032710\nEpoch 25, Batch 420/878, Loss: 0.034404\nEpoch 25, Batch 430/878, Loss: 0.031317\nEpoch 25, Batch 440/878, Loss: 0.031175\nEpoch 25, Batch 450/878, Loss: 0.034273\nEpoch 25, Batch 460/878, Loss: 0.020234\nEpoch 25, Batch 470/878, Loss: 0.025363\nEpoch 25, Batch 480/878, Loss: 0.031557\nEpoch 25, Batch 490/878, Loss: 0.034424\nEpoch 25, Batch 500/878, Loss: 0.029830\nEpoch 25, Batch 510/878, Loss: 0.031684\nEpoch 25, Batch 520/878, Loss: 0.039801\nEpoch 25, Batch 530/878, Loss: 0.042066\nEpoch 25, Batch 540/878, Loss: 0.022050\nEpoch 25, Batch 550/878, Loss: 0.034437\nEpoch 25, Batch 560/878, Loss: 0.032100\nEpoch 25, Batch 570/878, Loss: 0.031527\nEpoch 25, Batch 580/878, Loss: 0.036214\nEpoch 25, Batch 590/878, Loss: 0.037424\nEpoch 25, Batch 600/878, Loss: 0.028484\nEpoch 25, Batch 610/878, Loss: 0.034371\nEpoch 25, Batch 620/878, Loss: 0.030484\nEpoch 25, Batch 630/878, Loss: 0.031920\nEpoch 25, Batch 640/878, Loss: 0.030393\nEpoch 25, Batch 650/878, Loss: 0.030483\nEpoch 25, Batch 660/878, Loss: 0.029433\nEpoch 25, Batch 670/878, Loss: 0.042053\nEpoch 25, Batch 680/878, Loss: 0.026855\nEpoch 25, Batch 690/878, Loss: 0.031781\nEpoch 25, Batch 700/878, Loss: 0.036840\nEpoch 25, Batch 710/878, Loss: 0.033886\nEpoch 25, Batch 720/878, Loss: 0.025878\nEpoch 25, Batch 730/878, Loss: 0.030125\nEpoch 25, Batch 740/878, Loss: 0.028111\nEpoch 25, Batch 750/878, Loss: 0.033203\nEpoch 25, Batch 760/878, Loss: 0.032308\nEpoch 25, Batch 770/878, Loss: 0.037641\nEpoch 25, Batch 780/878, Loss: 0.028678\nEpoch 25, Batch 790/878, Loss: 0.029665\nEpoch 25, Batch 800/878, Loss: 0.034219\nEpoch 25, Batch 810/878, Loss: 0.038476\nEpoch 25, Batch 820/878, Loss: 0.024184\nEpoch 25, Batch 830/878, Loss: 0.027797\nEpoch 25, Batch 840/878, Loss: 0.030915\nEpoch 25, Batch 850/878, Loss: 0.038174\nEpoch 25, Batch 860/878, Loss: 0.032255\nEpoch 25, Batch 870/878, Loss: 0.028318\nEpoch 25/30, Average Loss: 0.031973, Time: 465s\nNew best model saved with loss: 0.031973\nEpoch 26, Batch 0/878, Loss: 0.028877\nEpoch 26, Batch 10/878, Loss: 0.033164\nEpoch 26, Batch 20/878, Loss: 0.037633\nEpoch 26, Batch 30/878, Loss: 0.037497\nEpoch 26, Batch 40/878, Loss: 0.033936\nEpoch 26, Batch 50/878, Loss: 0.032199\nEpoch 26, Batch 60/878, Loss: 0.035938\nEpoch 26, Batch 70/878, Loss: 0.030799\nEpoch 26, Batch 80/878, Loss: 0.024808\nEpoch 26, Batch 90/878, Loss: 0.023566\nEpoch 26, Batch 100/878, Loss: 0.031891\nEpoch 26, Batch 110/878, Loss: 0.031992\nEpoch 26, Batch 120/878, Loss: 0.032071\nEpoch 26, Batch 130/878, Loss: 0.027785\nEpoch 26, Batch 140/878, Loss: 0.025641\nEpoch 26, Batch 150/878, Loss: 0.033789\nEpoch 26, Batch 160/878, Loss: 0.040206\nEpoch 26, Batch 170/878, Loss: 0.042050\nEpoch 26, Batch 180/878, Loss: 0.033942\nEpoch 26, Batch 190/878, Loss: 0.026037\nEpoch 26, Batch 200/878, Loss: 0.028160\nEpoch 26, Batch 210/878, Loss: 0.036897\nEpoch 26, Batch 220/878, Loss: 0.031305\nEpoch 26, Batch 230/878, Loss: 0.031178\nEpoch 26, Batch 240/878, Loss: 0.031251\nEpoch 26, Batch 250/878, Loss: 0.038191\nEpoch 26, Batch 260/878, Loss: 0.028545\nEpoch 26, Batch 270/878, Loss: 0.039211\nEpoch 26, Batch 280/878, Loss: 0.038492\nEpoch 26, Batch 290/878, Loss: 0.034204\nEpoch 26, Batch 300/878, Loss: 0.040088\nEpoch 26, Batch 310/878, Loss: 0.044449\nEpoch 26, Batch 320/878, Loss: 0.038415\nEpoch 26, Batch 330/878, Loss: 0.029571\nEpoch 26, Batch 340/878, Loss: 0.035467\nEpoch 26, Batch 350/878, Loss: 0.033479\nEpoch 26, Batch 360/878, Loss: 0.025446\nEpoch 26, Batch 370/878, Loss: 0.032242\nEpoch 26, Batch 380/878, Loss: 0.034090\nEpoch 26, Batch 390/878, Loss: 0.032731\nEpoch 26, Batch 400/878, Loss: 0.025156\nEpoch 26, Batch 410/878, Loss: 0.037316\nEpoch 26, Batch 420/878, Loss: 0.032009\nEpoch 26, Batch 430/878, Loss: 0.037458\nEpoch 26, Batch 440/878, Loss: 0.034791\nEpoch 26, Batch 450/878, Loss: 0.030881\nEpoch 26, Batch 460/878, Loss: 0.028188\nEpoch 26, Batch 470/878, Loss: 0.031249\nEpoch 26, Batch 480/878, Loss: 0.032461\nEpoch 26, Batch 490/878, Loss: 0.029513\nEpoch 26, Batch 500/878, Loss: 0.029293\nEpoch 26, Batch 510/878, Loss: 0.038568\nEpoch 26, Batch 520/878, Loss: 0.031264\nEpoch 26, Batch 530/878, Loss: 0.032296\nEpoch 26, Batch 540/878, Loss: 0.034467\nEpoch 26, Batch 550/878, Loss: 0.032732\nEpoch 26, Batch 560/878, Loss: 0.036857\nEpoch 26, Batch 570/878, Loss: 0.031708\nEpoch 26, Batch 580/878, Loss: 0.034377\nEpoch 26, Batch 590/878, Loss: 0.032735\nEpoch 26, Batch 600/878, Loss: 0.030932\nEpoch 26, Batch 610/878, Loss: 0.030982\nEpoch 26, Batch 620/878, Loss: 0.032046\nEpoch 26, Batch 630/878, Loss: 0.032342\nEpoch 26, Batch 640/878, Loss: 0.028313\nEpoch 26, Batch 650/878, Loss: 0.027958\nEpoch 26, Batch 660/878, Loss: 0.028242\nEpoch 26, Batch 670/878, Loss: 0.036451\nEpoch 26, Batch 680/878, Loss: 0.029402\nEpoch 26, Batch 690/878, Loss: 0.029016\nEpoch 26, Batch 700/878, Loss: 0.031752\nEpoch 26, Batch 710/878, Loss: 0.028471\nEpoch 26, Batch 720/878, Loss: 0.027064\nEpoch 26, Batch 730/878, Loss: 0.028535\nEpoch 26, Batch 740/878, Loss: 0.032477\nEpoch 26, Batch 750/878, Loss: 0.032581\nEpoch 26, Batch 760/878, Loss: 0.041610\nEpoch 26, Batch 770/878, Loss: 0.028818\nEpoch 26, Batch 780/878, Loss: 0.036095\nEpoch 26, Batch 790/878, Loss: 0.029927\nEpoch 26, Batch 800/878, Loss: 0.034318\nEpoch 26, Batch 810/878, Loss: 0.034316\nEpoch 26, Batch 820/878, Loss: 0.023467\nEpoch 26, Batch 830/878, Loss: 0.036757\nEpoch 26, Batch 840/878, Loss: 0.030726\nEpoch 26, Batch 850/878, Loss: 0.027648\nEpoch 26, Batch 860/878, Loss: 0.028944\nEpoch 26, Batch 870/878, Loss: 0.028354\nEpoch 26/30, Average Loss: 0.031768, Time: 464s\nNew best model saved with loss: 0.031768\nEpoch 27, Batch 0/878, Loss: 0.029093\nEpoch 27, Batch 10/878, Loss: 0.027251\nEpoch 27, Batch 20/878, Loss: 0.040992\nEpoch 27, Batch 30/878, Loss: 0.030877\nEpoch 27, Batch 40/878, Loss: 0.041004\nEpoch 27, Batch 50/878, Loss: 0.030379\nEpoch 27, Batch 60/878, Loss: 0.039562\nEpoch 27, Batch 70/878, Loss: 0.025745\nEpoch 27, Batch 80/878, Loss: 0.027190\nEpoch 27, Batch 90/878, Loss: 0.025880\nEpoch 27, Batch 100/878, Loss: 0.031747\nEpoch 27, Batch 110/878, Loss: 0.030909\nEpoch 27, Batch 120/878, Loss: 0.031246\nEpoch 27, Batch 130/878, Loss: 0.034443\nEpoch 27, Batch 140/878, Loss: 0.023973\nEpoch 27, Batch 150/878, Loss: 0.031366\nEpoch 27, Batch 160/878, Loss: 0.031887\nEpoch 27, Batch 170/878, Loss: 0.024819\nEpoch 27, Batch 180/878, Loss: 0.038078\nEpoch 27, Batch 190/878, Loss: 0.026256\nEpoch 27, Batch 200/878, Loss: 0.029157\nEpoch 27, Batch 210/878, Loss: 0.029813\nEpoch 27, Batch 220/878, Loss: 0.031334\nEpoch 27, Batch 230/878, Loss: 0.027823\nEpoch 27, Batch 240/878, Loss: 0.034295\nEpoch 27, Batch 250/878, Loss: 0.023289\nEpoch 27, Batch 260/878, Loss: 0.031792\nEpoch 27, Batch 270/878, Loss: 0.032936\nEpoch 27, Batch 280/878, Loss: 0.026131\nEpoch 27, Batch 290/878, Loss: 0.028097\nEpoch 27, Batch 300/878, Loss: 0.033319\nEpoch 27, Batch 310/878, Loss: 0.035308\nEpoch 27, Batch 320/878, Loss: 0.027024\nEpoch 27, Batch 330/878, Loss: 0.030739\nEpoch 27, Batch 340/878, Loss: 0.028202\nEpoch 27, Batch 350/878, Loss: 0.030471\nEpoch 27, Batch 360/878, Loss: 0.029470\nEpoch 27, Batch 370/878, Loss: 0.028166\nEpoch 27, Batch 380/878, Loss: 0.025488\nEpoch 27, Batch 390/878, Loss: 0.027702\nEpoch 27, Batch 400/878, Loss: 0.031281\nEpoch 27, Batch 410/878, Loss: 0.036875\nEpoch 27, Batch 420/878, Loss: 0.025771\nEpoch 27, Batch 430/878, Loss: 0.025707\nEpoch 27, Batch 440/878, Loss: 0.028982\nEpoch 27, Batch 450/878, Loss: 0.025602\nEpoch 27, Batch 460/878, Loss: 0.034954\nEpoch 27, Batch 470/878, Loss: 0.022918\nEpoch 27, Batch 480/878, Loss: 0.024121\nEpoch 27, Batch 490/878, Loss: 0.031508\nEpoch 27, Batch 500/878, Loss: 0.032527\nEpoch 27, Batch 510/878, Loss: 0.044316\nEpoch 27, Batch 520/878, Loss: 0.027505\nEpoch 27, Batch 530/878, Loss: 0.026798\nEpoch 27, Batch 540/878, Loss: 0.024740\nEpoch 27, Batch 550/878, Loss: 0.027307\nEpoch 27, Batch 560/878, Loss: 0.026412\nEpoch 27, Batch 570/878, Loss: 0.026948\nEpoch 27, Batch 580/878, Loss: 0.028255\nEpoch 27, Batch 590/878, Loss: 0.032436\nEpoch 27, Batch 600/878, Loss: 0.029762\nEpoch 27, Batch 610/878, Loss: 0.034558\nEpoch 27, Batch 620/878, Loss: 0.028326\nEpoch 27, Batch 630/878, Loss: 0.034542\nEpoch 27, Batch 640/878, Loss: 0.032070\nEpoch 27, Batch 650/878, Loss: 0.035257\nEpoch 27, Batch 660/878, Loss: 0.033341\nEpoch 27, Batch 670/878, Loss: 0.034806\nEpoch 27, Batch 680/878, Loss: 0.044676\nEpoch 27, Batch 690/878, Loss: 0.030364\nEpoch 27, Batch 700/878, Loss: 0.034473\nEpoch 27, Batch 710/878, Loss: 0.032207\nEpoch 27, Batch 720/878, Loss: 0.031659\nEpoch 27, Batch 730/878, Loss: 0.034690\nEpoch 27, Batch 740/878, Loss: 0.035361\nEpoch 27, Batch 750/878, Loss: 0.033602\nEpoch 27, Batch 760/878, Loss: 0.035162\nEpoch 27, Batch 770/878, Loss: 0.028212\nEpoch 27, Batch 780/878, Loss: 0.038741\nEpoch 27, Batch 790/878, Loss: 0.030864\nEpoch 27, Batch 800/878, Loss: 0.023864\nEpoch 27, Batch 810/878, Loss: 0.028488\nEpoch 27, Batch 820/878, Loss: 0.026703\nEpoch 27, Batch 830/878, Loss: 0.027932\nEpoch 27, Batch 840/878, Loss: 0.036344\nEpoch 27, Batch 850/878, Loss: 0.028003\nEpoch 27, Batch 860/878, Loss: 0.034501\nEpoch 27, Batch 870/878, Loss: 0.031903\nEpoch 27/30, Average Loss: 0.031538, Time: 465s\nNew best model saved with loss: 0.031538\nEpoch 28, Batch 0/878, Loss: 0.033969\nEpoch 28, Batch 10/878, Loss: 0.032512\nEpoch 28, Batch 20/878, Loss: 0.027717\nEpoch 28, Batch 30/878, Loss: 0.039202\nEpoch 28, Batch 40/878, Loss: 0.031156\nEpoch 28, Batch 50/878, Loss: 0.026556\nEpoch 28, Batch 60/878, Loss: 0.037531\nEpoch 28, Batch 70/878, Loss: 0.028383\nEpoch 28, Batch 80/878, Loss: 0.034110\nEpoch 28, Batch 90/878, Loss: 0.035326\nEpoch 28, Batch 100/878, Loss: 0.035145\nEpoch 28, Batch 110/878, Loss: 0.038000\nEpoch 28, Batch 120/878, Loss: 0.031589\nEpoch 28, Batch 130/878, Loss: 0.029133\nEpoch 28, Batch 140/878, Loss: 0.027299\nEpoch 28, Batch 150/878, Loss: 0.035174\nEpoch 28, Batch 160/878, Loss: 0.032970\nEpoch 28, Batch 170/878, Loss: 0.029156\nEpoch 28, Batch 180/878, Loss: 0.032927\nEpoch 28, Batch 190/878, Loss: 0.024307\nEpoch 28, Batch 200/878, Loss: 0.031648\nEpoch 28, Batch 210/878, Loss: 0.021733\nEpoch 28, Batch 220/878, Loss: 0.025103\nEpoch 28, Batch 230/878, Loss: 0.036465\nEpoch 28, Batch 240/878, Loss: 0.029276\nEpoch 28, Batch 250/878, Loss: 0.028910\nEpoch 28, Batch 260/878, Loss: 0.023855\nEpoch 28, Batch 270/878, Loss: 0.033716\nEpoch 28, Batch 280/878, Loss: 0.032418\nEpoch 28, Batch 290/878, Loss: 0.033060\nEpoch 28, Batch 300/878, Loss: 0.032012\nEpoch 28, Batch 310/878, Loss: 0.030214\nEpoch 28, Batch 320/878, Loss: 0.031092\nEpoch 28, Batch 330/878, Loss: 0.024491\nEpoch 28, Batch 340/878, Loss: 0.034419\nEpoch 28, Batch 350/878, Loss: 0.033905\nEpoch 28, Batch 360/878, Loss: 0.035420\nEpoch 28, Batch 370/878, Loss: 0.034071\nEpoch 28, Batch 380/878, Loss: 0.031965\nEpoch 28, Batch 390/878, Loss: 0.026596\nEpoch 28, Batch 400/878, Loss: 0.025084\nEpoch 28, Batch 410/878, Loss: 0.046694\nEpoch 28, Batch 420/878, Loss: 0.022893\nEpoch 28, Batch 430/878, Loss: 0.035252\nEpoch 28, Batch 440/878, Loss: 0.027841\nEpoch 28, Batch 450/878, Loss: 0.043632\nEpoch 28, Batch 460/878, Loss: 0.031980\nEpoch 28, Batch 470/878, Loss: 0.027741\nEpoch 28, Batch 480/878, Loss: 0.034721\nEpoch 28, Batch 490/878, Loss: 0.027786\nEpoch 28, Batch 500/878, Loss: 0.031421\nEpoch 28, Batch 510/878, Loss: 0.033168\nEpoch 28, Batch 520/878, Loss: 0.029755\nEpoch 28, Batch 530/878, Loss: 0.027845\nEpoch 28, Batch 540/878, Loss: 0.029335\nEpoch 28, Batch 550/878, Loss: 0.020377\nEpoch 28, Batch 560/878, Loss: 0.032541\nEpoch 28, Batch 570/878, Loss: 0.038897\nEpoch 28, Batch 580/878, Loss: 0.033997\nEpoch 28, Batch 590/878, Loss: 0.029211\nEpoch 28, Batch 600/878, Loss: 0.027145\nEpoch 28, Batch 610/878, Loss: 0.032259\nEpoch 28, Batch 620/878, Loss: 0.035318\nEpoch 28, Batch 630/878, Loss: 0.038745\nEpoch 28, Batch 640/878, Loss: 0.039943\nEpoch 28, Batch 650/878, Loss: 0.021069\nEpoch 28, Batch 660/878, Loss: 0.028146\nEpoch 28, Batch 670/878, Loss: 0.034185\nEpoch 28, Batch 680/878, Loss: 0.031701\nEpoch 28, Batch 690/878, Loss: 0.024087\nEpoch 28, Batch 700/878, Loss: 0.033799\nEpoch 28, Batch 710/878, Loss: 0.024564\nEpoch 28, Batch 720/878, Loss: 0.025136\nEpoch 28, Batch 730/878, Loss: 0.024488\nEpoch 28, Batch 740/878, Loss: 0.029540\nEpoch 28, Batch 750/878, Loss: 0.025662\nEpoch 28, Batch 760/878, Loss: 0.035447\nEpoch 28, Batch 770/878, Loss: 0.031643\nEpoch 28, Batch 780/878, Loss: 0.027898\nEpoch 28, Batch 790/878, Loss: 0.035368\nEpoch 28, Batch 800/878, Loss: 0.028573\nEpoch 28, Batch 810/878, Loss: 0.020364\nEpoch 28, Batch 820/878, Loss: 0.028298\nEpoch 28, Batch 830/878, Loss: 0.029811\nEpoch 28, Batch 840/878, Loss: 0.028219\nEpoch 28, Batch 850/878, Loss: 0.029608\nEpoch 28, Batch 860/878, Loss: 0.031341\nEpoch 28, Batch 870/878, Loss: 0.033152\nEpoch 28/30, Average Loss: 0.031320, Time: 464s\nNew best model saved with loss: 0.031320\nEpoch 29, Batch 0/878, Loss: 0.027786\nEpoch 29, Batch 10/878, Loss: 0.035635\nEpoch 29, Batch 20/878, Loss: 0.029903\nEpoch 29, Batch 30/878, Loss: 0.031023\nEpoch 29, Batch 40/878, Loss: 0.028598\nEpoch 29, Batch 50/878, Loss: 0.033518\nEpoch 29, Batch 60/878, Loss: 0.045764\nEpoch 29, Batch 70/878, Loss: 0.034943\nEpoch 29, Batch 80/878, Loss: 0.045160\nEpoch 29, Batch 90/878, Loss: 0.030842\nEpoch 29, Batch 100/878, Loss: 0.034687\nEpoch 29, Batch 110/878, Loss: 0.028127\nEpoch 29, Batch 120/878, Loss: 0.037148\nEpoch 29, Batch 130/878, Loss: 0.029921\nEpoch 29, Batch 140/878, Loss: 0.025458\nEpoch 29, Batch 150/878, Loss: 0.032869\nEpoch 29, Batch 160/878, Loss: 0.028342\nEpoch 29, Batch 170/878, Loss: 0.026891\nEpoch 29, Batch 180/878, Loss: 0.031853\nEpoch 29, Batch 190/878, Loss: 0.039313\nEpoch 29, Batch 200/878, Loss: 0.029072\nEpoch 29, Batch 210/878, Loss: 0.030027\nEpoch 29, Batch 220/878, Loss: 0.026971\nEpoch 29, Batch 230/878, Loss: 0.028784\nEpoch 29, Batch 240/878, Loss: 0.034380\nEpoch 29, Batch 250/878, Loss: 0.036108\nEpoch 29, Batch 260/878, Loss: 0.026771\nEpoch 29, Batch 270/878, Loss: 0.039053\nEpoch 29, Batch 280/878, Loss: 0.028069\nEpoch 29, Batch 290/878, Loss: 0.033086\nEpoch 29, Batch 300/878, Loss: 0.026365\nEpoch 29, Batch 310/878, Loss: 0.030282\nEpoch 29, Batch 320/878, Loss: 0.025354\nEpoch 29, Batch 330/878, Loss: 0.036630\nEpoch 29, Batch 340/878, Loss: 0.030720\nEpoch 29, Batch 350/878, Loss: 0.035748\nEpoch 29, Batch 360/878, Loss: 0.027872\nEpoch 29, Batch 370/878, Loss: 0.037181\nEpoch 29, Batch 380/878, Loss: 0.026741\nEpoch 29, Batch 390/878, Loss: 0.020751\nEpoch 29, Batch 400/878, Loss: 0.033620\nEpoch 29, Batch 410/878, Loss: 0.033115\nEpoch 29, Batch 420/878, Loss: 0.032277\nEpoch 29, Batch 430/878, Loss: 0.035631\nEpoch 29, Batch 440/878, Loss: 0.035451\nEpoch 29, Batch 450/878, Loss: 0.029519\nEpoch 29, Batch 460/878, Loss: 0.029243\nEpoch 29, Batch 470/878, Loss: 0.033902\nEpoch 29, Batch 480/878, Loss: 0.034124\nEpoch 29, Batch 490/878, Loss: 0.029495\nEpoch 29, Batch 500/878, Loss: 0.039354\nEpoch 29, Batch 510/878, Loss: 0.035885\nEpoch 29, Batch 520/878, Loss: 0.026212\nEpoch 29, Batch 530/878, Loss: 0.029040\nEpoch 29, Batch 540/878, Loss: 0.027632\nEpoch 29, Batch 550/878, Loss: 0.025093\nEpoch 29, Batch 560/878, Loss: 0.037346\nEpoch 29, Batch 570/878, Loss: 0.029941\nEpoch 29, Batch 580/878, Loss: 0.030138\nEpoch 29, Batch 590/878, Loss: 0.029208\nEpoch 29, Batch 600/878, Loss: 0.026312\nEpoch 29, Batch 610/878, Loss: 0.032504\nEpoch 29, Batch 620/878, Loss: 0.034434\nEpoch 29, Batch 630/878, Loss: 0.040711\nEpoch 29, Batch 640/878, Loss: 0.030862\nEpoch 29, Batch 650/878, Loss: 0.036106\nEpoch 29, Batch 660/878, Loss: 0.032346\nEpoch 29, Batch 670/878, Loss: 0.032332\nEpoch 29, Batch 680/878, Loss: 0.036521\nEpoch 29, Batch 690/878, Loss: 0.029933\nEpoch 29, Batch 700/878, Loss: 0.032062\nEpoch 29, Batch 710/878, Loss: 0.021952\nEpoch 29, Batch 720/878, Loss: 0.033959\nEpoch 29, Batch 730/878, Loss: 0.027903\nEpoch 29, Batch 740/878, Loss: 0.031347\nEpoch 29, Batch 750/878, Loss: 0.038713\nEpoch 29, Batch 760/878, Loss: 0.027948\nEpoch 29, Batch 770/878, Loss: 0.029867\nEpoch 29, Batch 780/878, Loss: 0.029069\nEpoch 29, Batch 790/878, Loss: 0.034082\nEpoch 29, Batch 800/878, Loss: 0.024300\nEpoch 29, Batch 810/878, Loss: 0.032166\nEpoch 29, Batch 820/878, Loss: 0.029343\nEpoch 29, Batch 830/878, Loss: 0.029368\nEpoch 29, Batch 840/878, Loss: 0.033268\nEpoch 29, Batch 850/878, Loss: 0.030944\nEpoch 29, Batch 860/878, Loss: 0.032888\nEpoch 29, Batch 870/878, Loss: 0.048020\nEpoch 29/30, Average Loss: 0.031115, Time: 465s\nNew best model saved with loss: 0.031115\nEpoch 30, Batch 0/878, Loss: 0.023837\nEpoch 30, Batch 10/878, Loss: 0.033758\nEpoch 30, Batch 20/878, Loss: 0.032551\nEpoch 30, Batch 30/878, Loss: 0.037887\nEpoch 30, Batch 40/878, Loss: 0.026294\nEpoch 30, Batch 50/878, Loss: 0.026941\nEpoch 30, Batch 60/878, Loss: 0.031926\nEpoch 30, Batch 70/878, Loss: 0.027886\nEpoch 30, Batch 80/878, Loss: 0.026799\nEpoch 30, Batch 90/878, Loss: 0.033590\nEpoch 30, Batch 100/878, Loss: 0.029946\nEpoch 30, Batch 110/878, Loss: 0.027482\nEpoch 30, Batch 120/878, Loss: 0.029992\nEpoch 30, Batch 130/878, Loss: 0.032270\nEpoch 30, Batch 140/878, Loss: 0.037181\nEpoch 30, Batch 150/878, Loss: 0.024972\nEpoch 30, Batch 160/878, Loss: 0.037673\nEpoch 30, Batch 170/878, Loss: 0.038929\nEpoch 30, Batch 180/878, Loss: 0.043325\nEpoch 30, Batch 190/878, Loss: 0.026555\nEpoch 30, Batch 200/878, Loss: 0.019699\nEpoch 30, Batch 210/878, Loss: 0.021804\nEpoch 30, Batch 220/878, Loss: 0.029027\nEpoch 30, Batch 230/878, Loss: 0.027665\nEpoch 30, Batch 240/878, Loss: 0.034627\nEpoch 30, Batch 250/878, Loss: 0.035469\nEpoch 30, Batch 260/878, Loss: 0.033006\nEpoch 30, Batch 270/878, Loss: 0.028443\nEpoch 30, Batch 280/878, Loss: 0.031101\nEpoch 30, Batch 290/878, Loss: 0.043280\nEpoch 30, Batch 300/878, Loss: 0.028003\nEpoch 30, Batch 310/878, Loss: 0.026039\nEpoch 30, Batch 320/878, Loss: 0.032858\nEpoch 30, Batch 330/878, Loss: 0.031298\nEpoch 30, Batch 340/878, Loss: 0.026959\nEpoch 30, Batch 350/878, Loss: 0.037464\nEpoch 30, Batch 360/878, Loss: 0.032830\nEpoch 30, Batch 370/878, Loss: 0.029994\nEpoch 30, Batch 380/878, Loss: 0.028800\nEpoch 30, Batch 390/878, Loss: 0.026178\nEpoch 30, Batch 400/878, Loss: 0.026580\nEpoch 30, Batch 410/878, Loss: 0.027908\nEpoch 30, Batch 420/878, Loss: 0.037740\nEpoch 30, Batch 430/878, Loss: 0.034256\nEpoch 30, Batch 440/878, Loss: 0.033567\nEpoch 30, Batch 450/878, Loss: 0.031777\nEpoch 30, Batch 460/878, Loss: 0.028146\nEpoch 30, Batch 470/878, Loss: 0.022100\nEpoch 30, Batch 480/878, Loss: 0.026899\nEpoch 30, Batch 490/878, Loss: 0.030876\nEpoch 30, Batch 500/878, Loss: 0.031162\nEpoch 30, Batch 510/878, Loss: 0.033978\nEpoch 30, Batch 520/878, Loss: 0.030363\nEpoch 30, Batch 530/878, Loss: 0.033963\nEpoch 30, Batch 540/878, Loss: 0.028487\nEpoch 30, Batch 550/878, Loss: 0.031501\nEpoch 30, Batch 560/878, Loss: 0.023395\nEpoch 30, Batch 570/878, Loss: 0.030151\nEpoch 30, Batch 580/878, Loss: 0.028859\nEpoch 30, Batch 590/878, Loss: 0.037654\nEpoch 30, Batch 600/878, Loss: 0.030806\nEpoch 30, Batch 610/878, Loss: 0.030729\nEpoch 30, Batch 620/878, Loss: 0.024742\nEpoch 30, Batch 630/878, Loss: 0.028228\nEpoch 30, Batch 640/878, Loss: 0.035593\nEpoch 30, Batch 650/878, Loss: 0.024593\nEpoch 30, Batch 660/878, Loss: 0.029034\nEpoch 30, Batch 670/878, Loss: 0.027752\nEpoch 30, Batch 680/878, Loss: 0.031421\nEpoch 30, Batch 690/878, Loss: 0.030396\nEpoch 30, Batch 700/878, Loss: 0.029202\nEpoch 30, Batch 710/878, Loss: 0.024896\nEpoch 30, Batch 720/878, Loss: 0.031603\nEpoch 30, Batch 730/878, Loss: 0.029428\nEpoch 30, Batch 740/878, Loss: 0.033526\nEpoch 30, Batch 750/878, Loss: 0.028889\nEpoch 30, Batch 760/878, Loss: 0.045380\nEpoch 30, Batch 770/878, Loss: 0.033893\nEpoch 30, Batch 780/878, Loss: 0.031295\nEpoch 30, Batch 790/878, Loss: 0.043922\nEpoch 30, Batch 800/878, Loss: 0.029263\nEpoch 30, Batch 810/878, Loss: 0.028156\nEpoch 30, Batch 820/878, Loss: 0.031449\nEpoch 30, Batch 830/878, Loss: 0.031301\nEpoch 30, Batch 840/878, Loss: 0.033400\nEpoch 30, Batch 850/878, Loss: 0.026422\nEpoch 30, Batch 860/878, Loss: 0.032184\nEpoch 30, Batch 870/878, Loss: 0.035678\nEpoch 30/30, Average Loss: 0.030917, Time: 464s\nNew best model saved with loss: 0.030917\n","output_type":"stream"}],"execution_count":15},{"cell_type":"code","source":"test_model = load_model_for_inference(\"/kaggle/working/Hyper_U_NET_pytorch.pth\", device)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-08-25T13:47:31.919836Z","iopub.execute_input":"2025-08-25T13:47:31.920582Z","iopub.status.idle":"2025-08-25T13:47:32.399493Z","shell.execute_reply.started":"2025-08-25T13:47:31.920555Z","shell.execute_reply":"2025-08-25T13:47:32.398926Z"}},"outputs":[],"execution_count":19},{"cell_type":"code","source":"def prepare_test_image(img_path, dim=150):\n    img = cv2.imread(img_path)\n    img = cv2.resize(img, (dim, dim))\n\n    sz0, sz1 = img.shape[:2]\n    # R1 = img[:, :, 0].reshape(-1, 1)\n    # G1 = img[:, :, 1].reshape(-1, 1)\n    # B1 = img[:, :, 2].reshape(-1, 1)\n\n    R1 = img[:, :, 2].reshape(-1, 1)\n    G1 = img[:, :, 1].reshape(-1, 1)\n    B1 = img[:, :, 0].reshape(-1, 1)\n\n    # LAB2'ye çevir\n    L, A, B = RGB2LAB2(R1, G1, B1)\n    L = L.reshape(sz0, sz1, 1)\n\n    # Tensor formatına çevir\n    L_tensor = torch.FloatTensor(L).permute(2, 0, 1)  # (1, H, W)\n\n    return L_tensor, A.reshape(sz0, sz1), B.reshape(sz0, sz1)\n","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-08-25T13:47:38.528572Z","iopub.execute_input":"2025-08-25T13:47:38.529128Z","iopub.status.idle":"2025-08-25T13:47:38.534190Z","shell.execute_reply.started":"2025-08-25T13:47:38.529104Z","shell.execute_reply":"2025-08-25T13:47:38.533395Z"}},"outputs":[],"execution_count":20},{"cell_type":"code","source":"# Test resmi hazırla\nl_tensor, A_true, B_true = prepare_test_image(\"/kaggle/working/intel-image-classification/seg_pred/seg_pred/25.jpg\", dim=150)\n\n# Modelden tahmin\nab_pred = inference(test_model, l_tensor)  # (1,2,H,W) → numpy\nab_pred = ab_pred.squeeze(0)          # (2,H,W)\nA_pred, B_pred = ab_pred[0], ab_pred[1]\n\n# LAB2 → RGB geri çevir\nsz0, sz1 = A_pred.shape\nL = l_tensor.squeeze().numpy().reshape(-1,1)\nA = A_pred.reshape(-1,1)\nB = B_pred.reshape(-1,1)\n\nR, G, B = LAB22RGB(L, A, B)\nR = R.reshape(sz0, sz1)\nG = G.reshape(sz0, sz1)\nB = B.reshape(sz0, sz1)\n\nrgb_pred = cv2.merge([B,G,R])\n","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-08-25T13:51:52.586371Z","iopub.execute_input":"2025-08-25T13:51:52.587039Z","iopub.status.idle":"2025-08-25T13:51:52.623638Z","shell.execute_reply.started":"2025-08-25T13:51:52.587016Z","shell.execute_reply":"2025-08-25T13:51:52.623105Z"}},"outputs":[],"execution_count":33},{"cell_type":"code","source":"import matplotlib.pyplot as plt\n\nplt.subplot(1,2,1)\nplt.title(\"Grayscale (input)\")\nplt.imshow(l_tensor.squeeze(0).numpy(), cmap=\"gray\")\n\nplt.subplot(1,2,2)\nplt.title(\"Colorized (prediction)\")\nplt.imshow(cv2.cvtColor(rgb_pred, cv2.COLOR_BGR2RGB))\n\nplt.show()\n","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-08-25T13:51:52.795158Z","iopub.execute_input":"2025-08-25T13:51:52.795354Z","iopub.status.idle":"2025-08-25T13:51:53.064106Z","shell.execute_reply.started":"2025-08-25T13:51:52.795337Z","shell.execute_reply":"2025-08-25T13:51:53.063380Z"}},"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 640x480 with 2 Axes>","image/png":"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\n"},"metadata":{}}],"execution_count":34},{"cell_type":"code","source":"","metadata":{"trusted":true},"outputs":[],"execution_count":null}]}