diff --git "a/demo_sample.ipynb" "b/demo_sample.ipynb" deleted file mode 100644--- "a/demo_sample.ipynb" +++ /dev/null @@ -1,1961 +0,0 @@ -{ - "cells": [ - { - "cell_type": "markdown", - "metadata": { - "id": "X4cRE8IbIrIV" - }, - "source": [ - "If you're opening this Notebook on colab, you will probably need to install TorchVision and ImageNet-1K aatasets. Uncomment the following cell and run it." - ] - }, - { - "cell_type": "code", - "execution_count": 5, - "metadata": { - "id": "46CgrVgjg3E-" - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "--2025-05-22 07:47:03-- https://raw.githubusercontent.com/pytorch/vision/main/references/classification/presets.py\n", - "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 185.199.110.133, 185.199.111.133, 185.199.109.133, ...\n", - "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|185.199.110.133|:443... connected.\n", - "HTTP request sent, awaiting response... 200 OK\n", - "Length: 3885 (3.8K) [text/plain]\n", - "Saving to: ‘presets.py.1’\n", - "\n", - "presets.py.1 100%[===================>] 3.79K --.-KB/s in 0s \n", - "\n", - "2025-05-22 07:47:03 (24.3 MB/s) - ‘presets.py.1’ saved [3885/3885]\n", - "\n", - "--2025-05-22 07:47:04-- https://raw.githubusercontent.com/pytorch/vision/main/references/classification/sampler.py\n", - "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 185.199.109.133, 185.199.111.133, 185.199.110.133, ...\n", - "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|185.199.109.133|:443... connected.\n", - "HTTP request sent, awaiting response... 200 OK\n", - "Length: 2395 (2.3K) [text/plain]\n", - "Saving to: ‘sampler.py.1’\n", - "\n", - "sampler.py.1 100%[===================>] 2.34K --.-KB/s in 0s \n", - "\n", - "2025-05-22 07:47:04 (12.1 MB/s) - ‘sampler.py.1’ saved [2395/2395]\n", - "\n", - "--2025-05-22 07:47:04-- https://raw.githubusercontent.com/pytorch/vision/main/references/classification/train.py\n", - "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 185.199.110.133, 185.199.108.133, 185.199.111.133, ...\n", - "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|185.199.110.133|:443... connected.\n", - "HTTP request sent, awaiting response... 200 OK\n", - "Length: 23324 (23K) [text/plain]\n", - "Saving to: ‘train.py.1’\n", - "\n", - "train.py.1 100%[===================>] 22.78K --.-KB/s in 0.007s \n", - "\n", - "2025-05-22 07:47:04 (3.30 MB/s) - ‘train.py.1’ saved [23324/23324]\n", - "\n", - "--2025-05-22 07:47:04-- https://raw.githubusercontent.com/pytorch/vision/main/references/classification/train_quantization.py\n", - "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 185.199.108.133, 185.199.111.133, 185.199.109.133, ...\n", - "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|185.199.108.133|:443... connected.\n", - "HTTP request sent, awaiting response... 200 OK\n", - "Length: 11647 (11K) [text/plain]\n", - "Saving to: ‘train_quantization.py.1’\n", - "\n", - "train_quantization. 100%[===================>] 11.37K --.-KB/s in 0.002s \n", - "\n", - "2025-05-22 07:47:04 (6.26 MB/s) - ‘train_quantization.py.1’ saved [11647/11647]\n", - "\n", - "--2025-05-22 07:47:04-- https://raw.githubusercontent.com/pytorch/vision/main/references/classification/transformers.py\n", - "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 185.199.110.133, 185.199.111.133, 185.199.108.133, ...\n", - "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|185.199.110.133|:443... connected.\n", - "HTTP request sent, awaiting response... 404 Not Found\n", - "2025-05-22 07:47:05 ERROR 404: Not Found.\n", - "\n", - "--2025-05-22 07:47:05-- https://raw.githubusercontent.com/pytorch/vision/main/references/classification/utils.py\n", - "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 185.199.109.133, 185.199.108.133, 185.199.110.133, ...\n", - "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|185.199.109.133|:443... connected.\n", - "HTTP request sent, awaiting response... 200 OK\n", - "Length: 15791 (15K) [text/plain]\n", - "Saving to: ‘utils.py.1’\n", - "\n", - "utils.py.1 100%[===================>] 15.42K --.-KB/s in 0.005s \n", - "\n", - "2025-05-22 07:47:05 (3.21 MB/s) - ‘utils.py.1’ saved [15791/15791]\n", - "\n" - ] - } - ], - "source": [ - "! wget https://raw.githubusercontent.com/pytorch/vision/main/references/classification/presets.py\n", - "! wget https://raw.githubusercontent.com/pytorch/vision/main/references/classification/sampler.py\n", - "! wget https://raw.githubusercontent.com/pytorch/vision/main/references/classification/train.py\n", - "! wget https://raw.githubusercontent.com/pytorch/vision/main/references/classification/train_quantization.py\n", - "! wget https://raw.githubusercontent.com/pytorch/vision/main/references/classification/transformers.py\n", - "! wget https://raw.githubusercontent.com/pytorch/vision/main/references/classification/utils.py" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "HFASsisvIrIb" - }, - "source": [ - "You can find a script version of this notebook to fine-tune your model in a distributed fashion using multiple GPUs or TPUs [here](https://github.com/huggingface/transformers/tree/master/examples/language-modeling)." - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "KG5GeAWpg3E_" - }, - "source": [ - "We also quickly upload some telemetry - this tells us which examples and software versions are getting used so we know where to prioritize our maintenance efforts. We don't collect (or care about) any personally identifiable information, but if you'd prefer not to be counted, feel free to skip this step or delete this cell entirely." - ] - }, - { - "cell_type": "code", - "execution_count": 6, - "metadata": { - "id": "EaBokYCpg3FA" - }, - "outputs": [], - "source": [ - "import types\n", - "from typing import List, Callable\n", - "\n", - "import torch \n", - "from torch import nn, Tensor\n", - "from torch.nn import functional as F\n", - "from torchvision.models.resnet import BasicBlock\n", - "\n", - "\n", - "def trp_criterion(trp_blocks: nn.ModuleList, shared_head: Callable, criterion: Callable, lambdas: List[float], hidden_state: Tensor, logits: Tensor, targets: Tensor, loss_normalization=False):\n", - " losses, rewards = criterion(logits, targets)\n", - " returns = torch.ones_like(rewards, dtype=torch.float32, device=rewards.device)\n", - " if loss_normalization:\n", - " coeff = torch.mean(losses).detach()\n", - "\n", - " embeds = [hidden_state]\n", - " predictions = []\n", - " for k, w in enumerate(lambdas):\n", - " embeds.append(trp_blocks[k](embeds[-1]))\n", - " predictions.append(shared_head(embeds[-1]))\n", - " returns = returns + w * rewards\n", - " replica_losses, rewards = criterion(predictions[-1], targets, rewards)\n", - " losses = losses + replica_losses\n", - " loss = torch.mean(losses * returns)\n", - " \n", - " if loss_normalization:\n", - " with torch.no_grad():\n", - " coeff = torch.exp(coeff) / torch.exp(loss.detach())\n", - " loss = coeff * loss\n", - " \n", - " with torch.no_grad():\n", - " non_zero_returns = returns[returns != 0]\n", - " mean_returns = non_zero_returns.mean()\n", - " std_returns = non_zero_returns.std()\n", - "\n", - " return loss, mean_returns, std_returns\n", - "\n", - "\n", - "class TPBlock(nn.Module):\n", - " def __init__(self, depths: int, inplanes: int, planes: int):\n", - " super(TPBlock, self).__init__()\n", - "\n", - " blocks = [BasicBlock(inplanes=inplanes, planes=planes) for _ in range(depths)]\n", - " self.blocks = nn.Sequential(*blocks)\n", - " for name, param in self.blocks.named_parameters():\n", - " if 'conv' in name:\n", - " nn.init.zeros_(param) # Initialize weights\n", - " elif 'downsample' in name:\n", - " nn.init.zeros_(param) # Initialize biases\n", - "\n", - " def forward(self, x):\n", - " return self.blocks(x)\n", - "\n", - "\n", - "class ResNetConfig:\n", - " @staticmethod\n", - " def gen_criterion(label_smoothing=0.0, top_k=1):\n", - " def func(input, target, mask=None):\n", - " \"\"\"\n", - " Args:\n", - " input (Tensor): Input tensor of shape [B, C].\n", - " target (Tensor): Target labels of shape [B] or [B, C].\n", - "\n", - " Returns:\n", - " loss (Tensor): Scalar tensor representing the loss.\n", - " mask (Tensor): Boolean mask tensor of shape [B].\n", - " \"\"\"\n", - " label = torch.argmax(target, dim=1) if label_smoothing > 0.0 else target\n", - " \n", - " unmasked_loss = F.cross_entropy(input, label, reduction=\"none\", label_smoothing=label_smoothing)\n", - " if mask is None:\n", - " mask = torch.ones_like(unmasked_loss, dtype=torch.float32, device=target.device)\n", - " losses = mask * unmasked_loss\n", - "\n", - " with torch.no_grad():\n", - " topk_values, topk_indices = torch.topk(input, top_k, dim=-1)\n", - " mask = mask * torch.eq(topk_indices, label[:, None]).any(dim=-1).to(input.dtype)\n", - "\n", - " return losses, mask\n", - " return func\n", - " \n", - " @staticmethod\n", - " def gen_shared_head(self):\n", - " def func(x):\n", - " \"\"\"\n", - " Args:\n", - " x (Tensor): Hidden State tensor of shape [B, C, H, W].\n", - "\n", - " Returns:\n", - " logits (Tensor): Logits tensor of shape [B, C].\n", - " \"\"\"\n", - " x = self.layer4(x)\n", - " x = self.avgpool(x)\n", - " x = torch.flatten(x, 1)\n", - " logits = self.fc(x)\n", - " return logits\n", - " return func\n", - "\n", - " @staticmethod\n", - " def gen_forward(lambdas, loss_normalization=True, label_smoothing=0.0, top_k=1):\n", - " def func(self, x: Tensor, targets=None) -> Tensor:\n", - " x = self.conv1(x)\n", - " x = self.bn1(x)\n", - " x = self.relu(x)\n", - " x = self.maxpool(x)\n", - "\n", - " x = self.layer1(x)\n", - " x = self.layer2(x)\n", - " hidden_state = self.layer3(x)\n", - " x = self.layer4(hidden_state)\n", - " x = self.avgpool(x)\n", - " x = torch.flatten(x, 1)\n", - " logits = self.fc(x)\n", - "\n", - " if self.training:\n", - " shared_head = ResNetConfig.gen_shared_head(self)\n", - " criterion = ResNetConfig.gen_criterion(label_smoothing=label_smoothing, top_k=top_k)\n", - "\n", - " loss, mean_returns, std_returns = trp_criterion(self.trp_blocks, shared_head, criterion, lambdas, hidden_state, logits, targets, loss_normalization=loss_normalization)\n", - " \n", - " return logits, loss, mean_returns, std_returns\n", - "\n", - " return logits\n", - " \n", - " return func\n", - " \n", - "\n", - "def apply_trp(model, depths: List[int], planes: int, lambdas: List[float], **kwargs):\n", - " print(\"✅ Applying TRP to ResNet for Image Classification...\")\n", - " model.trp_blocks = torch.nn.ModuleList([TPBlock(depths=d, inplanes=planes, planes=planes) for d in depths])\n", - " model.forward = types.MethodType(ResNetConfig.gen_forward(lambdas), model) \n", - " return model" - ] - }, - { - "cell_type": "code", - "execution_count": 7, - "metadata": {}, - "outputs": [], - "source": [ - "import datetime\n", - "import os\n", - "import time\n", - "import warnings\n", - "\n", - "import presets\n", - "import torch\n", - "import torch.utils.data\n", - "import torchvision\n", - "import utils\n", - "from torch import nn\n", - "from torchvision.transforms.functional import InterpolationMode\n", - "\n", - "\n", - "def load_data(traindir, valdir):\n", - " # Data loading code\n", - " print(\"Loading data\")\n", - " interpolation = InterpolationMode(\"bilinear\")\n", - "\n", - " print(\"Loading training data\")\n", - " st = time.time()\n", - " dataset = torchvision.datasets.ImageFolder(\n", - " traindir,\n", - " presets.ClassificationPresetTrain(crop_size=224, interpolation=interpolation, auto_augment_policy=None, random_erase_prob=0.0, ra_magnitude=9, augmix_severity=3),\n", - " )\n", - " print(\"Took\", time.time() - st)\n", - "\n", - " print(\"Loading validation data\")\n", - " dataset_test = torchvision.datasets.ImageFolder(\n", - " valdir, \n", - " presets.ClassificationPresetEval(crop_size=224, resize_size=256, interpolation=interpolation)\n", - " )\n", - "\n", - " print(\"Creating data loaders\")\n", - " train_sampler = torch.utils.data.RandomSampler(dataset)\n", - " test_sampler = torch.utils.data.SequentialSampler(dataset_test)\n", - "\n", - " return dataset, dataset_test, train_sampler, test_sampler\n", - "\n", - "\n", - " \n", - "def train_one_epoch(model, optimizer, data_loader, device, epoch, args):\n", - " model.train()\n", - " metric_logger = utils.MetricLogger(delimiter=\" \")\n", - " metric_logger.add_meter(\"lr\", utils.SmoothedValue(window_size=1, fmt=\"{value}\"))\n", - " metric_logger.add_meter(\"img/s\", utils.SmoothedValue(window_size=10, fmt=\"{value}\"))\n", - "\n", - " header = f\"Epoch: [{epoch}]\"\n", - " for i, (image, target) in enumerate(metric_logger.log_every(data_loader, args.print_freq, header)):\n", - " start_time = time.time()\n", - " image, target = image.to(device), target.to(device)\n", - " with torch.amp.autocast(\"cuda\", enabled=False):\n", - " output, loss, mean_returns, std_returns = model(image, target)\n", - "\n", - " optimizer.zero_grad()\n", - " loss.backward()\n", - " optimizer.step()\n", - "\n", - " acc1, acc5 = utils.accuracy(output, target, topk=(1, 5))\n", - " batch_size = image.shape[0]\n", - " metric_logger.update(loss=loss.item(), lr=optimizer.param_groups[0][\"lr\"])\n", - " metric_logger.meters[\"acc1\"].update(acc1.item(), n=batch_size)\n", - " metric_logger.meters[\"acc5\"].update(acc5.item(), n=batch_size)\n", - " metric_logger.meters[\"meanQV\"].update(mean_returns.item(), n=batch_size)\n", - " metric_logger.meters[\"stdQV\"].update(std_returns.item(), n=batch_size)\n", - " metric_logger.meters[\"img/s\"].update(batch_size / (time.time() - start_time))\n", - "\n", - "\n", - "def evaluate(model, criterion, data_loader, device, print_freq=100, log_suffix=\"\"):\n", - " model.eval()\n", - " metric_logger = utils.MetricLogger(delimiter=\" \")\n", - " header = f\"Test: {log_suffix}\"\n", - "\n", - " num_processed_samples = 0\n", - " with torch.inference_mode():\n", - " for image, target in metric_logger.log_every(data_loader, print_freq, header):\n", - " image = image.to(device, non_blocking=True)\n", - " target = target.to(device, non_blocking=True)\n", - " output = model(image)\n", - " loss = criterion(output, target)\n", - "\n", - " acc1, acc5 = utils.accuracy(output, target, topk=(1, 5))\n", - " # FIXME need to take into account that the datasets\n", - " # could have been padded in distributed setup\n", - " batch_size = image.shape[0]\n", - " metric_logger.update(loss=loss.item())\n", - " metric_logger.meters[\"acc1\"].update(acc1.item(), n=batch_size)\n", - " metric_logger.meters[\"acc5\"].update(acc5.item(), n=batch_size)\n", - " num_processed_samples += batch_size\n", - " # gather the stats from all processes\n", - "\n", - " num_processed_samples = utils.reduce_across_processes(num_processed_samples)\n", - " if (\n", - " hasattr(data_loader.dataset, \"__len__\")\n", - " and len(data_loader.dataset) != num_processed_samples\n", - " and torch.distributed.get_rank() == 0\n", - " ):\n", - " # See FIXME above\n", - " warnings.warn(\n", - " f\"It looks like the dataset has {len(data_loader.dataset)} samples, but {num_processed_samples} \"\n", - " \"samples were used for the validation, which might bias the results. \"\n", - " \"Try adjusting the batch size and / or the world size. \"\n", - " \"Setting the world size to 1 is always a safe bet.\"\n", - " )\n", - "\n", - " metric_logger.synchronize_between_processes()\n", - "\n", - " print(f\"{header} Acc@1 {metric_logger.acc1.global_avg:.3f} Acc@5 {metric_logger.acc5.global_avg:.3f}\")\n", - " return metric_logger.acc1.global_avg\n", - "\n", - "\n", - "def main(args):\n", - " if args.output_dir:\n", - " utils.mkdir(args.output_dir)\n", - " print(args)\n", - "\n", - " device = torch.device(args.device)\n", - "\n", - " if args.use_deterministic_algorithms:\n", - " torch.backends.cudnn.benchmark = False\n", - " torch.use_deterministic_algorithms(True)\n", - " else:\n", - " torch.backends.cudnn.benchmark = True\n", - "\n", - " train_dir = os.path.join(args.data_path, \"train\")\n", - " val_dir = os.path.join(args.data_path, \"val\")\n", - " dataset, dataset_test, train_sampler, test_sampler = load_data(train_dir, val_dir)\n", - "\n", - " num_classes = len(dataset.classes)\n", - " data_loader = torch.utils.data.DataLoader(dataset, batch_size=args.batch_size, sampler=train_sampler, num_workers=16, pin_memory=True, collate_fn=None)\n", - " data_loader_test = torch.utils.data.DataLoader(dataset_test, batch_size=64, sampler=test_sampler, num_workers=16, pin_memory=True)\n", - "\n", - " print(\"Creating model\")\n", - " model = torchvision.models.get_model(args.model, weights=args.weights, num_classes=num_classes)\n", - " if args.apply_trp:\n", - " model = apply_trp(model, args.trp_depths, args.trp_planes, args.trp_lambdas)\n", - " model.to(device)\n", - "\n", - " parameters = utils.set_weight_decay(model, args.weight_decay, norm_weight_decay=None, custom_keys_weight_decay=None)\n", - " optimizer = torch.optim.SGD(parameters, lr=args.lr, momentum=args.momentum, weight_decay=args.weight_decay, nesterov=False)\n", - "\n", - " main_lr_scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=args.lr_step_size, gamma=args.lr_gamma)\n", - " warmup_lr_scheduler = torch.optim.lr_scheduler.ConstantLR(optimizer, factor=args.lr_warmup_decay, total_iters=args.lr_warmup_epochs)\n", - " lr_scheduler = torch.optim.lr_scheduler.SequentialLR(optimizer, schedulers=[warmup_lr_scheduler, main_lr_scheduler], milestones=[args.lr_warmup_epochs])\n", - "\n", - "\n", - " print(\"Start training\")\n", - " start_time = time.time()\n", - " for epoch in range(args.epochs):\n", - " train_one_epoch(model, optimizer, data_loader, device, epoch, args)\n", - " lr_scheduler.step()\n", - " evaluate(model, nn.CrossEntropyLoss(), data_loader_test, device=device)\n", - " if args.output_dir:\n", - " checkpoint = {\n", - " \"model\": model.state_dict() if not args.apply_trp else {k: v for k, v in model.state_dict().items() if not k.startswith(\"trp_blocks\")}, # NOTE: remove TRP heads\n", - " \"optimizer\": optimizer.state_dict(),\n", - " \"lr_scheduler\": lr_scheduler.state_dict(),\n", - " \"epoch\": epoch,\n", - " \"args\": args,\n", - " }\n", - " utils.save_on_master(checkpoint, os.path.join(args.output_dir, f\"model_{epoch}.pth\"))\n", - " utils.save_on_master(checkpoint, os.path.join(args.output_dir, \"checkpoint.pth\"))\n", - "\n", - " total_time = time.time() - start_time\n", - " total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n", - " print(f\"Training time {total_time_str}\")\n" - ] - }, - { - "cell_type": "code", - "execution_count": 8, - "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "namespace(data_path='/home/cs/Documents/datasets/imagenet', model='resnet18', device='cuda', batch_size=256, epochs=10, opt='sgd', lr=0.0004, momentum=0.9, weight_decay=0.0001, lr_warmup_epochs=1, lr_warmup_decay=0.0, lr_step_size=2, lr_gamma=0.5, print_freq=100, output_dir='resnet18', use_deterministic_algorithms=False, weights='ResNet18_Weights.IMAGENET1K_V1', apply_trp=True, trp_depths=[1, 1, 1], trp_planes=256, trp_lambdas=[0.4, 0.2, 0.1])\n", - "Loading data\n", - "Loading training data\n", - "Took 2.6400649547576904\n", - "Loading validation data\n", - "Creating data loaders\n", - "Creating model\n", - "✅ Applying TRP to ResNet for Image Classification...\n", - "Start training\n", - "Epoch: [0] [ 0/5005] eta: 5:44:15 lr: 0.0 img/s: 492.1456954856547 loss: 0.7194 (0.7194) acc1: 69.1406 (69.1406) acc5: 86.3281 (86.3281) meanQV: 1.4840 (1.4840) stdQV: 0.3240 (0.3240) time: 4.1269 data: 3.6067 max mem: 8962\n", - "Epoch: [0] [ 100/5005] eta: 0:46:06 lr: 0.0 img/s: 487.81080383655046 loss: 0.7315 (0.7366) acc1: 68.7500 (68.9434) acc5: 87.1094 (87.2022) meanQV: 1.4813 (1.4826) stdQV: 0.3240 (0.3238) time: 0.5212 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [ 200/5005] eta: 0:43:26 lr: 0.0 img/s: 494.6728449606332 loss: 0.7197 (0.7298) acc1: 68.7500 (69.0318) acc5: 87.5000 (87.2785) meanQV: 1.4813 (1.4832) stdQV: 0.3251 (0.3236) time: 0.5211 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [ 300/5005] eta: 0:41:58 lr: 0.0 img/s: 487.94536242744385 loss: 0.7055 (0.7270) acc1: 70.7031 (69.1757) acc5: 86.7188 (87.3053) meanQV: 1.4949 (1.4842) stdQV: 0.3192 (0.3232) time: 0.5198 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [ 400/5005] eta: 0:40:48 lr: 0.0 img/s: 493.8419675273725 loss: 0.7652 (0.7331) acc1: 67.9688 (68.9867) acc5: 87.1094 (87.2409) meanQV: 1.4758 (1.4829) stdQV: 0.3262 (0.3237) time: 0.5219 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [ 500/5005] eta: 0:39:45 lr: 0.0 img/s: 490.99324013193245 loss: 0.7112 (0.7341) acc1: 69.1406 (68.9067) acc5: 87.1094 (87.2653) meanQV: 1.4840 (1.4823) stdQV: 0.3228 (0.3239) time: 0.5201 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [ 600/5005] eta: 0:38:46 lr: 0.0 img/s: 493.63241471415074 loss: 0.7323 (0.7309) acc1: 69.5312 (69.0171) acc5: 87.8906 (87.3297) meanQV: 1.4867 (1.4831) stdQV: 0.3216 (0.3236) time: 0.5214 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [ 700/5005] eta: 0:37:49 lr: 0.0 img/s: 487.50804150353844 loss: 0.7489 (0.7310) acc1: 68.7500 (69.0442) acc5: 86.3281 (87.3167) meanQV: 1.4813 (1.4833) stdQV: 0.3240 (0.3235) time: 0.5206 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [ 800/5005] eta: 0:36:53 lr: 0.0 img/s: 491.46023226850616 loss: 0.7270 (0.7316) acc1: 69.1406 (69.0567) acc5: 87.8906 (87.3166) meanQV: 1.4840 (1.4834) stdQV: 0.3216 (0.3235) time: 0.5220 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [ 900/5005] eta: 0:35:58 lr: 0.0 img/s: 490.80023055787785 loss: 0.7573 (0.7338) acc1: 68.3594 (68.9898) acc5: 86.7188 (87.3010) meanQV: 1.4785 (1.4829) stdQV: 0.3262 (0.3237) time: 0.5208 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [1000/5005] eta: 0:35:03 lr: 0.0 img/s: 493.60609129495185 loss: 0.7472 (0.7342) acc1: 68.3594 (69.0087) acc5: 87.1094 (87.2924) meanQV: 1.4785 (1.4831) stdQV: 0.3262 (0.3237) time: 0.5212 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [1100/5005] eta: 0:34:09 lr: 0.0 img/s: 493.3883500186788 loss: 0.7273 (0.7337) acc1: 68.3594 (69.0157) acc5: 86.3281 (87.2832) meanQV: 1.4785 (1.4831) stdQV: 0.3251 (0.3236) time: 0.5200 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [1200/5005] eta: 0:33:15 lr: 0.0 img/s: 488.9129516588334 loss: 0.7296 (0.7333) acc1: 67.9688 (69.0434) acc5: 87.1094 (87.2827) meanQV: 1.4758 (1.4833) stdQV: 0.3251 (0.3235) time: 0.5207 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [1300/5005] eta: 0:32:22 lr: 0.0 img/s: 496.11964440830207 loss: 0.7214 (0.7334) acc1: 69.1406 (69.0587) acc5: 86.7188 (87.2835) meanQV: 1.4840 (1.4834) stdQV: 0.3240 (0.3235) time: 0.5209 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [1400/5005] eta: 0:31:29 lr: 0.0 img/s: 489.82739879256343 loss: 0.7335 (0.7330) acc1: 68.3594 (69.0832) acc5: 86.7188 (87.2881) meanQV: 1.4785 (1.4836) stdQV: 0.3240 (0.3234) time: 0.5208 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [1500/5005] eta: 0:30:35 lr: 0.0 img/s: 488.2193641437147 loss: 0.7221 (0.7324) acc1: 69.1406 (69.0979) acc5: 87.8906 (87.3184) meanQV: 1.4840 (1.4837) stdQV: 0.3228 (0.3234) time: 0.5221 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [1600/5005] eta: 0:29:43 lr: 0.0 img/s: 491.9973130670899 loss: 0.7626 (0.7324) acc1: 68.3594 (69.0891) acc5: 86.7188 (87.3175) meanQV: 1.4785 (1.4836) stdQV: 0.3251 (0.3234) time: 0.5204 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [1700/5005] eta: 0:28:50 lr: 0.0 img/s: 491.90219695845224 loss: 0.7566 (0.7321) acc1: 70.3125 (69.0922) acc5: 87.1094 (87.3330) meanQV: 1.4922 (1.4836) stdQV: 0.3192 (0.3234) time: 0.5211 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [1800/5005] eta: 0:27:57 lr: 0.0 img/s: 492.04353203565745 loss: 0.7514 (0.7323) acc1: 67.9688 (69.0897) acc5: 86.3281 (87.3206) meanQV: 1.4758 (1.4836) stdQV: 0.3262 (0.3234) time: 0.5207 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [1900/5005] eta: 0:27:04 lr: 0.0 img/s: 489.4212747065615 loss: 0.6998 (0.7323) acc1: 68.7500 (69.0662) acc5: 88.2812 (87.3253) meanQV: 1.4813 (1.4835) stdQV: 0.3240 (0.3235) time: 0.5216 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [2000/5005] eta: 0:26:11 lr: 0.0 img/s: 495.7586395402095 loss: 0.7553 (0.7335) acc1: 67.9688 (69.0319) acc5: 87.5000 (87.2974) meanQV: 1.4758 (1.4832) stdQV: 0.3273 (0.3236) time: 0.5215 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [2100/5005] eta: 0:25:19 lr: 0.0 img/s: 489.2644978319551 loss: 0.7095 (0.7331) acc1: 68.7500 (69.0454) acc5: 87.5000 (87.3048) meanQV: 1.4813 (1.4833) stdQV: 0.3240 (0.3235) time: 0.5210 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [2200/5005] eta: 0:24:26 lr: 0.0 img/s: 496.86875392465674 loss: 0.7402 (0.7334) acc1: 70.3125 (69.0735) acc5: 86.7188 (87.3154) meanQV: 1.4922 (1.4835) stdQV: 0.3204 (0.3234) time: 0.5210 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [2300/5005] eta: 0:23:34 lr: 0.0 img/s: 494.81235650908434 loss: 0.7301 (0.7332) acc1: 68.7500 (69.0787) acc5: 87.1094 (87.3180) meanQV: 1.4813 (1.4836) stdQV: 0.3251 (0.3234) time: 0.5205 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [2400/5005] eta: 0:22:41 lr: 0.0 img/s: 492.8439314901557 loss: 0.7250 (0.7332) acc1: 69.5312 (69.0799) acc5: 86.7188 (87.3132) meanQV: 1.4867 (1.4836) stdQV: 0.3228 (0.3234) time: 0.5205 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [2500/5005] eta: 0:21:49 lr: 0.0 img/s: 496.70487248638824 loss: 0.7133 (0.7329) acc1: 67.9688 (69.0827) acc5: 86.3281 (87.3009) meanQV: 1.4758 (1.4836) stdQV: 0.3262 (0.3234) time: 0.5208 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [2600/5005] eta: 0:20:56 lr: 0.0 img/s: 491.3833130219057 loss: 0.7494 (0.7331) acc1: 70.3125 (69.0795) acc5: 86.7188 (87.2938) meanQV: 1.4922 (1.4836) stdQV: 0.3204 (0.3234) time: 0.5203 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [2700/5005] eta: 0:20:04 lr: 0.0 img/s: 488.37014076537906 loss: 0.7300 (0.7330) acc1: 68.7500 (69.0900) acc5: 87.1094 (87.2955) meanQV: 1.4813 (1.4836) stdQV: 0.3240 (0.3234) time: 0.5204 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [2800/5005] eta: 0:19:12 lr: 0.0 img/s: 491.6548870066586 loss: 0.7110 (0.7331) acc1: 69.5312 (69.0993) acc5: 88.2812 (87.3015) meanQV: 1.4867 (1.4837) stdQV: 0.3228 (0.3234) time: 0.5216 data: 0.0005 max mem: 8962\n", - "Epoch: [0] [2900/5005] eta: 0:18:19 lr: 0.0 img/s: 490.05632164917586 loss: 0.7100 (0.7334) acc1: 69.5312 (69.0961) acc5: 87.5000 (87.3092) meanQV: 1.4867 (1.4837) stdQV: 0.3228 (0.3234) time: 0.5212 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [3000/5005] eta: 0:17:27 lr: 0.0 img/s: 491.02018376984734 loss: 0.7217 (0.7338) acc1: 69.5312 (69.0922) acc5: 88.2812 (87.3148) meanQV: 1.4867 (1.4836) stdQV: 0.3228 (0.3234) time: 0.5212 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [3100/5005] eta: 0:16:35 lr: 0.0 img/s: 488.71979426958876 loss: 0.6803 (0.7335) acc1: 68.3594 (69.0926) acc5: 87.1094 (87.3073) meanQV: 1.4785 (1.4836) stdQV: 0.3251 (0.3234) time: 0.5211 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [3200/5005] eta: 0:15:42 lr: 0.0 img/s: 492.2176638610237 loss: 0.7217 (0.7332) acc1: 70.7031 (69.1058) acc5: 87.5000 (87.3127) meanQV: 1.4949 (1.4837) stdQV: 0.3192 (0.3233) time: 0.5208 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [3300/5005] eta: 0:14:50 lr: 0.0 img/s: 491.9402840653201 loss: 0.7205 (0.7333) acc1: 69.1406 (69.1047) acc5: 87.1094 (87.3212) meanQV: 1.4840 (1.4837) stdQV: 0.3216 (0.3233) time: 0.5203 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [3400/5005] eta: 0:13:58 lr: 0.0 img/s: 487.95378866338956 loss: 0.7223 (0.7333) acc1: 69.5312 (69.0997) acc5: 86.7188 (87.3139) meanQV: 1.4867 (1.4837) stdQV: 0.3216 (0.3234) time: 0.5214 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [3500/5005] eta: 0:13:05 lr: 0.0 img/s: 1043.7659108075188 loss: 0.7439 (0.7335) acc1: 69.1406 (69.0932) acc5: 86.3281 (87.3065) meanQV: 1.4840 (1.4837) stdQV: 0.3216 (0.3234) time: 0.4635 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [3600/5005] eta: 0:12:07 lr: 0.0 img/s: 1048.0888024707265 loss: 0.7388 (0.7336) acc1: 69.1406 (69.0959) acc5: 86.7188 (87.3047) meanQV: 1.4840 (1.4837) stdQV: 0.3228 (0.3234) time: 0.2795 data: 0.0011 max mem: 8962\n", - "Epoch: [0] [3700/5005] eta: 0:11:15 lr: 0.0 img/s: 491.45843271282087 loss: 0.7387 (0.7335) acc1: 69.5312 (69.0927) acc5: 86.7188 (87.3025) meanQV: 1.4867 (1.4836) stdQV: 0.3228 (0.3234) time: 0.5209 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [3800/5005] eta: 0:10:23 lr: 0.0 img/s: 496.6446795362795 loss: 0.7232 (0.7336) acc1: 69.9219 (69.0989) acc5: 87.8906 (87.3037) meanQV: 1.4895 (1.4837) stdQV: 0.3204 (0.3234) time: 0.5210 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [3900/5005] eta: 0:09:32 lr: 0.0 img/s: 492.508007014193 loss: 0.7213 (0.7336) acc1: 67.9688 (69.0933) acc5: 87.1094 (87.2962) meanQV: 1.4758 (1.4837) stdQV: 0.3273 (0.3234) time: 0.5213 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [4000/5005] eta: 0:08:40 lr: 0.0 img/s: 488.4174580719702 loss: 0.7132 (0.7336) acc1: 68.7500 (69.0951) acc5: 87.5000 (87.2945) meanQV: 1.4813 (1.4837) stdQV: 0.3240 (0.3234) time: 0.5211 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [4100/5005] eta: 0:07:48 lr: 0.0 img/s: 495.24666769675684 loss: 0.7654 (0.7338) acc1: 68.3594 (69.0866) acc5: 87.1094 (87.2870) meanQV: 1.4785 (1.4836) stdQV: 0.3251 (0.3234) time: 0.5212 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [4200/5005] eta: 0:06:56 lr: 0.0 img/s: 490.48949008347176 loss: 0.7638 (0.7341) acc1: 66.7969 (69.0754) acc5: 86.7188 (87.2836) meanQV: 1.4676 (1.4835) stdQV: 0.3262 (0.3234) time: 0.5206 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [4300/5005] eta: 0:06:05 lr: 0.0 img/s: 487.7678139386137 loss: 0.7096 (0.7339) acc1: 69.1406 (69.0847) acc5: 87.1094 (87.2826) meanQV: 1.4840 (1.4836) stdQV: 0.3240 (0.3234) time: 0.5218 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [4400/5005] eta: 0:05:13 lr: 0.0 img/s: 493.6394498964211 loss: 0.7463 (0.7340) acc1: 68.3594 (69.0806) acc5: 87.8906 (87.2854) meanQV: 1.4785 (1.4836) stdQV: 0.3262 (0.3234) time: 0.5201 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [4500/5005] eta: 0:04:21 lr: 0.0 img/s: 485.9301278068194 loss: 0.7351 (0.7337) acc1: 69.1406 (69.0835) acc5: 87.1094 (87.2907) meanQV: 1.4840 (1.4836) stdQV: 0.3240 (0.3234) time: 0.5212 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [4600/5005] eta: 0:03:29 lr: 0.0 img/s: 488.56480022495765 loss: 0.7084 (0.7340) acc1: 68.3594 (69.0736) acc5: 86.7188 (87.2877) meanQV: 1.4785 (1.4835) stdQV: 0.3262 (0.3234) time: 0.5211 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [4700/5005] eta: 0:02:38 lr: 0.0 img/s: 494.1863106933297 loss: 0.7403 (0.7337) acc1: 68.7500 (69.0823) acc5: 87.1094 (87.2872) meanQV: 1.4813 (1.4836) stdQV: 0.3251 (0.3234) time: 0.5210 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [4800/5005] eta: 0:01:46 lr: 0.0 img/s: 491.70689603851804 loss: 0.7220 (0.7338) acc1: 71.0938 (69.0803) acc5: 86.7188 (87.2814) meanQV: 1.4977 (1.4836) stdQV: 0.3167 (0.3234) time: 0.5212 data: 0.0003 max mem: 8962\n", - "Epoch: [0] [4900/5005] eta: 0:00:54 lr: 0.0 img/s: 493.2093111277595 loss: 0.7027 (0.7336) acc1: 69.5312 (69.0827) acc5: 86.3281 (87.2757) meanQV: 1.4867 (1.4836) stdQV: 0.3228 (0.3234) time: 0.5207 data: 0.0004 max mem: 8962\n", - "Epoch: [0] [5000/5005] eta: 0:00:02 lr: 0.0 img/s: 488.95235848734404 loss: 0.7240 (0.7339) acc1: 69.5312 (69.0705) acc5: 86.7188 (87.2706) meanQV: 1.4867 (1.4835) stdQV: 0.3216 (0.3234) time: 0.5214 data: 0.0002 max mem: 8962\n", - "Epoch: [0] Total time: 0:43:17\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "/home/cs/anaconda3/envs/csenv/lib/python3.9/site-packages/torch/optim/lr_scheduler.py:243: UserWarning: The epoch parameter in `scheduler.step()` was not necessary and is being deprecated where possible. Please use `scheduler.step()` to step the scheduler. During the deprecation, if epoch is different from None, the closed form is used instead of the new chainable form, where available. Please open an issue if you are unable to replicate your use case: https://github.com/pytorch/pytorch/issues/new/choose.\n", - " warnings.warn(EPOCH_DEPRECATION_WARNING, UserWarning)\n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Test: [ 0/6250] eta: 2:26:50 loss: 0.8858 (0.8858) acc1: 75.0000 (75.0000) acc5: 100.0000 (100.0000) time: 1.4097 data: 0.7689 max mem: 8962\n", - "Test: [ 100/6250] eta: 0:02:03 loss: 0.2054 (0.6172) acc1: 100.0000 (83.7871) acc5: 100.0000 (95.6683) time: 0.0066 data: 0.0006 max mem: 8962\n", - "Test: [ 200/6250] eta: 0:01:22 loss: 0.6502 (0.6427) acc1: 87.5000 (84.3284) acc5: 87.5000 (95.0249) time: 0.0064 data: 0.0007 max mem: 8962\n", - "Test: [ 300/6250] eta: 0:01:08 loss: 1.1768 (0.8467) acc1: 62.5000 (78.3638) acc5: 87.5000 (93.2724) time: 0.0075 data: 0.0010 max mem: 8962\n", - "Test: [ 400/6250] eta: 0:00:59 loss: 1.1007 (0.9660) acc1: 75.0000 (75.4676) acc5: 87.5000 (92.2070) time: 0.0074 data: 0.0017 max mem: 8962\n", - "Test: [ 500/6250] eta: 0:00:55 loss: 1.1372 (1.0088) acc1: 75.0000 (74.0519) acc5: 87.5000 (91.8663) time: 0.0080 data: 0.0021 max mem: 8962\n", - "Test: [ 600/6250] eta: 0:00:51 loss: 0.5046 (0.9217) acc1: 87.5000 (76.2479) acc5: 100.0000 (92.6789) time: 0.0071 data: 0.0009 max mem: 8962\n", - "Test: [ 700/6250] eta: 0:00:49 loss: 0.5742 (0.9105) acc1: 87.5000 (76.6762) acc5: 100.0000 (92.7068) time: 0.0052 data: 0.0006 max mem: 8962\n", - "Test: [ 800/6250] eta: 0:00:47 loss: 0.8010 (0.9392) acc1: 75.0000 (76.1392) acc5: 87.5000 (92.3221) time: 0.0077 data: 0.0010 max mem: 8962\n", - "Test: [ 900/6250] eta: 0:00:46 loss: 0.4006 (0.8848) acc1: 87.5000 (77.4695) acc5: 100.0000 (92.8413) time: 0.0066 data: 0.0006 max mem: 8962\n", - "Test: [1000/6250] eta: 0:00:44 loss: 0.8992 (0.8725) acc1: 75.0000 (77.6474) acc5: 100.0000 (92.9570) time: 0.0070 data: 0.0007 max mem: 8962\n", - "Test: [1100/6250] eta: 0:00:43 loss: 1.0749 (0.9068) acc1: 62.5000 (76.6803) acc5: 100.0000 (92.8020) time: 0.0079 data: 0.0019 max mem: 8962\n", - "Test: [1200/6250] eta: 0:00:42 loss: 1.0901 (0.9146) acc1: 75.0000 (76.3114) acc5: 87.5000 (92.8809) time: 0.0081 data: 0.0007 max mem: 8962\n", - "Test: [1300/6250] eta: 0:00:41 loss: 0.6278 (0.9170) acc1: 75.0000 (76.0857) acc5: 100.0000 (93.0630) time: 0.0062 data: 0.0006 max mem: 8962\n", - "Test: [1400/6250] eta: 0:00:39 loss: 0.8946 (0.9150) acc1: 75.0000 (76.1153) acc5: 87.5000 (93.0496) time: 0.0070 data: 0.0006 max mem: 8962\n", - "Test: [1500/6250] eta: 0:00:38 loss: 0.7951 (0.9198) acc1: 75.0000 (75.8328) acc5: 100.0000 (93.0963) time: 0.0068 data: 0.0006 max mem: 8962\n", - "Test: [1600/6250] eta: 0:00:37 loss: 0.2566 (0.9163) acc1: 87.5000 (75.6636) acc5: 100.0000 (93.2386) time: 0.0072 data: 0.0007 max mem: 8962\n", - "Test: [1700/6250] eta: 0:00:36 loss: 1.0530 (0.9138) acc1: 75.0000 (75.6393) acc5: 100.0000 (93.3789) time: 0.0087 data: 0.0016 max mem: 8962\n", - "Test: [1800/6250] eta: 0:00:35 loss: 1.1882 (0.9209) acc1: 62.5000 (75.3956) acc5: 87.5000 (93.3509) time: 0.0053 data: 0.0004 max mem: 8962\n", - "Test: [1900/6250] eta: 0:00:34 loss: 1.0323 (0.9166) acc1: 62.5000 (75.6707) acc5: 100.0000 (93.4508) time: 0.0059 data: 0.0006 max mem: 8962\n", - "Test: [2000/6250] eta: 0:00:33 loss: 0.4827 (0.9227) acc1: 75.0000 (75.5622) acc5: 100.0000 (93.4283) time: 0.0062 data: 0.0005 max mem: 8962\n", - "Test: [2100/6250] eta: 0:00:32 loss: 0.4458 (0.9042) acc1: 87.5000 (76.1185) acc5: 100.0000 (93.5983) time: 0.0059 data: 0.0005 max mem: 8962\n", - "Test: [2200/6250] eta: 0:00:31 loss: 0.1983 (0.8953) acc1: 87.5000 (76.3233) acc5: 100.0000 (93.6620) time: 0.0058 data: 0.0005 max mem: 8962\n", - "Test: [2300/6250] eta: 0:00:30 loss: 0.6167 (0.8971) acc1: 87.5000 (76.2929) acc5: 100.0000 (93.6658) time: 0.0067 data: 0.0010 max mem: 8962\n", - "Test: [2400/6250] eta: 0:00:29 loss: 1.1606 (0.9060) acc1: 62.5000 (76.1245) acc5: 87.5000 (93.5652) time: 0.0069 data: 0.0006 max mem: 8962\n", - "Test: [2500/6250] eta: 0:00:29 loss: 0.9882 (0.9067) acc1: 75.0000 (76.2095) acc5: 100.0000 (93.5176) time: 0.0063 data: 0.0005 max mem: 8962\n", - "Test: [2600/6250] eta: 0:00:27 loss: 2.1575 (0.9289) acc1: 50.0000 (75.7785) acc5: 75.0000 (93.2045) time: 0.0058 data: 0.0006 max mem: 8962\n", - "Test: [2700/6250] eta: 0:00:27 loss: 0.6626 (0.9389) acc1: 75.0000 (75.6248) acc5: 100.0000 (93.0859) time: 0.0069 data: 0.0006 max mem: 8962\n", - "Test: [2800/6250] eta: 0:00:26 loss: 1.7137 (0.9616) acc1: 50.0000 (75.1205) acc5: 87.5000 (92.8106) time: 0.0062 data: 0.0006 max mem: 8962\n", - "Test: [2900/6250] eta: 0:00:25 loss: 1.7213 (0.9804) acc1: 50.0000 (74.7630) acc5: 87.5000 (92.5758) time: 0.0076 data: 0.0012 max mem: 8962\n", - "Test: [3000/6250] eta: 0:00:24 loss: 2.1883 (1.0015) acc1: 50.0000 (74.4585) acc5: 75.0000 (92.2567) time: 0.0066 data: 0.0013 max mem: 8962\n", - "Test: [3100/6250] eta: 0:00:23 loss: 1.9025 (1.0238) acc1: 50.0000 (73.9600) acc5: 87.5000 (91.9824) time: 0.0069 data: 0.0007 max mem: 8962\n", - "Test: [3200/6250] eta: 0:00:22 loss: 0.7818 (1.0439) acc1: 75.0000 (73.5786) acc5: 100.0000 (91.7252) time: 0.0066 data: 0.0009 max mem: 8962\n", - "Test: [3300/6250] eta: 0:00:21 loss: 1.3260 (1.0571) acc1: 50.0000 (73.2316) acc5: 87.5000 (91.5972) time: 0.0062 data: 0.0010 max mem: 8962\n", - "Test: [3400/6250] eta: 0:00:21 loss: 2.2889 (1.0708) acc1: 50.0000 (72.9638) acc5: 75.0000 (91.4069) time: 0.0054 data: 0.0004 max mem: 8962\n", - "Test: [3500/6250] eta: 0:00:20 loss: 1.8139 (1.0788) acc1: 62.5000 (72.8399) acc5: 87.5000 (91.2704) time: 0.0056 data: 0.0005 max mem: 8962\n", - "Test: [3600/6250] eta: 0:00:19 loss: 0.5227 (1.0757) acc1: 75.0000 (72.9797) acc5: 100.0000 (91.2837) time: 0.0061 data: 0.0004 max mem: 8962\n", - "Test: [3700/6250] eta: 0:00:18 loss: 1.5864 (1.0885) acc1: 50.0000 (72.7101) acc5: 87.5000 (91.0970) time: 0.0110 data: 0.0061 max mem: 8962\n", - "Test: [3800/6250] eta: 0:00:18 loss: 0.7844 (1.0958) acc1: 87.5000 (72.5993) acc5: 87.5000 (90.9826) time: 0.0059 data: 0.0005 max mem: 8962\n", - "Test: [3900/6250] eta: 0:00:17 loss: 2.1642 (1.1114) acc1: 37.5000 (72.2699) acc5: 75.0000 (90.7684) time: 0.0070 data: 0.0013 max mem: 8962\n", - "Test: [4000/6250] eta: 0:00:16 loss: 1.3791 (1.1249) acc1: 37.5000 (71.9789) acc5: 87.5000 (90.6086) time: 0.0061 data: 0.0009 max mem: 8962\n", - "Test: [4100/6250] eta: 0:00:15 loss: 1.8430 (1.1361) acc1: 50.0000 (71.8026) acc5: 87.5000 (90.4566) time: 0.0058 data: 0.0007 max mem: 8962\n", - "Test: [4200/6250] eta: 0:00:14 loss: 0.5273 (1.1408) acc1: 87.5000 (71.6556) acc5: 100.0000 (90.4428) time: 0.0054 data: 0.0004 max mem: 8962\n", - "Test: [4300/6250] eta: 0:00:14 loss: 0.5790 (1.1499) acc1: 75.0000 (71.5124) acc5: 87.5000 (90.3075) time: 0.0062 data: 0.0009 max mem: 8962\n", - "Test: [4400/6250] eta: 0:00:13 loss: 0.8274 (1.1557) acc1: 75.0000 (71.3730) acc5: 100.0000 (90.2721) time: 0.0063 data: 0.0009 max mem: 8962\n", - "Test: [4500/6250] eta: 0:00:12 loss: 1.3455 (1.1605) acc1: 75.0000 (71.3008) acc5: 87.5000 (90.2327) time: 0.0059 data: 0.0007 max mem: 8962\n", - "Test: [4600/6250] eta: 0:00:11 loss: 1.6417 (1.1701) acc1: 62.5000 (71.1476) acc5: 87.5000 (90.0619) time: 0.0066 data: 0.0008 max mem: 8962\n", - "Test: [4700/6250] eta: 0:00:11 loss: 1.4379 (1.1789) acc1: 62.5000 (70.8971) acc5: 87.5000 (89.9197) time: 0.0057 data: 0.0004 max mem: 8962\n", - "Test: [4800/6250] eta: 0:00:10 loss: 1.3181 (1.1859) acc1: 62.5000 (70.7899) acc5: 87.5000 (89.8198) time: 0.0061 data: 0.0006 max mem: 8962\n", - "Test: [4900/6250] eta: 0:00:09 loss: 0.5202 (1.1891) acc1: 87.5000 (70.7432) acc5: 100.0000 (89.7648) time: 0.0052 data: 0.0005 max mem: 8962\n", - "Test: [5000/6250] eta: 0:00:08 loss: 2.0605 (1.1999) acc1: 62.5000 (70.5934) acc5: 75.0000 (89.6296) time: 0.0062 data: 0.0012 max mem: 8962\n", - "Test: [5100/6250] eta: 0:00:08 loss: 1.3460 (1.2044) acc1: 62.5000 (70.5229) acc5: 87.5000 (89.5976) time: 0.0080 data: 0.0008 max mem: 8962\n", - "Test: [5200/6250] eta: 0:00:07 loss: 1.1308 (1.2106) acc1: 75.0000 (70.4264) acc5: 87.5000 (89.5236) time: 0.0055 data: 0.0005 max mem: 8962\n", - "Test: [5300/6250] eta: 0:00:06 loss: 1.5583 (1.2231) acc1: 62.5000 (70.1354) acc5: 87.5000 (89.3157) time: 0.0069 data: 0.0016 max mem: 8962\n", - "Test: [5400/6250] eta: 0:00:06 loss: 0.8486 (1.2267) acc1: 75.0000 (70.0542) acc5: 87.5000 (89.2728) time: 0.0064 data: 0.0012 max mem: 8962\n", - "Test: [5500/6250] eta: 0:00:05 loss: 1.0672 (1.2307) acc1: 62.5000 (69.9509) acc5: 87.5000 (89.2179) time: 0.0071 data: 0.0005 max mem: 8962\n", - "Test: [5600/6250] eta: 0:00:04 loss: 1.0070 (1.2359) acc1: 62.5000 (69.8648) acc5: 87.5000 (89.1202) time: 0.0096 data: 0.0041 max mem: 8962\n", - "Test: [5700/6250] eta: 0:00:03 loss: 2.0962 (1.2487) acc1: 37.5000 (69.5843) acc5: 87.5000 (89.0019) time: 0.0070 data: 0.0006 max mem: 8962\n", - "Test: [5800/6250] eta: 0:00:03 loss: 0.7288 (1.2454) acc1: 75.0000 (69.6582) acc5: 100.0000 (89.0471) time: 0.0061 data: 0.0005 max mem: 8962\n", - "Test: [5900/6250] eta: 0:00:02 loss: 1.1190 (1.2431) acc1: 75.0000 (69.6873) acc5: 100.0000 (89.0781) time: 0.0067 data: 0.0006 max mem: 8962\n", - "Test: [6000/6250] eta: 0:00:01 loss: 0.6253 (1.2358) acc1: 87.5000 (69.8425) acc5: 87.5000 (89.1622) time: 0.0068 data: 0.0006 max mem: 8962\n", - "Test: [6100/6250] eta: 0:00:01 loss: 1.1478 (1.2438) acc1: 62.5000 (69.6423) acc5: 100.0000 (89.0776) time: 0.0069 data: 0.0006 max mem: 8962\n", - "Test: [6200/6250] eta: 0:00:00 loss: 0.2636 (1.2384) acc1: 87.5000 (69.7650) acc5: 100.0000 (89.1469) time: 0.0071 data: 0.0024 max mem: 8962\n", - "Test: Total time: 0:00:44\n", - "Test: Acc@1 69.802 Acc@5 89.166\n", - "Epoch: [1] [ 0/5005] eta: 5:42:56 lr: 0.0004 img/s: 480.9318095087941 loss: 0.7240 (0.7240) acc1: 67.1875 (67.1875) acc5: 89.4531 (89.4531) meanQV: 1.4703 (1.4703) stdQV: 0.3293 (0.3293) time: 4.1113 data: 3.5789 max mem: 8962\n", - "Epoch: [1] [ 100/5005] eta: 0:45:51 lr: 0.0004 img/s: 493.27547238807784 loss: 0.7663 (0.7544) acc1: 68.3594 (68.9047) acc5: 85.9375 (87.0166) meanQV: 1.4785 (1.4823) stdQV: 0.3262 (0.3240) time: 0.5211 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [ 200/5005] eta: 0:43:21 lr: 0.0004 img/s: 494.15947332208856 loss: 0.7673 (0.7576) acc1: 68.7500 (69.0512) acc5: 86.3281 (86.9928) meanQV: 1.4813 (1.4833) stdQV: 0.3240 (0.3234) time: 0.5218 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [ 300/5005] eta: 0:41:55 lr: 0.0004 img/s: 490.8006792413158 loss: 0.7602 (0.7692) acc1: 69.9219 (68.8824) acc5: 88.2812 (86.9770) meanQV: 1.4895 (1.4821) stdQV: 0.3204 (0.3239) time: 0.5209 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [ 400/5005] eta: 0:40:46 lr: 0.0004 img/s: 488.3024018976519 loss: 0.7641 (0.7740) acc1: 67.9688 (68.7734) acc5: 87.5000 (86.9837) meanQV: 1.4758 (1.4814) stdQV: 0.3262 (0.3243) time: 0.5211 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [ 500/5005] eta: 0:39:44 lr: 0.0004 img/s: 491.2133664120496 loss: 0.7673 (0.7762) acc1: 69.9219 (68.8451) acc5: 87.5000 (87.0439) meanQV: 1.4895 (1.4819) stdQV: 0.3216 (0.3241) time: 0.5217 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [ 600/5005] eta: 0:38:45 lr: 0.0004 img/s: 489.9574830025097 loss: 0.7744 (0.7785) acc1: 68.7500 (68.8845) acc5: 87.5000 (87.0379) meanQV: 1.4813 (1.4821) stdQV: 0.3240 (0.3240) time: 0.5210 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [ 700/5005] eta: 0:37:48 lr: 0.0004 img/s: 490.1766813359385 loss: 0.7570 (0.7804) acc1: 69.9219 (68.9227) acc5: 87.1094 (87.0347) meanQV: 1.4895 (1.4824) stdQV: 0.3216 (0.3238) time: 0.5210 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [ 800/5005] eta: 0:36:53 lr: 0.0004 img/s: 488.54479381319 loss: 0.7706 (0.7828) acc1: 69.1406 (68.9124) acc5: 86.3281 (87.0435) meanQV: 1.4840 (1.4823) stdQV: 0.3226 (0.3239) time: 0.5213 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [ 900/5005] eta: 0:35:58 lr: 0.0004 img/s: 486.8855385165102 loss: 0.7941 (0.7848) acc1: 68.3594 (68.9390) acc5: 87.1094 (87.0465) meanQV: 1.4785 (1.4825) stdQV: 0.3251 (0.3238) time: 0.5219 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [1000/5005] eta: 0:35:03 lr: 0.0004 img/s: 492.5010040446165 loss: 0.7671 (0.7867) acc1: 69.5312 (68.9233) acc5: 87.1094 (87.0430) meanQV: 1.4867 (1.4824) stdQV: 0.3228 (0.3238) time: 0.5216 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [1100/5005] eta: 0:34:10 lr: 0.0004 img/s: 491.8478935923928 loss: 0.7918 (0.7885) acc1: 68.3594 (68.8866) acc5: 87.1094 (86.9948) meanQV: 1.4785 (1.4821) stdQV: 0.3251 (0.3240) time: 0.5219 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [1200/5005] eta: 0:33:16 lr: 0.0004 img/s: 492.04127725035846 loss: 0.8054 (0.7904) acc1: 68.3594 (68.8863) acc5: 87.1094 (87.0059) meanQV: 1.4785 (1.4821) stdQV: 0.3240 (0.3240) time: 0.5212 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [1300/5005] eta: 0:32:22 lr: 0.0004 img/s: 493.6503434775133 loss: 0.8003 (0.7921) acc1: 68.7500 (68.8953) acc5: 86.7188 (87.0238) meanQV: 1.4813 (1.4822) stdQV: 0.3240 (0.3239) time: 0.5208 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [1400/5005] eta: 0:31:29 lr: 0.0004 img/s: 493.7590786833788 loss: 0.8469 (0.7947) acc1: 67.5781 (68.8757) acc5: 86.3281 (87.0087) meanQV: 1.4730 (1.4821) stdQV: 0.3273 (0.3240) time: 0.5214 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [1500/5005] eta: 0:30:36 lr: 0.0004 img/s: 493.85105292835783 loss: 0.8278 (0.7964) acc1: 67.9688 (68.8734) acc5: 87.1094 (87.0126) meanQV: 1.4758 (1.4820) stdQV: 0.3260 (0.3240) time: 0.5209 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [1600/5005] eta: 0:29:43 lr: 0.0004 img/s: 489.0960134101013 loss: 0.8061 (0.7976) acc1: 69.5312 (68.8713) acc5: 86.7188 (87.0096) meanQV: 1.4867 (1.4820) stdQV: 0.3228 (0.3240) time: 0.5216 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [1700/5005] eta: 0:28:50 lr: 0.0004 img/s: 487.0217960638814 loss: 0.8219 (0.7990) acc1: 69.1406 (68.8761) acc5: 86.3281 (87.0118) meanQV: 1.4840 (1.4821) stdQV: 0.3204 (0.3240) time: 0.5216 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [1800/5005] eta: 0:27:57 lr: 0.0004 img/s: 487.4064667244674 loss: 0.8217 (0.8005) acc1: 67.1875 (68.8660) acc5: 85.9375 (86.9964) meanQV: 1.4703 (1.4820) stdQV: 0.3273 (0.3240) time: 0.5210 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [1900/5005] eta: 0:27:05 lr: 0.0004 img/s: 495.115358261054 loss: 0.8361 (0.8026) acc1: 67.9688 (68.8474) acc5: 86.3281 (86.9830) meanQV: 1.4746 (1.4818) stdQV: 0.3270 (0.3240) time: 0.5206 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [2000/5005] eta: 0:26:12 lr: 0.0004 img/s: 488.846174928818 loss: 0.8257 (0.8035) acc1: 69.1406 (68.8619) acc5: 87.1094 (86.9956) meanQV: 1.4840 (1.4819) stdQV: 0.3228 (0.3240) time: 0.5216 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [2100/5005] eta: 0:25:19 lr: 0.0004 img/s: 491.2070743506753 loss: 0.8428 (0.8053) acc1: 66.7969 (68.8420) acc5: 87.1094 (86.9922) meanQV: 1.4676 (1.4818) stdQV: 0.3303 (0.3241) time: 0.5210 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [2200/5005] eta: 0:24:27 lr: 0.0004 img/s: 488.1822949084529 loss: 0.8558 (0.8064) acc1: 68.3594 (68.8570) acc5: 87.5000 (87.0222) meanQV: 1.4785 (1.4819) stdQV: 0.3251 (0.3240) time: 0.5213 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [2300/5005] eta: 0:23:34 lr: 0.0004 img/s: 491.70599535560405 loss: 0.8227 (0.8072) acc1: 68.3594 (68.8542) acc5: 87.1094 (87.0248) meanQV: 1.4785 (1.4819) stdQV: 0.3262 (0.3240) time: 0.5208 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [2400/5005] eta: 0:22:42 lr: 0.0004 img/s: 488.04228171446755 loss: 0.8344 (0.8086) acc1: 69.5312 (68.8348) acc5: 87.5000 (87.0231) meanQV: 1.4867 (1.4817) stdQV: 0.3228 (0.3241) time: 0.5212 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [2500/5005] eta: 0:21:49 lr: 0.0004 img/s: 493.9346543052109 loss: 0.8142 (0.8099) acc1: 67.1875 (68.8248) acc5: 86.3281 (87.0107) meanQV: 1.4703 (1.4817) stdQV: 0.3273 (0.3241) time: 0.5206 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [2600/5005] eta: 0:20:57 lr: 0.0004 img/s: 493.71140181999846 loss: 0.8310 (0.8114) acc1: 69.5312 (68.8209) acc5: 87.8906 (87.0080) meanQV: 1.4855 (1.4816) stdQV: 0.3216 (0.3241) time: 0.5220 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [2700/5005] eta: 0:20:04 lr: 0.0004 img/s: 490.4764950910866 loss: 0.8228 (0.8124) acc1: 68.7500 (68.8167) acc5: 87.5000 (87.0055) meanQV: 1.4813 (1.4816) stdQV: 0.3238 (0.3241) time: 0.5214 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [2800/5005] eta: 0:19:12 lr: 0.0004 img/s: 489.5152105330343 loss: 0.8128 (0.8133) acc1: 69.1406 (68.8236) acc5: 87.1094 (87.0178) meanQV: 1.4840 (1.4817) stdQV: 0.3240 (0.3241) time: 0.5212 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [2900/5005] eta: 0:18:20 lr: 0.0004 img/s: 490.35195733551564 loss: 0.8154 (0.8143) acc1: 67.5781 (68.8168) acc5: 86.7188 (87.0159) meanQV: 1.4730 (1.4816) stdQV: 0.3273 (0.3241) time: 0.5220 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [3000/5005] eta: 0:17:27 lr: 0.0004 img/s: 496.8430038188373 loss: 0.8169 (0.8153) acc1: 69.1406 (68.8008) acc5: 88.2812 (87.0080) meanQV: 1.4840 (1.4815) stdQV: 0.3228 (0.3242) time: 0.5210 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [3100/5005] eta: 0:16:35 lr: 0.0004 img/s: 488.5263449079129 loss: 0.8308 (0.8165) acc1: 70.3125 (68.7841) acc5: 88.2812 (87.0036) meanQV: 1.4922 (1.4814) stdQV: 0.3192 (0.3242) time: 0.5215 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [3200/5005] eta: 0:15:43 lr: 0.0004 img/s: 491.9060279391008 loss: 0.8449 (0.8176) acc1: 68.3594 (68.7768) acc5: 86.3281 (86.9960) meanQV: 1.4785 (1.4813) stdQV: 0.3262 (0.3242) time: 0.5214 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [3300/5005] eta: 0:14:50 lr: 0.0004 img/s: 493.2963213941315 loss: 0.8492 (0.8186) acc1: 67.9688 (68.7659) acc5: 87.1094 (86.9994) meanQV: 1.4758 (1.4812) stdQV: 0.3248 (0.3243) time: 0.5203 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [3400/5005] eta: 0:13:58 lr: 0.0004 img/s: 489.8253877234111 loss: 0.8006 (0.8190) acc1: 68.3594 (68.7695) acc5: 86.7188 (87.0057) meanQV: 1.4785 (1.4813) stdQV: 0.3192 (0.3243) time: 0.5208 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [3500/5005] eta: 0:13:06 lr: 0.0004 img/s: 487.77468294972033 loss: 0.8505 (0.8198) acc1: 68.7500 (68.7579) acc5: 85.9375 (87.0026) meanQV: 1.4813 (1.4812) stdQV: 0.3248 (0.3243) time: 0.5211 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [3600/5005] eta: 0:12:13 lr: 0.0004 img/s: 493.46703224484054 loss: 0.8441 (0.8204) acc1: 68.3594 (68.7684) acc5: 87.1094 (86.9979) meanQV: 1.4785 (1.4813) stdQV: 0.3262 (0.3243) time: 0.5203 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [3700/5005] eta: 0:11:21 lr: 0.0004 img/s: 489.55359849141115 loss: 0.8019 (0.8213) acc1: 69.9219 (68.7655) acc5: 86.7188 (86.9992) meanQV: 1.4891 (1.4812) stdQV: 0.3214 (0.3243) time: 0.5221 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [3800/5005] eta: 0:10:29 lr: 0.0004 img/s: 493.67462881592127 loss: 0.8116 (0.8217) acc1: 69.1406 (68.7733) acc5: 86.7188 (87.0059) meanQV: 1.4840 (1.4813) stdQV: 0.3216 (0.3243) time: 0.5207 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [3900/5005] eta: 0:09:37 lr: 0.0004 img/s: 491.10327860529725 loss: 0.8894 (0.8224) acc1: 67.5781 (68.7723) acc5: 85.9375 (87.0182) meanQV: 1.4730 (1.4813) stdQV: 0.3273 (0.3243) time: 0.5204 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [4000/5005] eta: 0:08:42 lr: 0.0004 img/s: 658.0155167215349 loss: 0.8130 (0.8233) acc1: 68.7500 (68.7736) acc5: 86.7188 (87.0157) meanQV: 1.4813 (1.4813) stdQV: 0.3228 (0.3243) time: 0.3829 data: 0.0022 max mem: 8962\n", - "Epoch: [1] [4100/5005] eta: 0:07:48 lr: 0.0004 img/s: 495.21469028956733 loss: 0.8295 (0.8240) acc1: 69.1406 (68.7658) acc5: 87.5000 (87.0174) meanQV: 1.4816 (1.4812) stdQV: 0.3235 (0.3243) time: 0.5208 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [4200/5005] eta: 0:06:57 lr: 0.0004 img/s: 492.8694949672994 loss: 0.8703 (0.8249) acc1: 68.3594 (68.7643) acc5: 87.1094 (87.0155) meanQV: 1.4773 (1.4812) stdQV: 0.3262 (0.3243) time: 0.5213 data: 0.0005 max mem: 8962\n", - "Epoch: [1] [4300/5005] eta: 0:06:05 lr: 0.0004 img/s: 490.85205892006695 loss: 0.8436 (0.8255) acc1: 67.5781 (68.7648) acc5: 87.1094 (87.0136) meanQV: 1.4730 (1.4812) stdQV: 0.3262 (0.3243) time: 0.5211 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [4400/5005] eta: 0:05:13 lr: 0.0004 img/s: 493.98669227672667 loss: 0.8510 (0.8258) acc1: 68.3594 (68.7773) acc5: 86.7188 (87.0168) meanQV: 1.4785 (1.4813) stdQV: 0.3251 (0.3242) time: 0.5209 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [4500/5005] eta: 0:04:21 lr: 0.0004 img/s: 493.84991723495295 loss: 0.8170 (0.8264) acc1: 68.3594 (68.7790) acc5: 88.2812 (87.0156) meanQV: 1.4785 (1.4813) stdQV: 0.3238 (0.3242) time: 0.5213 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [4600/5005] eta: 0:03:30 lr: 0.0004 img/s: 488.5125647070813 loss: 0.8377 (0.8271) acc1: 68.3594 (68.7688) acc5: 87.5000 (87.0158) meanQV: 1.4770 (1.4812) stdQV: 0.3262 (0.3243) time: 0.5200 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [4700/5005] eta: 0:02:38 lr: 0.0004 img/s: 491.9828855090406 loss: 0.8226 (0.8275) acc1: 68.3594 (68.7709) acc5: 87.8906 (87.0220) meanQV: 1.4785 (1.4812) stdQV: 0.3216 (0.3243) time: 0.5212 data: 0.0004 max mem: 8962\n", - "Epoch: [1] [4800/5005] eta: 0:01:46 lr: 0.0004 img/s: 493.26708795863277 loss: 0.8528 (0.8281) acc1: 68.7500 (68.7686) acc5: 86.3281 (87.0159) meanQV: 1.4809 (1.4812) stdQV: 0.3251 (0.3243) time: 0.5216 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [4900/5005] eta: 0:00:54 lr: 0.0004 img/s: 491.2724747144307 loss: 0.8068 (0.8286) acc1: 69.5312 (68.7636) acc5: 87.1094 (87.0159) meanQV: 1.4867 (1.4812) stdQV: 0.3228 (0.3243) time: 0.5217 data: 0.0003 max mem: 8962\n", - "Epoch: [1] [5000/5005] eta: 0:00:02 lr: 0.0004 img/s: 492.8498131172718 loss: 0.8547 (0.8292) acc1: 67.9688 (68.7634) acc5: 86.3281 (87.0148) meanQV: 1.4758 (1.4812) stdQV: 0.3262 (0.3243) time: 0.5210 data: 0.0002 max mem: 8962\n", - "Epoch: [1] Total time: 0:43:16\n", - "Test: [ 0/6250] eta: 1:21:57 loss: 0.8673 (0.8673) acc1: 62.5000 (62.5000) acc5: 100.0000 (100.0000) time: 0.7867 data: 0.7725 max mem: 8962\n", - "Test: [ 100/6250] eta: 0:01:29 loss: 0.1868 (0.6206) acc1: 100.0000 (83.5396) acc5: 100.0000 (95.4208) time: 0.0063 data: 0.0008 max mem: 8962\n", - "Test: [ 200/6250] eta: 0:01:03 loss: 0.7102 (0.6552) acc1: 75.0000 (83.7687) acc5: 100.0000 (94.6517) time: 0.0056 data: 0.0004 max mem: 8962\n", - "Test: [ 300/6250] eta: 0:00:54 loss: 1.2776 (0.8690) acc1: 62.5000 (78.6130) acc5: 87.5000 (93.0233) time: 0.0061 data: 0.0009 max mem: 8962\n", - "Test: [ 400/6250] eta: 0:00:48 loss: 0.9489 (0.9890) acc1: 75.0000 (75.8416) acc5: 87.5000 (92.0823) time: 0.0060 data: 0.0005 max mem: 8962\n", - "Test: [ 500/6250] eta: 0:00:45 loss: 1.0595 (1.0403) acc1: 75.0000 (74.4261) acc5: 87.5000 (91.6168) time: 0.0053 data: 0.0007 max mem: 8962\n", - "Test: [ 600/6250] eta: 0:00:42 loss: 0.4378 (0.9454) acc1: 87.5000 (76.4975) acc5: 100.0000 (92.3877) time: 0.0051 data: 0.0004 max mem: 8962\n", - "Test: [ 700/6250] eta: 0:00:40 loss: 0.4026 (0.9303) acc1: 87.5000 (76.7653) acc5: 100.0000 (92.4394) time: 0.0056 data: 0.0006 max mem: 8962\n", - "Test: [ 800/6250] eta: 0:00:39 loss: 0.7303 (0.9549) acc1: 75.0000 (76.2797) acc5: 87.5000 (92.0880) time: 0.0063 data: 0.0006 max mem: 8962\n", - "Test: [ 900/6250] eta: 0:00:37 loss: 0.3813 (0.8983) acc1: 87.5000 (77.5943) acc5: 100.0000 (92.6471) time: 0.0063 data: 0.0006 max mem: 8962\n", - "Test: [1000/6250] eta: 0:00:36 loss: 0.8989 (0.8855) acc1: 75.0000 (77.7473) acc5: 100.0000 (92.7947) time: 0.0056 data: 0.0007 max mem: 8962\n", - "Test: [1100/6250] eta: 0:00:35 loss: 1.1069 (0.9229) acc1: 75.0000 (76.8847) acc5: 100.0000 (92.6203) time: 0.0064 data: 0.0022 max mem: 8962\n", - "Test: [1200/6250] eta: 0:00:34 loss: 1.0983 (0.9314) acc1: 75.0000 (76.4467) acc5: 87.5000 (92.6624) time: 0.0055 data: 0.0005 max mem: 8962\n", - "Test: [1300/6250] eta: 0:00:33 loss: 0.5355 (0.9346) acc1: 75.0000 (76.1145) acc5: 100.0000 (92.8132) time: 0.0057 data: 0.0005 max mem: 8962\n", - "Test: [1400/6250] eta: 0:00:32 loss: 0.7667 (0.9310) acc1: 75.0000 (76.2313) acc5: 100.0000 (92.9069) time: 0.0052 data: 0.0004 max mem: 8962\n", - "Test: [1500/6250] eta: 0:00:31 loss: 0.8760 (0.9357) acc1: 62.5000 (75.9494) acc5: 100.0000 (92.9880) time: 0.0053 data: 0.0007 max mem: 8962\n", - "Test: [1600/6250] eta: 0:00:30 loss: 0.0993 (0.9286) acc1: 100.0000 (75.9135) acc5: 100.0000 (93.1371) time: 0.0053 data: 0.0005 max mem: 8962\n", - "Test: [1700/6250] eta: 0:00:29 loss: 1.1072 (0.9257) acc1: 62.5000 (75.8157) acc5: 100.0000 (93.2760) time: 0.0068 data: 0.0021 max mem: 8962\n", - "Test: [1800/6250] eta: 0:00:28 loss: 1.1913 (0.9373) acc1: 62.5000 (75.5622) acc5: 87.5000 (93.2537) time: 0.0054 data: 0.0006 max mem: 8962\n", - "Test: [1900/6250] eta: 0:00:28 loss: 1.0258 (0.9308) acc1: 62.5000 (75.8417) acc5: 100.0000 (93.3653) time: 0.0050 data: 0.0006 max mem: 8962\n", - "Test: [2000/6250] eta: 0:00:27 loss: 0.5154 (0.9364) acc1: 87.5000 (75.7871) acc5: 100.0000 (93.3221) time: 0.0055 data: 0.0006 max mem: 8962\n", - "Test: [2100/6250] eta: 0:00:27 loss: 0.3852 (0.9175) acc1: 87.5000 (76.3267) acc5: 100.0000 (93.4793) time: 0.0061 data: 0.0012 max mem: 8962\n", - "Test: [2200/6250] eta: 0:00:26 loss: 0.2291 (0.9102) acc1: 87.5000 (76.4652) acc5: 100.0000 (93.5654) time: 0.0063 data: 0.0007 max mem: 8962\n", - "Test: [2300/6250] eta: 0:00:25 loss: 0.6943 (0.9109) acc1: 87.5000 (76.4776) acc5: 100.0000 (93.5843) time: 0.0062 data: 0.0005 max mem: 8962\n", - "Test: [2400/6250] eta: 0:00:25 loss: 0.8836 (0.9201) acc1: 75.0000 (76.3432) acc5: 87.5000 (93.4975) time: 0.0056 data: 0.0011 max mem: 8962\n", - "Test: [2500/6250] eta: 0:00:24 loss: 0.7644 (0.9205) acc1: 87.5000 (76.4794) acc5: 100.0000 (93.4626) time: 0.0061 data: 0.0010 max mem: 8962\n", - "Test: [2600/6250] eta: 0:00:23 loss: 2.2800 (0.9451) acc1: 37.5000 (75.9948) acc5: 75.0000 (93.1469) time: 0.0055 data: 0.0006 max mem: 8962\n", - "Test: [2700/6250] eta: 0:00:23 loss: 0.7858 (0.9555) acc1: 75.0000 (75.7960) acc5: 87.5000 (93.0118) time: 0.0060 data: 0.0006 max mem: 8962\n", - "Test: [2800/6250] eta: 0:00:22 loss: 1.6738 (0.9770) acc1: 62.5000 (75.3481) acc5: 75.0000 (92.7704) time: 0.0052 data: 0.0004 max mem: 8962\n", - "Test: [2900/6250] eta: 0:00:21 loss: 1.9858 (0.9960) acc1: 37.5000 (74.9138) acc5: 87.5000 (92.5241) time: 0.0071 data: 0.0006 max mem: 8962\n", - "Test: [3000/6250] eta: 0:00:21 loss: 2.4274 (1.0179) acc1: 50.0000 (74.6126) acc5: 75.0000 (92.1984) time: 0.0060 data: 0.0005 max mem: 8962\n", - "Test: [3100/6250] eta: 0:00:20 loss: 1.8322 (1.0420) acc1: 50.0000 (74.1172) acc5: 75.0000 (91.9461) time: 0.0064 data: 0.0005 max mem: 8962\n", - "Test: [3200/6250] eta: 0:00:19 loss: 0.7034 (1.0615) acc1: 75.0000 (73.6996) acc5: 100.0000 (91.7291) time: 0.0049 data: 0.0004 max mem: 8962\n", - "Test: [3300/6250] eta: 0:00:19 loss: 1.2400 (1.0739) acc1: 62.5000 (73.3528) acc5: 87.5000 (91.6238) time: 0.0050 data: 0.0006 max mem: 8962\n", - "Test: [3400/6250] eta: 0:00:18 loss: 2.3805 (1.0877) acc1: 37.5000 (73.0851) acc5: 75.0000 (91.4327) time: 0.0060 data: 0.0007 max mem: 8962\n", - "Test: [3500/6250] eta: 0:00:17 loss: 1.7449 (1.0943) acc1: 50.0000 (72.9899) acc5: 87.5000 (91.2953) time: 0.0074 data: 0.0023 max mem: 8962\n", - "Test: [3600/6250] eta: 0:00:16 loss: 0.4245 (1.0910) acc1: 87.5000 (73.1325) acc5: 100.0000 (91.2941) time: 0.0059 data: 0.0004 max mem: 8962\n", - "Test: [3700/6250] eta: 0:00:16 loss: 1.6126 (1.1052) acc1: 50.0000 (72.8519) acc5: 87.5000 (91.1105) time: 0.0121 data: 0.0065 max mem: 8962\n", - "Test: [3800/6250] eta: 0:00:15 loss: 0.7444 (1.1120) acc1: 87.5000 (72.7605) acc5: 87.5000 (90.9761) time: 0.0057 data: 0.0005 max mem: 8962\n", - "Test: [3900/6250] eta: 0:00:14 loss: 2.3107 (1.1276) acc1: 50.0000 (72.4494) acc5: 75.0000 (90.7652) time: 0.0052 data: 0.0005 max mem: 8962\n", - "Test: [4000/6250] eta: 0:00:14 loss: 1.4615 (1.1421) acc1: 62.5000 (72.1788) acc5: 87.5000 (90.5992) time: 0.0056 data: 0.0004 max mem: 8962\n", - "Test: [4100/6250] eta: 0:00:13 loss: 1.7447 (1.1522) acc1: 62.5000 (72.0129) acc5: 87.5000 (90.4840) time: 0.0059 data: 0.0005 max mem: 8962\n", - "Test: [4200/6250] eta: 0:00:12 loss: 0.4051 (1.1574) acc1: 87.5000 (71.8609) acc5: 100.0000 (90.4725) time: 0.0061 data: 0.0005 max mem: 8962\n", - "Test: [4300/6250] eta: 0:00:12 loss: 0.4819 (1.1662) acc1: 75.0000 (71.7420) acc5: 87.5000 (90.3191) time: 0.0058 data: 0.0007 max mem: 8962\n", - "Test: [4400/6250] eta: 0:00:11 loss: 0.8488 (1.1727) acc1: 75.0000 (71.5633) acc5: 100.0000 (90.2835) time: 0.0052 data: 0.0005 max mem: 8962\n", - "Test: [4500/6250] eta: 0:00:11 loss: 1.1410 (1.1786) acc1: 62.5000 (71.4619) acc5: 87.5000 (90.2383) time: 0.0058 data: 0.0004 max mem: 8962\n", - "Test: [4600/6250] eta: 0:00:10 loss: 1.5850 (1.1891) acc1: 50.0000 (71.2834) acc5: 87.5000 (90.0701) time: 0.0064 data: 0.0010 max mem: 8962\n", - "Test: [4700/6250] eta: 0:00:09 loss: 1.3859 (1.1988) acc1: 62.5000 (70.9982) acc5: 87.5000 (89.9463) time: 0.0051 data: 0.0005 max mem: 8962\n", - "Test: [4800/6250] eta: 0:00:09 loss: 1.2475 (1.2056) acc1: 62.5000 (70.8811) acc5: 87.5000 (89.8563) time: 0.0054 data: 0.0005 max mem: 8962\n", - "Test: [4900/6250] eta: 0:00:08 loss: 0.4116 (1.2104) acc1: 87.5000 (70.8121) acc5: 100.0000 (89.7750) time: 0.0058 data: 0.0009 max mem: 8962\n", - "Test: [5000/6250] eta: 0:00:07 loss: 2.0346 (1.2224) acc1: 62.5000 (70.6259) acc5: 75.0000 (89.6321) time: 0.0059 data: 0.0005 max mem: 8962\n", - "Test: [5100/6250] eta: 0:00:07 loss: 1.2888 (1.2271) acc1: 62.5000 (70.5327) acc5: 87.5000 (89.5952) time: 0.0069 data: 0.0006 max mem: 8962\n", - "Test: [5200/6250] eta: 0:00:06 loss: 1.0585 (1.2335) acc1: 75.0000 (70.4216) acc5: 87.5000 (89.5236) time: 0.0053 data: 0.0005 max mem: 8962\n", - "Test: [5300/6250] eta: 0:00:05 loss: 1.5391 (1.2470) acc1: 62.5000 (70.1401) acc5: 87.5000 (89.3440) time: 0.0058 data: 0.0006 max mem: 8962\n", - "Test: [5400/6250] eta: 0:00:05 loss: 0.9108 (1.2497) acc1: 75.0000 (70.0889) acc5: 87.5000 (89.2936) time: 0.0057 data: 0.0010 max mem: 8962\n", - "Test: [5500/6250] eta: 0:00:04 loss: 0.9651 (1.2539) acc1: 75.0000 (69.9668) acc5: 87.5000 (89.2383) time: 0.0060 data: 0.0006 max mem: 8962\n", - "Test: [5600/6250] eta: 0:00:04 loss: 0.8866 (1.2592) acc1: 62.5000 (69.8915) acc5: 87.5000 (89.1381) time: 0.0056 data: 0.0005 max mem: 8962\n", - "Test: [5700/6250] eta: 0:00:03 loss: 1.9771 (1.2739) acc1: 37.5000 (69.5799) acc5: 75.0000 (88.9954) time: 0.0055 data: 0.0004 max mem: 8962\n", - "Test: [5800/6250] eta: 0:00:02 loss: 0.5795 (1.2705) acc1: 75.0000 (69.6410) acc5: 100.0000 (89.0407) time: 0.0060 data: 0.0006 max mem: 8962\n", - "Test: [5900/6250] eta: 0:00:02 loss: 1.1301 (1.2675) acc1: 75.0000 (69.6831) acc5: 100.0000 (89.0824) time: 0.0058 data: 0.0005 max mem: 8962\n", - "Test: [6000/6250] eta: 0:00:01 loss: 0.5545 (1.2598) acc1: 87.5000 (69.8384) acc5: 87.5000 (89.1685) time: 0.0059 data: 0.0004 max mem: 8962\n", - "Test: [6100/6250] eta: 0:00:00 loss: 1.2558 (1.2676) acc1: 62.5000 (69.6648) acc5: 87.5000 (89.0797) time: 0.0053 data: 0.0004 max mem: 8962\n", - "Test: [6200/6250] eta: 0:00:00 loss: 0.1734 (1.2610) acc1: 87.5000 (69.7831) acc5: 100.0000 (89.1610) time: 0.0060 data: 0.0007 max mem: 8962\n", - "Test: Total time: 0:00:38\n", - "Test: Acc@1 69.816 Acc@5 89.174\n", - "Epoch: [2] [ 0/5005] eta: 5:24:29 lr: 0.0004 img/s: 488.37947021283276 loss: 0.7970 (0.7970) acc1: 72.2656 (72.2656) acc5: 87.1094 (87.1094) meanQV: 1.5059 (1.5059) stdQV: 0.3140 (0.3140) time: 3.8899 data: 3.3657 max mem: 8962\n", - "Epoch: [2] [ 100/5005] eta: 0:45:25 lr: 0.0004 img/s: 493.19209398441154 loss: 0.8641 (0.8689) acc1: 67.9688 (68.4483) acc5: 87.1094 (86.6530) meanQV: 1.4758 (1.4788) stdQV: 0.3251 (0.3251) time: 0.5211 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [ 200/5005] eta: 0:43:07 lr: 0.0004 img/s: 491.1342779344896 loss: 0.8738 (0.8655) acc1: 68.7500 (68.5595) acc5: 87.5000 (86.8898) meanQV: 1.4813 (1.4796) stdQV: 0.3240 (0.3248) time: 0.5215 data: 0.0004 max mem: 8962\n", - "Epoch: [2] [ 300/5005] eta: 0:41:47 lr: 0.0004 img/s: 487.00059052814623 loss: 0.8533 (0.8685) acc1: 68.3594 (68.4191) acc5: 87.5000 (86.8005) meanQV: 1.4785 (1.4786) stdQV: 0.3251 (0.3252) time: 0.5215 data: 0.0004 max mem: 8962\n", - "Epoch: [2] [ 400/5005] eta: 0:40:40 lr: 0.0004 img/s: 488.72224116185464 loss: 0.8569 (0.8636) acc1: 69.1406 (68.5932) acc5: 87.1094 (86.8931) meanQV: 1.4840 (1.4799) stdQV: 0.3216 (0.3247) time: 0.5220 data: 0.0004 max mem: 8962\n", - "Epoch: [2] [ 500/5005] eta: 0:39:39 lr: 0.0004 img/s: 485.98511188265843 loss: 0.8240 (0.8635) acc1: 69.5312 (68.5403) acc5: 87.8906 (86.8723) meanQV: 1.4867 (1.4795) stdQV: 0.3215 (0.3249) time: 0.5207 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [ 600/5005] eta: 0:38:41 lr: 0.0004 img/s: 493.6265143933163 loss: 0.8603 (0.8641) acc1: 68.7500 (68.5791) acc5: 86.3281 (86.8962) meanQV: 1.4801 (1.4798) stdQV: 0.3251 (0.3248) time: 0.5222 data: 0.0004 max mem: 8962\n", - "Epoch: [2] [ 700/5005] eta: 0:37:44 lr: 0.0004 img/s: 495.0733539463698 loss: 0.8482 (0.8641) acc1: 68.3594 (68.5656) acc5: 87.1094 (86.9194) meanQV: 1.4785 (1.4797) stdQV: 0.3260 (0.3249) time: 0.5209 data: 0.0004 max mem: 8962\n", - "Epoch: [2] [ 800/5005] eta: 0:36:49 lr: 0.0004 img/s: 490.3255347956828 loss: 0.8578 (0.8649) acc1: 69.1406 (68.5959) acc5: 86.7188 (86.9270) meanQV: 1.4840 (1.4799) stdQV: 0.3240 (0.3248) time: 0.5212 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [ 900/5005] eta: 0:35:55 lr: 0.0004 img/s: 487.8480383248159 loss: 0.8300 (0.8626) acc1: 69.1406 (68.6646) acc5: 87.5000 (86.9767) meanQV: 1.4840 (1.4804) stdQV: 0.3228 (0.3246) time: 0.5215 data: 0.0004 max mem: 8962\n", - "Epoch: [2] [1000/5005] eta: 0:35:01 lr: 0.0004 img/s: 491.5558526149826 loss: 0.8513 (0.8634) acc1: 69.1406 (68.6805) acc5: 87.5000 (86.9775) meanQV: 1.4840 (1.4805) stdQV: 0.3226 (0.3245) time: 0.5211 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [1100/5005] eta: 0:34:07 lr: 0.0004 img/s: 490.84757118315395 loss: 0.8577 (0.8636) acc1: 68.3594 (68.6847) acc5: 86.7188 (86.9710) meanQV: 1.4785 (1.4805) stdQV: 0.3251 (0.3245) time: 0.5209 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [1200/5005] eta: 0:33:14 lr: 0.0004 img/s: 494.1517410611354 loss: 0.8542 (0.8649) acc1: 68.7500 (68.6462) acc5: 86.3281 (86.9607) meanQV: 1.4813 (1.4802) stdQV: 0.3240 (0.3246) time: 0.5209 data: 0.0004 max mem: 8962\n", - "Epoch: [2] [1300/5005] eta: 0:32:20 lr: 0.0004 img/s: 487.5868519700021 loss: 0.8400 (0.8649) acc1: 69.1406 (68.6611) acc5: 86.3281 (86.9457) meanQV: 1.4840 (1.4803) stdQV: 0.3240 (0.3246) time: 0.5207 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [1400/5005] eta: 0:31:27 lr: 0.0004 img/s: 493.3883500186788 loss: 0.8124 (0.8646) acc1: 69.1406 (68.6887) acc5: 87.8906 (86.9658) meanQV: 1.4840 (1.4805) stdQV: 0.3240 (0.3245) time: 0.5210 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [1500/5005] eta: 0:30:34 lr: 0.0004 img/s: 495.2411855622316 loss: 0.8914 (0.8663) acc1: 67.5781 (68.6449) acc5: 87.5000 (86.9530) meanQV: 1.4719 (1.4802) stdQV: 0.3280 (0.3246) time: 0.5220 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [1600/5005] eta: 0:29:41 lr: 0.0004 img/s: 492.28401072281827 loss: 0.8554 (0.8661) acc1: 69.9219 (68.6358) acc5: 87.5000 (86.9669) meanQV: 1.4891 (1.4802) stdQV: 0.3216 (0.3246) time: 0.5215 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [1700/5005] eta: 0:28:49 lr: 0.0004 img/s: 488.3163923205836 loss: 0.8374 (0.8661) acc1: 69.5312 (68.6529) acc5: 87.5000 (86.9794) meanQV: 1.4867 (1.4803) stdQV: 0.3228 (0.3246) time: 0.5213 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [1800/5005] eta: 0:27:56 lr: 0.0004 img/s: 493.57046837953203 loss: 0.8886 (0.8664) acc1: 68.3594 (68.6639) acc5: 86.7188 (86.9808) meanQV: 1.4785 (1.4804) stdQV: 0.3251 (0.3245) time: 0.5211 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [1900/5005] eta: 0:27:03 lr: 0.0004 img/s: 490.1453552810568 loss: 0.8548 (0.8668) acc1: 68.3594 (68.6598) acc5: 87.1094 (86.9768) meanQV: 1.4785 (1.4803) stdQV: 0.3249 (0.3245) time: 0.5213 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [2000/5005] eta: 0:26:11 lr: 0.0004 img/s: 488.1270344653269 loss: 0.9081 (0.8673) acc1: 67.9688 (68.6596) acc5: 85.9375 (86.9690) meanQV: 1.4754 (1.4803) stdQV: 0.3270 (0.3245) time: 0.5214 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [2100/5005] eta: 0:25:18 lr: 0.0004 img/s: 489.5038292141757 loss: 0.8766 (0.8678) acc1: 68.7500 (68.6416) acc5: 87.5000 (86.9612) meanQV: 1.4813 (1.4802) stdQV: 0.3240 (0.3246) time: 0.5213 data: 0.0004 max mem: 8962\n", - "Epoch: [2] [2200/5005] eta: 0:24:26 lr: 0.0004 img/s: 493.80040608063206 loss: 0.8591 (0.8679) acc1: 68.3594 (68.6433) acc5: 86.7188 (86.9536) meanQV: 1.4785 (1.4802) stdQV: 0.3259 (0.3246) time: 0.5222 data: 0.0004 max mem: 8962\n", - "Epoch: [2] [2300/5005] eta: 0:23:33 lr: 0.0004 img/s: 487.89902333117954 loss: 0.8695 (0.8681) acc1: 67.5781 (68.6313) acc5: 86.3281 (86.9520) meanQV: 1.4703 (1.4801) stdQV: 0.3262 (0.3246) time: 0.5212 data: 0.0004 max mem: 8962\n", - "Epoch: [2] [2400/5005] eta: 0:22:41 lr: 0.0004 img/s: 495.53373554749567 loss: 0.9173 (0.8689) acc1: 67.5781 (68.6229) acc5: 86.3281 (86.9381) meanQV: 1.4730 (1.4801) stdQV: 0.3270 (0.3247) time: 0.5204 data: 0.0004 max mem: 8962\n", - "Epoch: [2] [2500/5005] eta: 0:21:48 lr: 0.0004 img/s: 492.0162505212319 loss: 0.8747 (0.8695) acc1: 67.9688 (68.6191) acc5: 87.1094 (86.9432) meanQV: 1.4758 (1.4800) stdQV: 0.3262 (0.3247) time: 0.5214 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [2600/5005] eta: 0:20:56 lr: 0.0004 img/s: 489.42819036679475 loss: 0.8458 (0.8694) acc1: 69.9219 (68.6381) acc5: 86.7188 (86.9568) meanQV: 1.4895 (1.4802) stdQV: 0.3216 (0.3246) time: 0.5213 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [2700/5005] eta: 0:20:04 lr: 0.0004 img/s: 491.20055774275215 loss: 0.8641 (0.8696) acc1: 67.1875 (68.6336) acc5: 86.7188 (86.9582) meanQV: 1.4703 (1.4801) stdQV: 0.3277 (0.3246) time: 0.5221 data: 0.0004 max mem: 8962\n", - "Epoch: [2] [2800/5005] eta: 0:19:11 lr: 0.0004 img/s: 494.1210418244721 loss: 0.8495 (0.8697) acc1: 69.5312 (68.6488) acc5: 87.5000 (86.9604) meanQV: 1.4867 (1.4802) stdQV: 0.3228 (0.3246) time: 0.5220 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [2900/5005] eta: 0:18:19 lr: 0.0004 img/s: 491.0172647329166 loss: 0.8790 (0.8703) acc1: 68.3594 (68.6341) acc5: 86.7188 (86.9494) meanQV: 1.4785 (1.4801) stdQV: 0.3251 (0.3246) time: 0.5208 data: 0.0004 max mem: 8962\n", - "Epoch: [2] [3000/5005] eta: 0:17:27 lr: 0.0004 img/s: 485.94684164959335 loss: 0.8755 (0.8706) acc1: 68.3594 (68.6326) acc5: 87.5000 (86.9472) meanQV: 1.4785 (1.4801) stdQV: 0.3192 (0.3246) time: 0.5216 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [3100/5005] eta: 0:16:34 lr: 0.0004 img/s: 494.1287731247124 loss: 0.9319 (0.8717) acc1: 66.4062 (68.6098) acc5: 84.7656 (86.9271) meanQV: 1.4645 (1.4800) stdQV: 0.3303 (0.3247) time: 0.5206 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [3200/5005] eta: 0:15:42 lr: 0.0004 img/s: 489.1026970888204 loss: 0.8573 (0.8715) acc1: 67.1875 (68.6226) acc5: 87.1094 (86.9358) meanQV: 1.4699 (1.4801) stdQV: 0.3291 (0.3246) time: 0.5214 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [3300/5005] eta: 0:14:50 lr: 0.0004 img/s: 493.99282848070345 loss: 0.9075 (0.8719) acc1: 68.3594 (68.6160) acc5: 85.9375 (86.9377) meanQV: 1.4770 (1.4800) stdQV: 0.3257 (0.3247) time: 0.5221 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [3400/5005] eta: 0:13:58 lr: 0.0004 img/s: 491.0915987137075 loss: 0.8952 (0.8724) acc1: 67.1875 (68.6177) acc5: 87.5000 (86.9409) meanQV: 1.4688 (1.4800) stdQV: 0.3283 (0.3247) time: 0.5217 data: 0.0004 max mem: 8962\n", - "Epoch: [2] [3500/5005] eta: 0:13:05 lr: 0.0004 img/s: 491.85172372724526 loss: 0.8616 (0.8726) acc1: 67.9688 (68.6169) acc5: 87.5000 (86.9428) meanQV: 1.4758 (1.4800) stdQV: 0.3262 (0.3247) time: 0.5212 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [3600/5005] eta: 0:12:13 lr: 0.0004 img/s: 493.1658175040636 loss: 0.8484 (0.8728) acc1: 68.7500 (68.6238) acc5: 88.2812 (86.9481) meanQV: 1.4813 (1.4801) stdQV: 0.3237 (0.3246) time: 0.5210 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [3700/5005] eta: 0:11:21 lr: 0.0004 img/s: 488.61460107584765 loss: 0.8767 (0.8732) acc1: 67.9688 (68.6150) acc5: 86.7188 (86.9436) meanQV: 1.4758 (1.4800) stdQV: 0.3240 (0.3247) time: 0.5216 data: 0.0004 max mem: 8962\n", - "Epoch: [2] [3800/5005] eta: 0:10:29 lr: 0.0004 img/s: 494.11217386622536 loss: 0.8734 (0.8735) acc1: 66.7969 (68.6138) acc5: 86.3281 (86.9486) meanQV: 1.4676 (1.4800) stdQV: 0.3251 (0.3247) time: 0.5209 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [3900/5005] eta: 0:09:36 lr: 0.0004 img/s: 490.99728148911294 loss: 0.8477 (0.8733) acc1: 69.1406 (68.6243) acc5: 87.1094 (86.9573) meanQV: 1.4840 (1.4801) stdQV: 0.3240 (0.3246) time: 0.5215 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [4000/5005] eta: 0:08:44 lr: 0.0004 img/s: 491.4019783456549 loss: 0.8635 (0.8735) acc1: 68.7500 (68.6289) acc5: 87.5000 (86.9596) meanQV: 1.4813 (1.4801) stdQV: 0.3240 (0.3246) time: 0.5211 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [4100/5005] eta: 0:07:52 lr: 0.0004 img/s: 494.1733465144578 loss: 0.8801 (0.8738) acc1: 68.3594 (68.6256) acc5: 85.9375 (86.9581) meanQV: 1.4773 (1.4801) stdQV: 0.3228 (0.3246) time: 0.5216 data: 0.0004 max mem: 8962\n", - "Epoch: [2] [4200/5005] eta: 0:07:00 lr: 0.0004 img/s: 489.6234710977453 loss: 0.8273 (0.8740) acc1: 70.3125 (68.6327) acc5: 87.8906 (86.9577) meanQV: 1.4922 (1.4801) stdQV: 0.3190 (0.3246) time: 0.5215 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [4300/5005] eta: 0:06:08 lr: 0.0004 img/s: 494.2581946216013 loss: 0.8741 (0.8740) acc1: 68.7500 (68.6423) acc5: 87.5000 (86.9612) meanQV: 1.4813 (1.4802) stdQV: 0.3251 (0.3246) time: 0.5204 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [4400/5005] eta: 0:05:15 lr: 0.0004 img/s: 745.4293803117672 loss: 0.8636 (0.8744) acc1: 68.7500 (68.6341) acc5: 87.1094 (86.9601) meanQV: 1.4813 (1.4801) stdQV: 0.3240 (0.3246) time: 0.4351 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [4500/5005] eta: 0:04:21 lr: 0.0004 img/s: 486.88907097457 loss: 0.8677 (0.8746) acc1: 67.9688 (68.6392) acc5: 86.3281 (86.9591) meanQV: 1.4758 (1.4802) stdQV: 0.3262 (0.3246) time: 0.3115 data: 0.0013 max mem: 8962\n", - "Epoch: [2] [4600/5005] eta: 0:03:29 lr: 0.0004 img/s: 489.810640387527 loss: 0.8474 (0.8748) acc1: 68.7500 (68.6407) acc5: 86.7188 (86.9571) meanQV: 1.4813 (1.4802) stdQV: 0.3240 (0.3246) time: 0.5214 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [4700/5005] eta: 0:02:38 lr: 0.0004 img/s: 494.03101273102885 loss: 0.8843 (0.8749) acc1: 67.1875 (68.6427) acc5: 85.9375 (86.9571) meanQV: 1.4691 (1.4802) stdQV: 0.3270 (0.3246) time: 0.5212 data: 0.0004 max mem: 8962\n", - "Epoch: [2] [4800/5005] eta: 0:01:46 lr: 0.0004 img/s: 491.7528352428831 loss: 0.9494 (0.8752) acc1: 68.3594 (68.6450) acc5: 85.9375 (86.9551) meanQV: 1.4785 (1.4802) stdQV: 0.3251 (0.3246) time: 0.5206 data: 0.0003 max mem: 8962\n", - "Epoch: [2] [4900/5005] eta: 0:00:54 lr: 0.0004 img/s: 489.231282062264 loss: 0.8820 (0.8755) acc1: 67.1875 (68.6437) acc5: 85.9375 (86.9490) meanQV: 1.4691 (1.4802) stdQV: 0.3290 (0.3246) time: 0.5212 data: 0.0004 max mem: 8962\n", - "Epoch: [2] [5000/5005] eta: 0:00:02 lr: 0.0004 img/s: 489.3046302811125 loss: 0.8896 (0.8759) acc1: 68.3594 (68.6439) acc5: 86.7188 (86.9525) meanQV: 1.4785 (1.4802) stdQV: 0.3248 (0.3246) time: 0.5207 data: 0.0002 max mem: 8962\n", - "Epoch: [2] Total time: 0:43:15\n", - "Test: [ 0/6250] eta: 1:13:07 loss: 0.9734 (0.9734) acc1: 62.5000 (62.5000) acc5: 100.0000 (100.0000) time: 0.7020 data: 0.6951 max mem: 8962\n", - "Test: [ 100/6250] eta: 0:01:26 loss: 0.1843 (0.6218) acc1: 87.5000 (83.2921) acc5: 100.0000 (95.2970) time: 0.0060 data: 0.0006 max mem: 8962\n", - "Test: [ 200/6250] eta: 0:01:02 loss: 0.7251 (0.6534) acc1: 87.5000 (84.0174) acc5: 87.5000 (94.8383) time: 0.0060 data: 0.0005 max mem: 8962\n", - "Test: [ 300/6250] eta: 0:00:53 loss: 1.2929 (0.8644) acc1: 62.5000 (78.9037) acc5: 87.5000 (93.1894) time: 0.0062 data: 0.0005 max mem: 8962\n", - "Test: [ 400/6250] eta: 0:00:48 loss: 0.8992 (0.9909) acc1: 75.0000 (76.0910) acc5: 87.5000 (92.3005) time: 0.0060 data: 0.0004 max mem: 8962\n", - "Test: [ 500/6250] eta: 0:00:45 loss: 1.1745 (1.0423) acc1: 75.0000 (74.8503) acc5: 87.5000 (91.9661) time: 0.0065 data: 0.0010 max mem: 8962\n", - "Test: [ 600/6250] eta: 0:00:42 loss: 0.4571 (0.9501) acc1: 87.5000 (76.8095) acc5: 100.0000 (92.5957) time: 0.0061 data: 0.0005 max mem: 8962\n", - "Test: [ 700/6250] eta: 0:00:40 loss: 0.4039 (0.9331) acc1: 87.5000 (77.1220) acc5: 100.0000 (92.5820) time: 0.0057 data: 0.0006 max mem: 8962\n", - "Test: [ 800/6250] eta: 0:00:39 loss: 0.7346 (0.9629) acc1: 75.0000 (76.5605) acc5: 87.5000 (92.1660) time: 0.0058 data: 0.0006 max mem: 8962\n", - "Test: [ 900/6250] eta: 0:00:37 loss: 0.3460 (0.9073) acc1: 87.5000 (77.8857) acc5: 100.0000 (92.6609) time: 0.0059 data: 0.0005 max mem: 8962\n", - "Test: [1000/6250] eta: 0:00:36 loss: 0.8425 (0.8922) acc1: 75.0000 (78.1094) acc5: 100.0000 (92.8322) time: 0.0057 data: 0.0008 max mem: 8962\n", - "Test: [1100/6250] eta: 0:00:35 loss: 0.9796 (0.9278) acc1: 75.0000 (77.2252) acc5: 100.0000 (92.6544) time: 0.0073 data: 0.0016 max mem: 8962\n", - "Test: [1200/6250] eta: 0:00:34 loss: 1.1206 (0.9326) acc1: 75.0000 (76.8110) acc5: 87.5000 (92.7560) time: 0.0060 data: 0.0011 max mem: 8962\n", - "Test: [1300/6250] eta: 0:00:33 loss: 0.4536 (0.9371) acc1: 87.5000 (76.5661) acc5: 100.0000 (92.8997) time: 0.0059 data: 0.0005 max mem: 8962\n", - "Test: [1400/6250] eta: 0:00:32 loss: 0.7222 (0.9336) acc1: 75.0000 (76.6149) acc5: 87.5000 (92.9425) time: 0.0059 data: 0.0006 max mem: 8962\n", - "Test: [1500/6250] eta: 0:00:31 loss: 0.9426 (0.9376) acc1: 62.5000 (76.3658) acc5: 100.0000 (93.0213) time: 0.0057 data: 0.0005 max mem: 8962\n", - "Test: [1600/6250] eta: 0:00:30 loss: 0.1496 (0.9321) acc1: 100.0000 (76.2336) acc5: 100.0000 (93.1839) time: 0.0053 data: 0.0005 max mem: 8962\n", - "Test: [1700/6250] eta: 0:00:29 loss: 1.0775 (0.9295) acc1: 75.0000 (76.0802) acc5: 100.0000 (93.3128) time: 0.0051 data: 0.0006 max mem: 8962\n", - "Test: [1800/6250] eta: 0:00:29 loss: 1.1900 (0.9398) acc1: 62.5000 (75.8537) acc5: 87.5000 (93.2954) time: 0.0047 data: 0.0004 max mem: 8962\n", - "Test: [1900/6250] eta: 0:00:28 loss: 0.9377 (0.9324) acc1: 75.0000 (76.1376) acc5: 100.0000 (93.4179) time: 0.0056 data: 0.0005 max mem: 8962\n", - "Test: [2000/6250] eta: 0:00:27 loss: 0.4859 (0.9373) acc1: 87.5000 (76.0557) acc5: 100.0000 (93.4158) time: 0.0057 data: 0.0005 max mem: 8962\n", - "Test: [2100/6250] eta: 0:00:27 loss: 0.3588 (0.9185) acc1: 87.5000 (76.5826) acc5: 100.0000 (93.5745) time: 0.0059 data: 0.0004 max mem: 8962\n", - "Test: [2200/6250] eta: 0:00:26 loss: 0.1941 (0.9123) acc1: 87.5000 (76.7208) acc5: 100.0000 (93.6336) time: 0.0066 data: 0.0011 max mem: 8962\n", - "Test: [2300/6250] eta: 0:00:25 loss: 0.6498 (0.9137) acc1: 87.5000 (76.6949) acc5: 100.0000 (93.5952) time: 0.0052 data: 0.0008 max mem: 8962\n", - "Test: [2400/6250] eta: 0:00:25 loss: 0.9289 (0.9226) acc1: 75.0000 (76.5723) acc5: 87.5000 (93.5079) time: 0.0067 data: 0.0004 max mem: 8962\n", - "Test: [2500/6250] eta: 0:00:24 loss: 0.9120 (0.9237) acc1: 87.5000 (76.6593) acc5: 100.0000 (93.4726) time: 0.0054 data: 0.0004 max mem: 8962\n", - "Test: [2600/6250] eta: 0:00:23 loss: 2.2991 (0.9477) acc1: 50.0000 (76.1774) acc5: 62.5000 (93.1565) time: 0.0057 data: 0.0005 max mem: 8962\n", - "Test: [2700/6250] eta: 0:00:23 loss: 0.7023 (0.9584) acc1: 75.0000 (76.0089) acc5: 100.0000 (93.0442) time: 0.0061 data: 0.0007 max mem: 8962\n", - "Test: [2800/6250] eta: 0:00:22 loss: 1.6450 (0.9808) acc1: 62.5000 (75.5400) acc5: 87.5000 (92.7749) time: 0.0056 data: 0.0007 max mem: 8962\n", - "Test: [2900/6250] eta: 0:00:21 loss: 1.8880 (1.0003) acc1: 50.0000 (75.1034) acc5: 87.5000 (92.5414) time: 0.0057 data: 0.0005 max mem: 8962\n", - "Test: [3000/6250] eta: 0:00:21 loss: 2.2391 (1.0214) acc1: 50.0000 (74.7959) acc5: 75.0000 (92.2193) time: 0.0051 data: 0.0004 max mem: 8962\n", - "Test: [3100/6250] eta: 0:00:20 loss: 1.8099 (1.0473) acc1: 50.0000 (74.2785) acc5: 75.0000 (91.9502) time: 0.0058 data: 0.0006 max mem: 8962\n", - "Test: [3200/6250] eta: 0:00:19 loss: 0.7831 (1.0671) acc1: 62.5000 (73.8558) acc5: 100.0000 (91.7565) time: 0.0058 data: 0.0009 max mem: 8962\n", - "Test: [3300/6250] eta: 0:00:18 loss: 1.2720 (1.0789) acc1: 62.5000 (73.5497) acc5: 87.5000 (91.6540) time: 0.0060 data: 0.0010 max mem: 8962\n", - "Test: [3400/6250] eta: 0:00:18 loss: 2.3960 (1.0930) acc1: 37.5000 (73.2946) acc5: 75.0000 (91.4363) time: 0.0070 data: 0.0010 max mem: 8962\n", - "Test: [3500/6250] eta: 0:00:17 loss: 1.8152 (1.0998) acc1: 50.0000 (73.1612) acc5: 75.0000 (91.3203) time: 0.0075 data: 0.0021 max mem: 8962\n", - "Test: [3600/6250] eta: 0:00:17 loss: 0.4449 (1.0965) acc1: 87.5000 (73.3026) acc5: 100.0000 (91.3219) time: 0.0058 data: 0.0004 max mem: 8962\n", - "Test: [3700/6250] eta: 0:00:16 loss: 1.6940 (1.1099) acc1: 62.5000 (73.0276) acc5: 87.5000 (91.1477) time: 0.0131 data: 0.0075 max mem: 8962\n", - "Test: [3800/6250] eta: 0:00:15 loss: 0.7985 (1.1155) acc1: 87.5000 (72.9644) acc5: 87.5000 (91.0484) time: 0.0054 data: 0.0006 max mem: 8962\n", - "Test: [3900/6250] eta: 0:00:15 loss: 2.2583 (1.1305) acc1: 50.0000 (72.6833) acc5: 75.0000 (90.8293) time: 0.0053 data: 0.0004 max mem: 8962\n", - "Test: [4000/6250] eta: 0:00:14 loss: 1.4313 (1.1455) acc1: 62.5000 (72.4256) acc5: 87.5000 (90.6492) time: 0.0056 data: 0.0005 max mem: 8962\n", - "Test: [4100/6250] eta: 0:00:13 loss: 1.7358 (1.1560) acc1: 62.5000 (72.2507) acc5: 87.5000 (90.5389) time: 0.0056 data: 0.0004 max mem: 8962\n", - "Test: [4200/6250] eta: 0:00:13 loss: 0.4747 (1.1604) acc1: 87.5000 (72.1168) acc5: 100.0000 (90.5380) time: 0.0064 data: 0.0013 max mem: 8962\n", - "Test: [4300/6250] eta: 0:00:12 loss: 0.5343 (1.1704) acc1: 75.0000 (71.9513) acc5: 87.5000 (90.3743) time: 0.0054 data: 0.0004 max mem: 8962\n", - "Test: [4400/6250] eta: 0:00:11 loss: 0.8444 (1.1773) acc1: 75.0000 (71.7422) acc5: 100.0000 (90.3175) time: 0.0050 data: 0.0003 max mem: 8962\n", - "Test: [4500/6250] eta: 0:00:11 loss: 1.2144 (1.1832) acc1: 62.5000 (71.6119) acc5: 87.5000 (90.2744) time: 0.0060 data: 0.0006 max mem: 8962\n", - "Test: [4600/6250] eta: 0:00:10 loss: 1.7535 (1.1929) acc1: 50.0000 (71.4600) acc5: 87.5000 (90.1190) time: 0.0061 data: 0.0005 max mem: 8962\n", - "Test: [4700/6250] eta: 0:00:09 loss: 1.4817 (1.2036) acc1: 62.5000 (71.1498) acc5: 87.5000 (89.9702) time: 0.0050 data: 0.0006 max mem: 8962\n", - "Test: [4800/6250] eta: 0:00:09 loss: 1.3074 (1.2111) acc1: 62.5000 (71.0034) acc5: 87.5000 (89.8563) time: 0.0054 data: 0.0005 max mem: 8962\n", - "Test: [4900/6250] eta: 0:00:08 loss: 0.4546 (1.2157) acc1: 75.0000 (70.9090) acc5: 100.0000 (89.7852) time: 0.0053 data: 0.0006 max mem: 8962\n", - "Test: [5000/6250] eta: 0:00:07 loss: 2.0130 (1.2274) acc1: 62.5000 (70.7359) acc5: 75.0000 (89.6296) time: 0.0059 data: 0.0004 max mem: 8962\n", - "Test: [5100/6250] eta: 0:00:07 loss: 1.3178 (1.2327) acc1: 62.5000 (70.6136) acc5: 87.5000 (89.5903) time: 0.0061 data: 0.0005 max mem: 8962\n", - "Test: [5200/6250] eta: 0:00:06 loss: 1.0106 (1.2388) acc1: 75.0000 (70.5177) acc5: 87.5000 (89.5236) time: 0.0052 data: 0.0004 max mem: 8962\n", - "Test: [5300/6250] eta: 0:00:05 loss: 1.5488 (1.2526) acc1: 62.5000 (70.2297) acc5: 87.5000 (89.3298) time: 0.0053 data: 0.0005 max mem: 8962\n", - "Test: [5400/6250] eta: 0:00:05 loss: 0.8032 (1.2557) acc1: 75.0000 (70.1768) acc5: 87.5000 (89.2728) time: 0.0058 data: 0.0004 max mem: 8962\n", - "Test: [5500/6250] eta: 0:00:04 loss: 0.8721 (1.2595) acc1: 75.0000 (70.0827) acc5: 87.5000 (89.2201) time: 0.0061 data: 0.0006 max mem: 8962\n", - "Test: [5600/6250] eta: 0:00:04 loss: 1.0362 (1.2646) acc1: 62.5000 (70.0009) acc5: 87.5000 (89.1269) time: 0.0053 data: 0.0004 max mem: 8962\n", - "Test: [5700/6250] eta: 0:00:03 loss: 2.2192 (1.2800) acc1: 37.5000 (69.7049) acc5: 75.0000 (88.9756) time: 0.0052 data: 0.0005 max mem: 8962\n", - "Test: [5800/6250] eta: 0:00:02 loss: 0.5529 (1.2755) acc1: 75.0000 (69.7897) acc5: 100.0000 (89.0321) time: 0.0052 data: 0.0005 max mem: 8962\n", - "Test: [5900/6250] eta: 0:00:02 loss: 1.1504 (1.2730) acc1: 75.0000 (69.8187) acc5: 87.5000 (89.0612) time: 0.0059 data: 0.0006 max mem: 8962\n", - "Test: [6000/6250] eta: 0:00:01 loss: 0.5121 (1.2652) acc1: 87.5000 (69.9779) acc5: 87.5000 (89.1518) time: 0.0060 data: 0.0005 max mem: 8962\n", - "Test: [6100/6250] eta: 0:00:00 loss: 1.3071 (1.2734) acc1: 62.5000 (69.7959) acc5: 87.5000 (89.0407) time: 0.0048 data: 0.0004 max mem: 8962\n", - "Test: [6200/6250] eta: 0:00:00 loss: 0.2227 (1.2671) acc1: 100.0000 (69.9161) acc5: 100.0000 (89.1207) time: 0.0062 data: 0.0005 max mem: 8962\n", - "Test: Total time: 0:00:38\n", - "Test: Acc@1 69.952 Acc@5 89.136\n", - "Epoch: [3] [ 0/5005] eta: 5:49:16 lr: 0.0002 img/s: 494.13332106139956 loss: 0.8680 (0.8680) acc1: 68.3594 (68.3594) acc5: 87.8906 (87.8906) meanQV: 1.4785 (1.4785) stdQV: 0.3262 (0.3262) time: 4.1871 data: 3.6690 max mem: 8962\n", - "Epoch: [3] [ 100/5005] eta: 0:45:34 lr: 0.0002 img/s: 493.44095322259034 loss: 0.8700 (0.8771) acc1: 69.5312 (68.9124) acc5: 87.1094 (87.1906) meanQV: 1.4867 (1.4820) stdQV: 0.3216 (0.3238) time: 0.5208 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [ 200/5005] eta: 0:43:11 lr: 0.0002 img/s: 491.6118921765009 loss: 0.8557 (0.8772) acc1: 69.1406 (69.1056) acc5: 87.1094 (87.0666) meanQV: 1.4840 (1.4834) stdQV: 0.3228 (0.3233) time: 0.5215 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [ 300/5005] eta: 0:41:50 lr: 0.0002 img/s: 488.24378408069873 loss: 0.8735 (0.8765) acc1: 68.3594 (69.1173) acc5: 87.1094 (87.2859) meanQV: 1.4773 (1.4834) stdQV: 0.3259 (0.3233) time: 0.5215 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [ 400/5005] eta: 0:40:42 lr: 0.0002 img/s: 491.1491050840666 loss: 0.8561 (0.8749) acc1: 69.1406 (69.1338) acc5: 87.1094 (87.3091) meanQV: 1.4816 (1.4835) stdQV: 0.3235 (0.3232) time: 0.5215 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [ 500/5005] eta: 0:39:40 lr: 0.0002 img/s: 494.98206480200105 loss: 0.8668 (0.8747) acc1: 68.3594 (69.2046) acc5: 87.5000 (87.3082) meanQV: 1.4785 (1.4840) stdQV: 0.3228 (0.3230) time: 0.5210 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [ 600/5005] eta: 0:38:42 lr: 0.0002 img/s: 492.9600064641953 loss: 0.8776 (0.8775) acc1: 69.1406 (69.1595) acc5: 87.5000 (87.2979) meanQV: 1.4840 (1.4837) stdQV: 0.3226 (0.3231) time: 0.5206 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [ 700/5005] eta: 0:37:46 lr: 0.0002 img/s: 488.8984818192125 loss: 0.8980 (0.8788) acc1: 69.1406 (69.1512) acc5: 86.3281 (87.2465) meanQV: 1.4836 (1.4837) stdQV: 0.3228 (0.3231) time: 0.5213 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [ 800/5005] eta: 0:36:50 lr: 0.0002 img/s: 489.088661574488 loss: 0.8472 (0.8815) acc1: 68.7500 (69.0777) acc5: 86.7188 (87.1776) meanQV: 1.4801 (1.4831) stdQV: 0.3248 (0.3233) time: 0.5208 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [ 900/5005] eta: 0:35:56 lr: 0.0002 img/s: 491.4127733969181 loss: 0.8806 (0.8807) acc1: 68.3594 (69.0921) acc5: 86.7188 (87.1657) meanQV: 1.4762 (1.4832) stdQV: 0.3262 (0.3233) time: 0.5213 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [1000/5005] eta: 0:35:01 lr: 0.0002 img/s: 491.22662521193246 loss: 0.9323 (0.8815) acc1: 67.9688 (69.0762) acc5: 86.7188 (87.1488) meanQV: 1.4758 (1.4831) stdQV: 0.3273 (0.3233) time: 0.5216 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [1100/5005] eta: 0:34:08 lr: 0.0002 img/s: 490.98717822090043 loss: 0.8727 (0.8828) acc1: 69.1406 (69.0541) acc5: 87.5000 (87.1292) meanQV: 1.4828 (1.4830) stdQV: 0.3237 (0.3234) time: 0.5207 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [1200/5005] eta: 0:33:14 lr: 0.0002 img/s: 489.31912417248043 loss: 0.8732 (0.8839) acc1: 68.7500 (69.0385) acc5: 86.3281 (87.0990) meanQV: 1.4801 (1.4829) stdQV: 0.3248 (0.3234) time: 0.5208 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [1300/5005] eta: 0:32:21 lr: 0.0002 img/s: 488.60570732215643 loss: 0.8841 (0.8833) acc1: 69.1406 (69.0139) acc5: 87.1094 (87.1034) meanQV: 1.4840 (1.4827) stdQV: 0.3228 (0.3235) time: 0.5211 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [1400/5005] eta: 0:31:27 lr: 0.0002 img/s: 489.7130175490766 loss: 0.8099 (0.8833) acc1: 69.9219 (69.0252) acc5: 87.1094 (87.0971) meanQV: 1.4895 (1.4828) stdQV: 0.3188 (0.3235) time: 0.5214 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [1500/5005] eta: 0:30:35 lr: 0.0002 img/s: 490.8552003847328 loss: 0.8780 (0.8834) acc1: 69.1406 (69.0147) acc5: 86.7188 (87.0748) meanQV: 1.4840 (1.4827) stdQV: 0.3228 (0.3235) time: 0.5206 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [1600/5005] eta: 0:29:42 lr: 0.0002 img/s: 492.7211280428486 loss: 0.9144 (0.8839) acc1: 69.1406 (69.0086) acc5: 86.7188 (87.0847) meanQV: 1.4828 (1.4827) stdQV: 0.3237 (0.3235) time: 0.5212 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [1700/5005] eta: 0:28:49 lr: 0.0002 img/s: 494.0639741221239 loss: 0.8737 (0.8834) acc1: 69.5312 (69.0187) acc5: 86.3281 (87.0809) meanQV: 1.4867 (1.4827) stdQV: 0.3228 (0.3235) time: 0.5216 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [1800/5005] eta: 0:27:56 lr: 0.0002 img/s: 494.14514608859565 loss: 0.9103 (0.8836) acc1: 67.5781 (69.0094) acc5: 87.1094 (87.0679) meanQV: 1.4730 (1.4827) stdQV: 0.3283 (0.3235) time: 0.5212 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [1900/5005] eta: 0:27:04 lr: 0.0002 img/s: 488.1301411462633 loss: 0.8608 (0.8841) acc1: 69.1406 (68.9943) acc5: 86.7188 (87.0798) meanQV: 1.4840 (1.4826) stdQV: 0.3228 (0.3236) time: 0.5210 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [2000/5005] eta: 0:26:11 lr: 0.0002 img/s: 493.9826015587656 loss: 0.9005 (0.8840) acc1: 69.1406 (69.0001) acc5: 87.1094 (87.0899) meanQV: 1.4840 (1.4826) stdQV: 0.3216 (0.3236) time: 0.5209 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [2100/5005] eta: 0:25:18 lr: 0.0002 img/s: 491.0731815518808 loss: 0.8610 (0.8835) acc1: 70.7031 (69.0287) acc5: 87.8906 (87.0908) meanQV: 1.4949 (1.4828) stdQV: 0.3180 (0.3235) time: 0.5210 data: 0.0005 max mem: 8962\n", - "Epoch: [3] [2200/5005] eta: 0:24:26 lr: 0.0002 img/s: 486.65824433670525 loss: 0.8879 (0.8840) acc1: 67.1875 (69.0139) acc5: 87.1094 (87.0943) meanQV: 1.4703 (1.4827) stdQV: 0.3262 (0.3235) time: 0.5212 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [2300/5005] eta: 0:23:33 lr: 0.0002 img/s: 489.28055002159914 loss: 0.8677 (0.8844) acc1: 69.1406 (69.0091) acc5: 88.6719 (87.0997) meanQV: 1.4840 (1.4826) stdQV: 0.3192 (0.3235) time: 0.5220 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [2400/5005] eta: 0:22:41 lr: 0.0002 img/s: 489.7322262906224 loss: 0.8478 (0.8846) acc1: 69.5312 (68.9997) acc5: 88.2812 (87.0944) meanQV: 1.4867 (1.4826) stdQV: 0.3226 (0.3236) time: 0.5206 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [2500/5005] eta: 0:21:49 lr: 0.0002 img/s: 494.49583330838766 loss: 0.9224 (0.8849) acc1: 68.7500 (68.9983) acc5: 86.3281 (87.0841) meanQV: 1.4813 (1.4826) stdQV: 0.3240 (0.3236) time: 0.5211 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [2600/5005] eta: 0:20:56 lr: 0.0002 img/s: 491.60829085668 loss: 0.8767 (0.8851) acc1: 69.9219 (69.0107) acc5: 87.8906 (87.0868) meanQV: 1.4883 (1.4826) stdQV: 0.3214 (0.3235) time: 0.5205 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [2700/5005] eta: 0:20:04 lr: 0.0002 img/s: 494.5233905012099 loss: 0.8684 (0.8851) acc1: 69.5312 (69.0087) acc5: 87.5000 (87.0942) meanQV: 1.4855 (1.4826) stdQV: 0.3216 (0.3235) time: 0.5203 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [2800/5005] eta: 0:19:11 lr: 0.0002 img/s: 491.0152438815262 loss: 0.9061 (0.8855) acc1: 69.1406 (68.9994) acc5: 86.7188 (87.0876) meanQV: 1.4840 (1.4826) stdQV: 0.3228 (0.3236) time: 0.5209 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [2900/5005] eta: 0:18:19 lr: 0.0002 img/s: 490.7789190394875 loss: 0.8594 (0.8853) acc1: 69.1406 (68.9987) acc5: 87.5000 (87.0913) meanQV: 1.4820 (1.4826) stdQV: 0.3215 (0.3236) time: 0.5220 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [3000/5005] eta: 0:17:27 lr: 0.0002 img/s: 490.72441483952787 loss: 0.8432 (0.8854) acc1: 69.5312 (68.9989) acc5: 87.8906 (87.0884) meanQV: 1.4867 (1.4826) stdQV: 0.3202 (0.3236) time: 0.5214 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [3100/5005] eta: 0:16:34 lr: 0.0002 img/s: 491.1839299405131 loss: 0.8755 (0.8852) acc1: 70.3125 (69.0101) acc5: 87.5000 (87.0990) meanQV: 1.4898 (1.4826) stdQV: 0.3192 (0.3235) time: 0.5210 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [3200/5005] eta: 0:15:42 lr: 0.0002 img/s: 491.87718437506726 loss: 0.8693 (0.8855) acc1: 69.5312 (69.0059) acc5: 86.3281 (87.0879) meanQV: 1.4863 (1.4826) stdQV: 0.3216 (0.3235) time: 0.5207 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [3300/5005] eta: 0:14:50 lr: 0.0002 img/s: 489.23663194863695 loss: 0.8853 (0.8859) acc1: 69.9219 (69.0043) acc5: 85.9375 (87.0807) meanQV: 1.4871 (1.4826) stdQV: 0.3202 (0.3235) time: 0.5208 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [3400/5005] eta: 0:13:58 lr: 0.0002 img/s: 487.7279331806508 loss: 0.8678 (0.8862) acc1: 68.3594 (68.9936) acc5: 86.3281 (87.0718) meanQV: 1.4785 (1.4825) stdQV: 0.3262 (0.3236) time: 0.5219 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [3500/5005] eta: 0:13:05 lr: 0.0002 img/s: 492.66031437817423 loss: 0.8921 (0.8862) acc1: 68.3594 (68.9935) acc5: 86.3281 (87.0740) meanQV: 1.4781 (1.4825) stdQV: 0.3260 (0.3236) time: 0.5205 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [3600/5005] eta: 0:12:13 lr: 0.0002 img/s: 491.45280918624405 loss: 0.8688 (0.8865) acc1: 69.5312 (68.9917) acc5: 87.1094 (87.0669) meanQV: 1.4855 (1.4825) stdQV: 0.3226 (0.3236) time: 0.5213 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [3700/5005] eta: 0:11:21 lr: 0.0002 img/s: 491.6204455225456 loss: 0.8652 (0.8867) acc1: 68.3594 (68.9841) acc5: 86.3281 (87.0616) meanQV: 1.4773 (1.4824) stdQV: 0.3262 (0.3236) time: 0.5209 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [3800/5005] eta: 0:10:29 lr: 0.0002 img/s: 487.71309034173106 loss: 0.8378 (0.8872) acc1: 69.1406 (68.9696) acc5: 86.7188 (87.0521) meanQV: 1.4840 (1.4823) stdQV: 0.3228 (0.3236) time: 0.5206 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [3900/5005] eta: 0:09:36 lr: 0.0002 img/s: 493.498784340653 loss: 0.9257 (0.8877) acc1: 68.3594 (68.9631) acc5: 87.1094 (87.0480) meanQV: 1.4785 (1.4823) stdQV: 0.3240 (0.3237) time: 0.5218 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [4000/5005] eta: 0:08:44 lr: 0.0002 img/s: 494.29778520681486 loss: 0.8728 (0.8877) acc1: 68.3594 (68.9672) acc5: 87.5000 (87.0530) meanQV: 1.4785 (1.4823) stdQV: 0.3240 (0.3236) time: 0.5206 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [4100/5005] eta: 0:07:52 lr: 0.0002 img/s: 491.56395393020676 loss: 0.9223 (0.8879) acc1: 68.7500 (68.9693) acc5: 86.7188 (87.0528) meanQV: 1.4789 (1.4823) stdQV: 0.3249 (0.3236) time: 0.5214 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [4200/5005] eta: 0:07:00 lr: 0.0002 img/s: 489.26494371188943 loss: 0.8813 (0.8883) acc1: 67.5781 (68.9555) acc5: 86.3281 (87.0487) meanQV: 1.4730 (1.4822) stdQV: 0.3273 (0.3237) time: 0.5211 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [4300/5005] eta: 0:06:07 lr: 0.0002 img/s: 493.6955115781962 loss: 0.8671 (0.8883) acc1: 70.7031 (68.9589) acc5: 86.3281 (87.0572) meanQV: 1.4949 (1.4823) stdQV: 0.3192 (0.3237) time: 0.5204 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [4400/5005] eta: 0:05:15 lr: 0.0002 img/s: 493.47927899772964 loss: 0.9207 (0.8885) acc1: 68.3594 (68.9580) acc5: 85.9375 (87.0487) meanQV: 1.4785 (1.4823) stdQV: 0.3262 (0.3237) time: 0.5198 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [4500/5005] eta: 0:04:23 lr: 0.0002 img/s: 492.8457412066241 loss: 0.8825 (0.8885) acc1: 69.5312 (68.9623) acc5: 87.1094 (87.0524) meanQV: 1.4867 (1.4823) stdQV: 0.3228 (0.3237) time: 0.5207 data: 0.0004 max mem: 8962\n", - "Epoch: [3] [4600/5005] eta: 0:03:31 lr: 0.0002 img/s: 491.5558526149826 loss: 0.8631 (0.8888) acc1: 68.7500 (68.9587) acc5: 86.7188 (87.0511) meanQV: 1.4813 (1.4823) stdQV: 0.3240 (0.3237) time: 0.5209 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [4700/5005] eta: 0:02:39 lr: 0.0002 img/s: 491.1044017010751 loss: 0.9060 (0.8890) acc1: 67.9688 (68.9607) acc5: 86.7188 (87.0516) meanQV: 1.4754 (1.4823) stdQV: 0.3262 (0.3237) time: 0.5205 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [4800/5005] eta: 0:01:46 lr: 0.0002 img/s: 488.3679195209023 loss: 0.8826 (0.8891) acc1: 70.3125 (68.9605) acc5: 87.5000 (87.0509) meanQV: 1.4922 (1.4823) stdQV: 0.3204 (0.3237) time: 0.5207 data: 0.0003 max mem: 8962\n", - "Epoch: [3] [4900/5005] eta: 0:00:54 lr: 0.0002 img/s: 657.254661723217 loss: 0.8709 (0.8890) acc1: 69.1406 (68.9624) acc5: 87.1094 (87.0525) meanQV: 1.4840 (1.4823) stdQV: 0.3208 (0.3237) time: 0.3795 data: 0.0027 max mem: 8962\n", - "Epoch: [3] [5000/5005] eta: 0:00:02 lr: 0.0002 img/s: 489.617889485732 loss: 0.8860 (0.8894) acc1: 67.5781 (68.9478) acc5: 86.3281 (87.0502) meanQV: 1.4730 (1.4822) stdQV: 0.3259 (0.3237) time: 0.5212 data: 0.0002 max mem: 8962\n", - "Epoch: [3] Total time: 0:43:15\n", - "Test: [ 0/6250] eta: 1:13:37 loss: 0.8178 (0.8178) acc1: 62.5000 (62.5000) acc5: 100.0000 (100.0000) time: 0.7068 data: 0.7000 max mem: 8962\n", - "Test: [ 100/6250] eta: 0:01:28 loss: 0.1440 (0.6006) acc1: 87.5000 (83.4158) acc5: 100.0000 (95.4208) time: 0.0070 data: 0.0009 max mem: 8962\n", - "Test: [ 200/6250] eta: 0:01:06 loss: 0.6774 (0.6502) acc1: 87.5000 (84.2040) acc5: 100.0000 (94.8383) time: 0.0073 data: 0.0006 max mem: 8962\n", - "Test: [ 300/6250] eta: 0:00:56 loss: 1.2668 (0.8598) acc1: 62.5000 (79.2774) acc5: 87.5000 (93.1478) time: 0.0066 data: 0.0007 max mem: 8962\n", - "Test: [ 400/6250] eta: 0:00:51 loss: 0.8958 (0.9858) acc1: 75.0000 (76.1845) acc5: 100.0000 (92.2693) time: 0.0062 data: 0.0008 max mem: 8962\n", - "Test: [ 500/6250] eta: 0:00:48 loss: 1.1632 (1.0374) acc1: 75.0000 (74.9002) acc5: 87.5000 (91.9411) time: 0.0061 data: 0.0007 max mem: 8962\n", - "Test: [ 600/6250] eta: 0:00:45 loss: 0.3993 (0.9455) acc1: 87.5000 (76.9343) acc5: 100.0000 (92.6165) time: 0.0064 data: 0.0006 max mem: 8962\n", - "Test: [ 700/6250] eta: 0:00:43 loss: 0.4326 (0.9321) acc1: 87.5000 (77.1576) acc5: 100.0000 (92.6712) time: 0.0067 data: 0.0006 max mem: 8962\n", - "Test: [ 800/6250] eta: 0:00:42 loss: 0.8489 (0.9613) acc1: 75.0000 (76.5918) acc5: 87.5000 (92.2441) time: 0.0066 data: 0.0008 max mem: 8962\n", - "Test: [ 900/6250] eta: 0:00:41 loss: 0.3832 (0.9033) acc1: 87.5000 (77.9550) acc5: 100.0000 (92.7303) time: 0.0068 data: 0.0005 max mem: 8962\n", - "Test: [1000/6250] eta: 0:00:39 loss: 0.9345 (0.8886) acc1: 75.0000 (78.1094) acc5: 100.0000 (92.9196) time: 0.0061 data: 0.0009 max mem: 8962\n", - "Test: [1100/6250] eta: 0:00:38 loss: 1.0317 (0.9263) acc1: 75.0000 (77.0777) acc5: 100.0000 (92.7679) time: 0.0071 data: 0.0008 max mem: 8962\n", - "Test: [1200/6250] eta: 0:00:37 loss: 1.0881 (0.9341) acc1: 75.0000 (76.5820) acc5: 87.5000 (92.8393) time: 0.0069 data: 0.0007 max mem: 8962\n", - "Test: [1300/6250] eta: 0:00:36 loss: 0.4206 (0.9358) acc1: 75.0000 (76.3259) acc5: 100.0000 (93.0150) time: 0.0074 data: 0.0009 max mem: 8962\n", - "Test: [1400/6250] eta: 0:00:35 loss: 0.8187 (0.9322) acc1: 75.0000 (76.4632) acc5: 87.5000 (93.0496) time: 0.0061 data: 0.0009 max mem: 8962\n", - "Test: [1500/6250] eta: 0:00:34 loss: 0.8115 (0.9372) acc1: 75.0000 (76.2408) acc5: 100.0000 (93.1379) time: 0.0061 data: 0.0006 max mem: 8962\n", - "Test: [1600/6250] eta: 0:00:33 loss: 0.1571 (0.9323) acc1: 100.0000 (76.1165) acc5: 100.0000 (93.2933) time: 0.0064 data: 0.0006 max mem: 8962\n", - "Test: [1700/6250] eta: 0:00:32 loss: 0.9674 (0.9303) acc1: 75.0000 (76.0068) acc5: 100.0000 (93.3936) time: 0.0071 data: 0.0007 max mem: 8962\n", - "Test: [1800/6250] eta: 0:00:32 loss: 1.2235 (0.9404) acc1: 62.5000 (75.7843) acc5: 87.5000 (93.3717) time: 0.0083 data: 0.0008 max mem: 8962\n", - "Test: [1900/6250] eta: 0:00:31 loss: 1.0283 (0.9336) acc1: 62.5000 (76.0652) acc5: 100.0000 (93.4903) time: 0.0054 data: 0.0006 max mem: 8962\n", - "Test: [2000/6250] eta: 0:00:30 loss: 0.4590 (0.9386) acc1: 87.5000 (76.0057) acc5: 100.0000 (93.4533) time: 0.0056 data: 0.0009 max mem: 8962\n", - "Test: [2100/6250] eta: 0:00:29 loss: 0.3773 (0.9199) acc1: 87.5000 (76.5469) acc5: 100.0000 (93.6042) time: 0.0062 data: 0.0007 max mem: 8962\n", - "Test: [2200/6250] eta: 0:00:28 loss: 0.2202 (0.9123) acc1: 87.5000 (76.7208) acc5: 100.0000 (93.6733) time: 0.0054 data: 0.0006 max mem: 8962\n", - "Test: [2300/6250] eta: 0:00:27 loss: 0.6553 (0.9135) acc1: 87.5000 (76.7058) acc5: 100.0000 (93.6821) time: 0.0067 data: 0.0020 max mem: 8962\n", - "Test: [2400/6250] eta: 0:00:27 loss: 0.8205 (0.9230) acc1: 75.0000 (76.5775) acc5: 87.5000 (93.5964) time: 0.0078 data: 0.0017 max mem: 8962\n", - "Test: [2500/6250] eta: 0:00:26 loss: 0.7730 (0.9230) acc1: 87.5000 (76.6693) acc5: 100.0000 (93.5726) time: 0.0055 data: 0.0006 max mem: 8962\n", - "Test: [2600/6250] eta: 0:00:25 loss: 2.1682 (0.9463) acc1: 50.0000 (76.2351) acc5: 75.0000 (93.2526) time: 0.0053 data: 0.0004 max mem: 8962\n", - "Test: [2700/6250] eta: 0:00:24 loss: 0.8635 (0.9574) acc1: 75.0000 (76.0459) acc5: 87.5000 (93.1137) time: 0.0055 data: 0.0005 max mem: 8962\n", - "Test: [2800/6250] eta: 0:00:23 loss: 1.4083 (0.9782) acc1: 62.5000 (75.6025) acc5: 87.5000 (92.8865) time: 0.0055 data: 0.0004 max mem: 8962\n", - "Test: [2900/6250] eta: 0:00:22 loss: 1.8681 (0.9967) acc1: 37.5000 (75.2025) acc5: 87.5000 (92.6706) time: 0.0063 data: 0.0009 max mem: 8962\n", - "Test: [3000/6250] eta: 0:00:22 loss: 2.3759 (1.0186) acc1: 50.0000 (74.8542) acc5: 75.0000 (92.3484) time: 0.0055 data: 0.0006 max mem: 8962\n", - "Test: [3100/6250] eta: 0:00:21 loss: 1.6813 (1.0453) acc1: 50.0000 (74.2986) acc5: 75.0000 (92.0872) time: 0.0058 data: 0.0007 max mem: 8962\n", - "Test: [3200/6250] eta: 0:00:20 loss: 0.7106 (1.0641) acc1: 62.5000 (73.9261) acc5: 100.0000 (91.8658) time: 0.0050 data: 0.0004 max mem: 8962\n", - "Test: [3300/6250] eta: 0:00:19 loss: 1.2310 (1.0768) acc1: 62.5000 (73.5800) acc5: 87.5000 (91.7487) time: 0.0060 data: 0.0005 max mem: 8962\n", - "Test: [3400/6250] eta: 0:00:19 loss: 2.4905 (1.0913) acc1: 50.0000 (73.3203) acc5: 75.0000 (91.5466) time: 0.0059 data: 0.0005 max mem: 8962\n", - "Test: [3500/6250] eta: 0:00:18 loss: 1.7458 (1.0984) acc1: 50.0000 (73.2112) acc5: 87.5000 (91.4310) time: 0.0060 data: 0.0007 max mem: 8962\n", - "Test: [3600/6250] eta: 0:00:17 loss: 0.4052 (1.0948) acc1: 87.5000 (73.3477) acc5: 100.0000 (91.4295) time: 0.0060 data: 0.0008 max mem: 8962\n", - "Test: [3700/6250] eta: 0:00:16 loss: 1.7100 (1.1095) acc1: 62.5000 (73.0444) acc5: 87.5000 (91.2456) time: 0.0104 data: 0.0052 max mem: 8962\n", - "Test: [3800/6250] eta: 0:00:16 loss: 0.7023 (1.1154) acc1: 87.5000 (72.9742) acc5: 87.5000 (91.1471) time: 0.0053 data: 0.0005 max mem: 8962\n", - "Test: [3900/6250] eta: 0:00:15 loss: 2.4000 (1.1317) acc1: 37.5000 (72.6609) acc5: 75.0000 (90.9382) time: 0.0053 data: 0.0005 max mem: 8962\n", - "Test: [4000/6250] eta: 0:00:14 loss: 1.5344 (1.1465) acc1: 62.5000 (72.3882) acc5: 87.5000 (90.7523) time: 0.0056 data: 0.0006 max mem: 8962\n", - "Test: [4100/6250] eta: 0:00:14 loss: 1.4738 (1.1567) acc1: 62.5000 (72.2171) acc5: 87.5000 (90.6364) time: 0.0052 data: 0.0004 max mem: 8962\n", - "Test: [4200/6250] eta: 0:00:13 loss: 0.3974 (1.1608) acc1: 87.5000 (72.0930) acc5: 100.0000 (90.6362) time: 0.0068 data: 0.0005 max mem: 8962\n", - "Test: [4300/6250] eta: 0:00:12 loss: 0.4784 (1.1701) acc1: 75.0000 (71.9542) acc5: 100.0000 (90.4935) time: 0.0052 data: 0.0006 max mem: 8962\n", - "Test: [4400/6250] eta: 0:00:12 loss: 0.9728 (1.1769) acc1: 75.0000 (71.7763) acc5: 100.0000 (90.4368) time: 0.0052 data: 0.0005 max mem: 8962\n", - "Test: [4500/6250] eta: 0:00:11 loss: 1.1849 (1.1823) acc1: 75.0000 (71.6785) acc5: 87.5000 (90.3855) time: 0.0058 data: 0.0004 max mem: 8962\n", - "Test: [4600/6250] eta: 0:00:10 loss: 1.6483 (1.1923) acc1: 50.0000 (71.5116) acc5: 87.5000 (90.2413) time: 0.0056 data: 0.0007 max mem: 8962\n", - "Test: [4700/6250] eta: 0:00:10 loss: 1.3664 (1.2023) acc1: 62.5000 (71.2189) acc5: 87.5000 (90.1032) time: 0.0055 data: 0.0007 max mem: 8962\n", - "Test: [4800/6250] eta: 0:00:09 loss: 1.3209 (1.2095) acc1: 62.5000 (71.0841) acc5: 87.5000 (89.9969) time: 0.0051 data: 0.0008 max mem: 8962\n", - "Test: [4900/6250] eta: 0:00:08 loss: 0.4727 (1.2146) acc1: 87.5000 (71.0161) acc5: 100.0000 (89.9077) time: 0.0049 data: 0.0004 max mem: 8962\n", - "Test: [5000/6250] eta: 0:00:08 loss: 2.0785 (1.2263) acc1: 62.5000 (70.8483) acc5: 75.0000 (89.7520) time: 0.0054 data: 0.0005 max mem: 8962\n", - "Test: [5100/6250] eta: 0:00:07 loss: 1.2702 (1.2315) acc1: 62.5000 (70.7410) acc5: 87.5000 (89.7153) time: 0.0059 data: 0.0007 max mem: 8962\n", - "Test: [5200/6250] eta: 0:00:06 loss: 1.0106 (1.2380) acc1: 75.0000 (70.6138) acc5: 87.5000 (89.6510) time: 0.0047 data: 0.0004 max mem: 8962\n", - "Test: [5300/6250] eta: 0:00:06 loss: 1.4912 (1.2517) acc1: 50.0000 (70.3122) acc5: 87.5000 (89.4619) time: 0.0056 data: 0.0005 max mem: 8962\n", - "Test: [5400/6250] eta: 0:00:05 loss: 0.9010 (1.2548) acc1: 75.0000 (70.2671) acc5: 100.0000 (89.4071) time: 0.0057 data: 0.0006 max mem: 8962\n", - "Test: [5500/6250] eta: 0:00:04 loss: 0.9291 (1.2585) acc1: 75.0000 (70.1622) acc5: 87.5000 (89.3656) time: 0.0056 data: 0.0005 max mem: 8962\n", - "Test: [5600/6250] eta: 0:00:04 loss: 0.9468 (1.2638) acc1: 62.5000 (70.0857) acc5: 87.5000 (89.2586) time: 0.0060 data: 0.0016 max mem: 8962\n", - "Test: [5700/6250] eta: 0:00:03 loss: 2.1460 (1.2784) acc1: 37.5000 (69.7926) acc5: 75.0000 (89.1203) time: 0.0050 data: 0.0006 max mem: 8962\n", - "Test: [5800/6250] eta: 0:00:02 loss: 0.6562 (1.2747) acc1: 75.0000 (69.8522) acc5: 100.0000 (89.1657) time: 0.0053 data: 0.0008 max mem: 8962\n", - "Test: [5900/6250] eta: 0:00:02 loss: 1.0713 (1.2722) acc1: 75.0000 (69.8759) acc5: 100.0000 (89.2095) time: 0.0053 data: 0.0007 max mem: 8962\n", - "Test: [6000/6250] eta: 0:00:01 loss: 0.5802 (1.2647) acc1: 87.5000 (70.0342) acc5: 87.5000 (89.2955) time: 0.0057 data: 0.0006 max mem: 8962\n", - "Test: [6100/6250] eta: 0:00:00 loss: 1.2877 (1.2731) acc1: 62.5000 (69.8328) acc5: 87.5000 (89.2046) time: 0.0052 data: 0.0008 max mem: 8962\n", - "Test: [6200/6250] eta: 0:00:00 loss: 0.2487 (1.2671) acc1: 87.5000 (69.9544) acc5: 100.0000 (89.2739) time: 0.0061 data: 0.0008 max mem: 8962\n", - "Test: Total time: 0:00:39\n", - "Test: Acc@1 69.994 Acc@5 89.292\n", - "Epoch: [4] [ 0/5005] eta: 6:27:39 lr: 0.0002 img/s: 465.53450010860746 loss: 0.8434 (0.8434) acc1: 71.0938 (71.0938) acc5: 87.5000 (87.5000) meanQV: 1.4977 (1.4977) stdQV: 0.3180 (0.3180) time: 4.6472 data: 4.0972 max mem: 8962\n", - "Epoch: [4] [ 100/5005] eta: 0:45:55 lr: 0.0002 img/s: 490.74347871791 loss: 0.9550 (0.8984) acc1: 67.1875 (69.1097) acc5: 86.7188 (87.1519) meanQV: 1.4691 (1.4833) stdQV: 0.3260 (0.3231) time: 0.5208 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [ 200/5005] eta: 0:43:22 lr: 0.0002 img/s: 493.9701025625455 loss: 0.8835 (0.9029) acc1: 68.7500 (68.8783) acc5: 87.5000 (87.0375) meanQV: 1.4809 (1.4816) stdQV: 0.3240 (0.3238) time: 0.5217 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [ 300/5005] eta: 0:41:56 lr: 0.0002 img/s: 491.63170037870236 loss: 0.9149 (0.9015) acc1: 67.9688 (68.7227) acc5: 87.8906 (87.0367) meanQV: 1.4758 (1.4805) stdQV: 0.3260 (0.3243) time: 0.5222 data: 0.0005 max mem: 8962\n", - "Epoch: [4] [ 400/5005] eta: 0:40:46 lr: 0.0002 img/s: 489.97201559708594 loss: 0.8910 (0.8983) acc1: 68.7500 (68.8055) acc5: 86.3281 (87.0305) meanQV: 1.4797 (1.4812) stdQV: 0.3246 (0.3240) time: 0.5209 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [ 500/5005] eta: 0:39:44 lr: 0.0002 img/s: 494.2242973791529 loss: 0.8698 (0.8962) acc1: 69.9219 (68.8599) acc5: 86.7188 (87.0556) meanQV: 1.4895 (1.4815) stdQV: 0.3214 (0.3239) time: 0.5205 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [ 600/5005] eta: 0:38:45 lr: 0.0002 img/s: 491.0853097714258 loss: 0.8708 (0.8966) acc1: 70.3125 (68.8962) acc5: 87.1094 (87.0457) meanQV: 1.4922 (1.4818) stdQV: 0.3188 (0.3238) time: 0.5203 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [ 700/5005] eta: 0:37:48 lr: 0.0002 img/s: 493.5375728016555 loss: 0.9078 (0.8985) acc1: 68.7500 (68.8353) acc5: 87.1094 (87.0018) meanQV: 1.4813 (1.4814) stdQV: 0.3248 (0.3240) time: 0.5213 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [ 800/5005] eta: 0:36:52 lr: 0.0002 img/s: 490.2532232540004 loss: 0.9188 (0.8979) acc1: 67.9688 (68.8904) acc5: 86.3281 (86.9953) meanQV: 1.4746 (1.4818) stdQV: 0.3270 (0.3238) time: 0.5209 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [ 900/5005] eta: 0:35:57 lr: 0.0002 img/s: 493.18552460181996 loss: 0.9238 (0.8994) acc1: 67.9688 (68.8575) acc5: 86.3281 (86.9680) meanQV: 1.4758 (1.4815) stdQV: 0.3262 (0.3239) time: 0.5208 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [1000/5005] eta: 0:35:03 lr: 0.0002 img/s: 490.8677666455156 loss: 0.8737 (0.8983) acc1: 69.9219 (68.8835) acc5: 87.1094 (86.9892) meanQV: 1.4895 (1.4817) stdQV: 0.3192 (0.3238) time: 0.5206 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [1100/5005] eta: 0:34:09 lr: 0.0002 img/s: 494.22497982810273 loss: 0.8912 (0.8984) acc1: 69.1406 (68.9065) acc5: 86.7188 (87.0005) meanQV: 1.4840 (1.4819) stdQV: 0.3204 (0.3237) time: 0.5215 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [1200/5005] eta: 0:33:16 lr: 0.0002 img/s: 491.8136503316876 loss: 0.9000 (0.8996) acc1: 68.7500 (68.8476) acc5: 87.5000 (86.9653) meanQV: 1.4801 (1.4814) stdQV: 0.3248 (0.3239) time: 0.5211 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [1300/5005] eta: 0:32:22 lr: 0.0002 img/s: 494.33123736827326 loss: 0.8564 (0.8992) acc1: 70.3125 (68.8515) acc5: 87.5000 (86.9686) meanQV: 1.4918 (1.4815) stdQV: 0.3192 (0.3239) time: 0.5202 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [1400/5005] eta: 0:31:29 lr: 0.0002 img/s: 493.6809842522038 loss: 0.8581 (0.8993) acc1: 69.5312 (68.8392) acc5: 87.5000 (86.9747) meanQV: 1.4867 (1.4814) stdQV: 0.3216 (0.3239) time: 0.5203 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [1500/5005] eta: 0:30:36 lr: 0.0002 img/s: 492.017828708216 loss: 0.8798 (0.8990) acc1: 68.7500 (68.8564) acc5: 86.7188 (86.9821) meanQV: 1.4801 (1.4815) stdQV: 0.3216 (0.3239) time: 0.5223 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [1600/5005] eta: 0:29:43 lr: 0.0002 img/s: 494.6657802925495 loss: 0.8522 (0.8990) acc1: 69.9219 (68.8637) acc5: 87.1094 (86.9842) meanQV: 1.4887 (1.4816) stdQV: 0.3213 (0.3239) time: 0.5216 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [1700/5005] eta: 0:28:50 lr: 0.0002 img/s: 499.96709108899364 loss: 0.8898 (0.8983) acc1: 69.5312 (68.8800) acc5: 86.7188 (86.9978) meanQV: 1.4867 (1.4817) stdQV: 0.3215 (0.3238) time: 0.5219 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [1800/5005] eta: 0:27:57 lr: 0.0002 img/s: 491.00491534530653 loss: 0.9142 (0.8985) acc1: 68.3594 (68.8823) acc5: 86.7188 (86.9988) meanQV: 1.4781 (1.4817) stdQV: 0.3251 (0.3238) time: 0.5209 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [1900/5005] eta: 0:27:05 lr: 0.0002 img/s: 489.01270554630025 loss: 0.8833 (0.8986) acc1: 68.7500 (68.8708) acc5: 87.1094 (86.9945) meanQV: 1.4801 (1.4816) stdQV: 0.3248 (0.3238) time: 0.5220 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [2000/5005] eta: 0:26:12 lr: 0.0002 img/s: 485.6457289761854 loss: 0.8689 (0.8982) acc1: 69.1406 (68.8986) acc5: 87.5000 (87.0073) meanQV: 1.4824 (1.4818) stdQV: 0.3235 (0.3238) time: 0.5208 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [2100/5005] eta: 0:25:19 lr: 0.0002 img/s: 490.798211492559 loss: 0.8893 (0.8982) acc1: 68.7500 (68.8917) acc5: 87.5000 (87.0095) meanQV: 1.4813 (1.4817) stdQV: 0.3251 (0.3238) time: 0.5210 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [2200/5005] eta: 0:24:27 lr: 0.0002 img/s: 493.2516794802064 loss: 0.8902 (0.8984) acc1: 68.3594 (68.8870) acc5: 87.1094 (87.0112) meanQV: 1.4785 (1.4817) stdQV: 0.3248 (0.3238) time: 0.5206 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [2300/5005] eta: 0:23:34 lr: 0.0002 img/s: 487.6654664365519 loss: 0.8884 (0.8983) acc1: 68.7500 (68.8970) acc5: 86.3281 (87.0163) meanQV: 1.4813 (1.4818) stdQV: 0.3240 (0.3238) time: 0.5224 data: 0.0002 max mem: 8962\n", - "Epoch: [4] [2400/5005] eta: 0:22:42 lr: 0.0002 img/s: 487.97818203963027 loss: 0.9129 (0.8987) acc1: 68.3594 (68.9003) acc5: 87.5000 (87.0210) meanQV: 1.4785 (1.4818) stdQV: 0.3251 (0.3238) time: 0.5218 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [2500/5005] eta: 0:21:49 lr: 0.0002 img/s: 493.4107957036287 loss: 0.8948 (0.8988) acc1: 68.7500 (68.9018) acc5: 86.3281 (87.0080) meanQV: 1.4809 (1.4818) stdQV: 0.3240 (0.3238) time: 0.5208 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [2600/5005] eta: 0:20:57 lr: 0.0002 img/s: 491.60288897587867 loss: 0.8879 (0.8991) acc1: 68.7500 (68.9072) acc5: 86.3281 (87.0161) meanQV: 1.4813 (1.4818) stdQV: 0.3226 (0.3237) time: 0.5208 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [2700/5005] eta: 0:20:04 lr: 0.0002 img/s: 491.1488804236057 loss: 0.8520 (0.8993) acc1: 69.9219 (68.9277) acc5: 87.8906 (87.0213) meanQV: 1.4895 (1.4820) stdQV: 0.3204 (0.3237) time: 0.5217 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [2800/5005] eta: 0:19:12 lr: 0.0002 img/s: 493.3804151641137 loss: 0.8855 (0.8992) acc1: 69.1406 (68.9271) acc5: 87.1094 (87.0320) meanQV: 1.4828 (1.4820) stdQV: 0.3228 (0.3237) time: 0.5209 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [2900/5005] eta: 0:18:20 lr: 0.0002 img/s: 494.2067818344083 loss: 0.9216 (0.8993) acc1: 68.7500 (68.9377) acc5: 86.3281 (87.0301) meanQV: 1.4813 (1.4821) stdQV: 0.3251 (0.3237) time: 0.5215 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [3000/5005] eta: 0:17:27 lr: 0.0002 img/s: 491.1448365704542 loss: 0.9209 (0.8991) acc1: 68.3594 (68.9410) acc5: 86.7188 (87.0258) meanQV: 1.4785 (1.4821) stdQV: 0.3249 (0.3237) time: 0.5223 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [3100/5005] eta: 0:16:35 lr: 0.0002 img/s: 492.0568356895722 loss: 0.8835 (0.8994) acc1: 68.7500 (68.9353) acc5: 87.1094 (87.0167) meanQV: 1.4813 (1.4820) stdQV: 0.3226 (0.3237) time: 0.5203 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [3200/5005] eta: 0:15:43 lr: 0.0002 img/s: 491.44133759107467 loss: 0.8832 (0.8996) acc1: 69.9219 (68.9327) acc5: 87.1094 (87.0172) meanQV: 1.4883 (1.4820) stdQV: 0.3214 (0.3237) time: 0.5212 data: 0.0005 max mem: 8962\n", - "Epoch: [4] [3300/5005] eta: 0:14:50 lr: 0.0002 img/s: 487.73613035951706 loss: 0.8594 (0.8997) acc1: 69.1406 (68.9283) acc5: 87.8906 (87.0199) meanQV: 1.4840 (1.4820) stdQV: 0.3228 (0.3237) time: 0.5213 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [3400/5005] eta: 0:13:58 lr: 0.0002 img/s: 491.204377802288 loss: 0.9090 (0.8999) acc1: 68.7500 (68.9294) acc5: 86.3281 (87.0150) meanQV: 1.4789 (1.4820) stdQV: 0.3246 (0.3237) time: 0.5213 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [3500/5005] eta: 0:13:06 lr: 0.0002 img/s: 495.01811999100084 loss: 0.9384 (0.9003) acc1: 68.3594 (68.9144) acc5: 86.3281 (87.0097) meanQV: 1.4785 (1.4819) stdQV: 0.3262 (0.3237) time: 0.5219 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [3600/5005] eta: 0:12:13 lr: 0.0002 img/s: 488.85240665181243 loss: 0.8764 (0.9000) acc1: 69.9219 (68.9205) acc5: 87.8906 (87.0174) meanQV: 1.4895 (1.4819) stdQV: 0.3204 (0.3237) time: 0.5217 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [3700/5005] eta: 0:11:21 lr: 0.0002 img/s: 488.62905411616515 loss: 0.8644 (0.8999) acc1: 69.9219 (68.9198) acc5: 86.7188 (87.0206) meanQV: 1.4895 (1.4819) stdQV: 0.3204 (0.3237) time: 0.5218 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [3800/5005] eta: 0:10:29 lr: 0.0002 img/s: 488.65351505542134 loss: 0.8901 (0.8997) acc1: 68.7500 (68.9341) acc5: 87.5000 (87.0288) meanQV: 1.4789 (1.4820) stdQV: 0.3248 (0.3237) time: 0.5216 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [3900/5005] eta: 0:09:37 lr: 0.0002 img/s: 493.65601740808734 loss: 0.8595 (0.8997) acc1: 68.3594 (68.9418) acc5: 87.1094 (87.0309) meanQV: 1.4785 (1.4821) stdQV: 0.3262 (0.3237) time: 0.5212 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [4000/5005] eta: 0:08:44 lr: 0.0002 img/s: 492.87718715199964 loss: 0.8968 (0.8997) acc1: 69.9219 (68.9440) acc5: 87.5000 (87.0334) meanQV: 1.4895 (1.4821) stdQV: 0.3204 (0.3237) time: 0.5208 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [4100/5005] eta: 0:07:52 lr: 0.0002 img/s: 488.15255480193474 loss: 0.8810 (0.8995) acc1: 68.3594 (68.9461) acc5: 87.5000 (87.0394) meanQV: 1.4785 (1.4821) stdQV: 0.3251 (0.3237) time: 0.5210 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [4200/5005] eta: 0:07:00 lr: 0.0002 img/s: 490.5603024844127 loss: 0.9360 (0.8999) acc1: 68.7500 (68.9386) acc5: 85.9375 (87.0355) meanQV: 1.4813 (1.4820) stdQV: 0.3240 (0.3237) time: 0.5213 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [4300/5005] eta: 0:06:08 lr: 0.0002 img/s: 493.6328685908266 loss: 0.8927 (0.9000) acc1: 67.9688 (68.9293) acc5: 87.1094 (87.0393) meanQV: 1.4758 (1.4820) stdQV: 0.3273 (0.3237) time: 0.5209 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [4400/5005] eta: 0:05:15 lr: 0.0002 img/s: 500.0339136176978 loss: 0.8938 (0.9000) acc1: 67.9688 (68.9281) acc5: 87.5000 (87.0415) meanQV: 1.4746 (1.4820) stdQV: 0.3237 (0.3237) time: 0.5203 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [4500/5005] eta: 0:04:23 lr: 0.0002 img/s: 491.36240054840806 loss: 0.8792 (0.9000) acc1: 69.5312 (68.9270) acc5: 87.1094 (87.0380) meanQV: 1.4832 (1.4820) stdQV: 0.3213 (0.3237) time: 0.5218 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [4600/5005] eta: 0:03:31 lr: 0.0002 img/s: 494.30324647494075 loss: 0.9115 (0.9000) acc1: 69.1406 (68.9302) acc5: 86.7188 (87.0371) meanQV: 1.4828 (1.4820) stdQV: 0.3238 (0.3237) time: 0.5208 data: 0.0003 max mem: 8962\n", - "Epoch: [4] [4700/5005] eta: 0:02:39 lr: 0.0002 img/s: 494.0173748211402 loss: 0.8860 (0.9000) acc1: 69.9219 (68.9322) acc5: 87.1094 (87.0404) meanQV: 1.4895 (1.4820) stdQV: 0.3202 (0.3237) time: 0.5212 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [4800/5005] eta: 0:01:47 lr: 0.0002 img/s: 493.9932830203957 loss: 0.9224 (0.9002) acc1: 68.7500 (68.9312) acc5: 87.1094 (87.0430) meanQV: 1.4801 (1.4820) stdQV: 0.3248 (0.3237) time: 0.5215 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [4900/5005] eta: 0:00:54 lr: 0.0002 img/s: 494.8479308631688 loss: 0.8930 (0.9002) acc1: 67.9688 (68.9332) acc5: 86.7188 (87.0443) meanQV: 1.4758 (1.4820) stdQV: 0.3262 (0.3237) time: 0.5202 data: 0.0004 max mem: 8962\n", - "Epoch: [4] [5000/5005] eta: 0:00:02 lr: 0.0002 img/s: 492.44182560982995 loss: 0.8594 (0.8999) acc1: 70.3125 (68.9486) acc5: 87.5000 (87.0523) meanQV: 1.4922 (1.4821) stdQV: 0.3192 (0.3236) time: 0.5203 data: 0.0002 max mem: 8962\n", - "Epoch: [4] Total time: 0:43:33\n", - "Test: [ 0/6250] eta: 1:16:02 loss: 0.7909 (0.7909) acc1: 75.0000 (75.0000) acc5: 100.0000 (100.0000) time: 0.7299 data: 0.7206 max mem: 8962\n", - "Test: [ 100/6250] eta: 0:01:29 loss: 0.1878 (0.6247) acc1: 87.5000 (83.4158) acc5: 100.0000 (95.2970) time: 0.0065 data: 0.0006 max mem: 8962\n", - "Test: [ 200/6250] eta: 0:01:04 loss: 0.6666 (0.6644) acc1: 87.5000 (84.3284) acc5: 100.0000 (94.5274) time: 0.0077 data: 0.0020 max mem: 8962\n", - "Test: [ 300/6250] eta: 0:00:54 loss: 1.2709 (0.8699) acc1: 62.5000 (78.9037) acc5: 87.5000 (93.0648) time: 0.0064 data: 0.0008 max mem: 8962\n", - "Test: [ 400/6250] eta: 0:00:49 loss: 0.8989 (0.9948) acc1: 75.0000 (76.1222) acc5: 100.0000 (92.1135) time: 0.0065 data: 0.0005 max mem: 8962\n", - "Test: [ 500/6250] eta: 0:00:45 loss: 1.1087 (1.0409) acc1: 75.0000 (74.9251) acc5: 87.5000 (91.7914) time: 0.0068 data: 0.0025 max mem: 8962\n", - "Test: [ 600/6250] eta: 0:00:43 loss: 0.4547 (0.9505) acc1: 87.5000 (76.8927) acc5: 100.0000 (92.4293) time: 0.0059 data: 0.0005 max mem: 8962\n", - "Test: [ 700/6250] eta: 0:00:40 loss: 0.3772 (0.9356) acc1: 87.5000 (77.1755) acc5: 100.0000 (92.4215) time: 0.0055 data: 0.0004 max mem: 8962\n", - "Test: [ 800/6250] eta: 0:00:40 loss: 0.8068 (0.9664) acc1: 75.0000 (76.5762) acc5: 87.5000 (92.0568) time: 0.0060 data: 0.0004 max mem: 8962\n", - "Test: [ 900/6250] eta: 0:00:38 loss: 0.3764 (0.9114) acc1: 87.5000 (77.8996) acc5: 100.0000 (92.5361) time: 0.0062 data: 0.0004 max mem: 8962\n", - "Test: [1000/6250] eta: 0:00:37 loss: 0.9310 (0.8980) acc1: 75.0000 (78.1094) acc5: 100.0000 (92.7572) time: 0.0063 data: 0.0006 max mem: 8962\n", - "Test: [1100/6250] eta: 0:00:37 loss: 1.1327 (0.9363) acc1: 75.0000 (77.1571) acc5: 100.0000 (92.5636) time: 0.0073 data: 0.0006 max mem: 8962\n", - "Test: [1200/6250] eta: 0:00:36 loss: 1.0381 (0.9410) acc1: 75.0000 (76.7902) acc5: 87.5000 (92.6728) time: 0.0076 data: 0.0006 max mem: 8962\n", - "Test: [1300/6250] eta: 0:00:35 loss: 0.4042 (0.9409) acc1: 87.5000 (76.5853) acc5: 100.0000 (92.8997) time: 0.0066 data: 0.0007 max mem: 8962\n", - "Test: [1400/6250] eta: 0:00:34 loss: 0.7884 (0.9356) acc1: 75.0000 (76.6774) acc5: 87.5000 (92.9961) time: 0.0063 data: 0.0005 max mem: 8962\n", - "Test: [1500/6250] eta: 0:00:33 loss: 0.9547 (0.9392) acc1: 75.0000 (76.4074) acc5: 100.0000 (93.1046) time: 0.0058 data: 0.0005 max mem: 8962\n", - "Test: [1600/6250] eta: 0:00:32 loss: 0.1298 (0.9348) acc1: 87.5000 (76.2492) acc5: 100.0000 (93.2464) time: 0.0063 data: 0.0006 max mem: 8962\n", - "Test: [1700/6250] eta: 0:00:32 loss: 1.0550 (0.9324) acc1: 75.0000 (76.1023) acc5: 100.0000 (93.3715) time: 0.0092 data: 0.0041 max mem: 8962\n", - "Test: [1800/6250] eta: 0:00:31 loss: 1.2747 (0.9422) acc1: 62.5000 (75.8537) acc5: 87.5000 (93.3648) time: 0.0064 data: 0.0004 max mem: 8962\n", - "Test: [1900/6250] eta: 0:00:30 loss: 0.9273 (0.9352) acc1: 75.0000 (76.1244) acc5: 100.0000 (93.4705) time: 0.0065 data: 0.0005 max mem: 8962\n", - "Test: [2000/6250] eta: 0:00:30 loss: 0.4722 (0.9411) acc1: 87.5000 (76.0120) acc5: 100.0000 (93.4658) time: 0.0064 data: 0.0006 max mem: 8962\n", - "Test: [2100/6250] eta: 0:00:29 loss: 0.3972 (0.9220) acc1: 87.5000 (76.5469) acc5: 100.0000 (93.6161) time: 0.0064 data: 0.0005 max mem: 8962\n", - "Test: [2200/6250] eta: 0:00:28 loss: 0.1838 (0.9145) acc1: 87.5000 (76.6924) acc5: 100.0000 (93.6733) time: 0.0052 data: 0.0004 max mem: 8962\n", - "Test: [2300/6250] eta: 0:00:27 loss: 0.7064 (0.9173) acc1: 87.5000 (76.6515) acc5: 100.0000 (93.6441) time: 0.0061 data: 0.0019 max mem: 8962\n", - "Test: [2400/6250] eta: 0:00:26 loss: 0.9312 (0.9272) acc1: 75.0000 (76.5618) acc5: 87.5000 (93.5392) time: 0.0057 data: 0.0006 max mem: 8962\n", - "Test: [2500/6250] eta: 0:00:26 loss: 0.9157 (0.9291) acc1: 75.0000 (76.6144) acc5: 100.0000 (93.4976) time: 0.0060 data: 0.0006 max mem: 8962\n", - "Test: [2600/6250] eta: 0:00:25 loss: 2.2484 (0.9531) acc1: 50.0000 (76.1246) acc5: 62.5000 (93.1949) time: 0.0059 data: 0.0005 max mem: 8962\n", - "Test: [2700/6250] eta: 0:00:24 loss: 0.8119 (0.9641) acc1: 75.0000 (75.9210) acc5: 87.5000 (93.0766) time: 0.0057 data: 0.0004 max mem: 8962\n", - "Test: [2800/6250] eta: 0:00:23 loss: 1.5186 (0.9860) acc1: 62.5000 (75.4418) acc5: 75.0000 (92.8240) time: 0.0058 data: 0.0004 max mem: 8962\n", - "Test: [2900/6250] eta: 0:00:23 loss: 1.7899 (1.0040) acc1: 37.5000 (75.0603) acc5: 87.5000 (92.6060) time: 0.0078 data: 0.0007 max mem: 8962\n", - "Test: [3000/6250] eta: 0:00:22 loss: 2.3188 (1.0252) acc1: 50.0000 (74.7418) acc5: 75.0000 (92.2817) time: 0.0062 data: 0.0006 max mem: 8962\n", - "Test: [3100/6250] eta: 0:00:21 loss: 1.8096 (1.0517) acc1: 50.0000 (74.2341) acc5: 75.0000 (92.0106) time: 0.0056 data: 0.0004 max mem: 8962\n", - "Test: [3200/6250] eta: 0:00:20 loss: 0.6736 (1.0715) acc1: 75.0000 (73.8402) acc5: 100.0000 (91.8072) time: 0.0057 data: 0.0004 max mem: 8962\n", - "Test: [3300/6250] eta: 0:00:19 loss: 1.2702 (1.0843) acc1: 62.5000 (73.4853) acc5: 87.5000 (91.6919) time: 0.0060 data: 0.0009 max mem: 8962\n", - "Test: [3400/6250] eta: 0:00:19 loss: 2.4354 (1.0978) acc1: 37.5000 (73.2468) acc5: 75.0000 (91.5062) time: 0.0059 data: 0.0005 max mem: 8962\n", - "Test: [3500/6250] eta: 0:00:18 loss: 1.7555 (1.1037) acc1: 50.0000 (73.1720) acc5: 75.0000 (91.3846) time: 0.0058 data: 0.0006 max mem: 8962\n", - "Test: [3600/6250] eta: 0:00:17 loss: 0.5104 (1.0999) acc1: 87.5000 (73.3164) acc5: 100.0000 (91.3948) time: 0.0054 data: 0.0005 max mem: 8962\n", - "Test: [3700/6250] eta: 0:00:17 loss: 1.7609 (1.1145) acc1: 62.5000 (73.0276) acc5: 87.5000 (91.2017) time: 0.0116 data: 0.0071 max mem: 8962\n", - "Test: [3800/6250] eta: 0:00:16 loss: 0.6589 (1.1207) acc1: 87.5000 (72.9446) acc5: 87.5000 (91.1076) time: 0.0069 data: 0.0008 max mem: 8962\n", - "Test: [3900/6250] eta: 0:00:15 loss: 2.3705 (1.1368) acc1: 50.0000 (72.6224) acc5: 75.0000 (90.8837) time: 0.0063 data: 0.0010 max mem: 8962\n", - "Test: [4000/6250] eta: 0:00:14 loss: 1.3760 (1.1507) acc1: 62.5000 (72.3663) acc5: 87.5000 (90.7179) time: 0.0052 data: 0.0004 max mem: 8962\n", - "Test: [4100/6250] eta: 0:00:14 loss: 1.3587 (1.1609) acc1: 62.5000 (72.1836) acc5: 87.5000 (90.6059) time: 0.0056 data: 0.0005 max mem: 8962\n", - "Test: [4200/6250] eta: 0:00:13 loss: 0.4211 (1.1651) acc1: 87.5000 (72.0543) acc5: 100.0000 (90.5975) time: 0.0050 data: 0.0004 max mem: 8962\n", - "Test: [4300/6250] eta: 0:00:12 loss: 0.4845 (1.1744) acc1: 75.0000 (71.9339) acc5: 87.5000 (90.4354) time: 0.0056 data: 0.0006 max mem: 8962\n", - "Test: [4400/6250] eta: 0:00:12 loss: 0.9234 (1.1813) acc1: 75.0000 (71.7422) acc5: 100.0000 (90.3885) time: 0.0054 data: 0.0005 max mem: 8962\n", - "Test: [4500/6250] eta: 0:00:11 loss: 1.2129 (1.1870) acc1: 62.5000 (71.6230) acc5: 87.5000 (90.3299) time: 0.0060 data: 0.0008 max mem: 8962\n", - "Test: [4600/6250] eta: 0:00:10 loss: 1.6169 (1.1973) acc1: 50.0000 (71.4519) acc5: 87.5000 (90.1652) time: 0.0059 data: 0.0006 max mem: 8962\n", - "Test: [4700/6250] eta: 0:00:10 loss: 1.4572 (1.2073) acc1: 62.5000 (71.1471) acc5: 87.5000 (90.0181) time: 0.0054 data: 0.0005 max mem: 8962\n", - "Test: [4800/6250] eta: 0:00:09 loss: 1.1656 (1.2140) acc1: 62.5000 (71.0347) acc5: 87.5000 (89.9188) time: 0.0057 data: 0.0005 max mem: 8962\n", - "Test: [4900/6250] eta: 0:00:08 loss: 0.4505 (1.2183) acc1: 87.5000 (70.9779) acc5: 100.0000 (89.8312) time: 0.0050 data: 0.0005 max mem: 8962\n", - "Test: [5000/6250] eta: 0:00:08 loss: 2.1536 (1.2300) acc1: 62.5000 (70.8008) acc5: 75.0000 (89.6821) time: 0.0061 data: 0.0004 max mem: 8962\n", - "Test: [5100/6250] eta: 0:00:07 loss: 1.3107 (1.2352) acc1: 62.5000 (70.6920) acc5: 100.0000 (89.6442) time: 0.0064 data: 0.0008 max mem: 8962\n", - "Test: [5200/6250] eta: 0:00:06 loss: 1.0626 (1.2412) acc1: 75.0000 (70.5850) acc5: 87.5000 (89.5741) time: 0.0053 data: 0.0005 max mem: 8962\n", - "Test: [5300/6250] eta: 0:00:06 loss: 1.4655 (1.2552) acc1: 62.5000 (70.2650) acc5: 87.5000 (89.3864) time: 0.0063 data: 0.0005 max mem: 8962\n", - "Test: [5400/6250] eta: 0:00:05 loss: 0.8139 (1.2579) acc1: 75.0000 (70.2162) acc5: 87.5000 (89.3469) time: 0.0053 data: 0.0004 max mem: 8962\n", - "Test: [5500/6250] eta: 0:00:04 loss: 0.9340 (1.2610) acc1: 75.0000 (70.1213) acc5: 87.5000 (89.3065) time: 0.0055 data: 0.0004 max mem: 8962\n", - "Test: [5600/6250] eta: 0:00:04 loss: 1.0789 (1.2661) acc1: 62.5000 (70.0433) acc5: 87.5000 (89.2140) time: 0.0064 data: 0.0013 max mem: 8962\n", - "Test: [5700/6250] eta: 0:00:03 loss: 2.0526 (1.2814) acc1: 37.5000 (69.7422) acc5: 75.0000 (89.0567) time: 0.0050 data: 0.0004 max mem: 8962\n", - "Test: [5800/6250] eta: 0:00:02 loss: 0.5955 (1.2775) acc1: 75.0000 (69.8220) acc5: 100.0000 (89.0989) time: 0.0068 data: 0.0007 max mem: 8962\n", - "Test: [5900/6250] eta: 0:00:02 loss: 1.1023 (1.2751) acc1: 62.5000 (69.8483) acc5: 100.0000 (89.1353) time: 0.0056 data: 0.0006 max mem: 8962\n", - "Test: [6000/6250] eta: 0:00:01 loss: 0.5431 (1.2677) acc1: 87.5000 (69.9988) acc5: 87.5000 (89.2289) time: 0.0058 data: 0.0004 max mem: 8962\n", - "Test: [6100/6250] eta: 0:00:00 loss: 1.3317 (1.2758) acc1: 62.5000 (69.8164) acc5: 87.5000 (89.1370) time: 0.0054 data: 0.0004 max mem: 8962\n", - "Test: [6200/6250] eta: 0:00:00 loss: 0.2130 (1.2696) acc1: 87.5000 (69.9444) acc5: 100.0000 (89.2114) time: 0.0065 data: 0.0005 max mem: 8962\n", - "Test: Total time: 0:00:40\n", - "Test: Acc@1 69.982 Acc@5 89.216\n", - "Epoch: [5] [ 0/5005] eta: 6:26:06 lr: 0.0001 img/s: 472.08087977640673 loss: 0.9772 (0.9772) acc1: 70.3125 (70.3125) acc5: 87.8906 (87.8906) meanQV: 1.4922 (1.4922) stdQV: 0.3204 (0.3204) time: 4.6287 data: 4.0864 max mem: 8962\n", - "Epoch: [5] [ 100/5005] eta: 0:45:55 lr: 0.0001 img/s: 493.412156113808 loss: 0.8671 (0.8977) acc1: 69.1406 (69.1445) acc5: 87.1094 (87.3028) meanQV: 1.4840 (1.4833) stdQV: 0.3228 (0.3230) time: 0.5215 data: 0.0004 max mem: 8962\n", - "Epoch: [5] [ 200/5005] eta: 0:42:34 lr: 0.0001 img/s: 633.4831033504211 loss: 0.8743 (0.9023) acc1: 69.5312 (69.1465) acc5: 87.1094 (87.1774) meanQV: 1.4867 (1.4834) stdQV: 0.3216 (0.3230) time: 0.4189 data: 0.0004 max mem: 8962\n", - "Epoch: [5] [ 300/5005] eta: 0:37:51 lr: 0.0001 img/s: 491.4145726182402 loss: 0.8887 (0.9009) acc1: 69.5312 (69.1938) acc5: 87.8906 (87.1872) meanQV: 1.4867 (1.4838) stdQV: 0.3204 (0.3229) time: 0.3729 data: 0.0032 max mem: 8962\n", - "Epoch: [5] [ 400/5005] eta: 0:37:47 lr: 0.0001 img/s: 488.7140108034497 loss: 0.9087 (0.9030) acc1: 68.3594 (69.0539) acc5: 87.5000 (87.1396) meanQV: 1.4781 (1.4828) stdQV: 0.3251 (0.3233) time: 0.5207 data: 0.0003 max mem: 8962\n", - "Epoch: [5] [ 500/5005] eta: 0:37:23 lr: 0.0001 img/s: 489.01871879991273 loss: 0.8969 (0.9043) acc1: 67.5781 (68.9605) acc5: 87.1094 (87.0493) meanQV: 1.4730 (1.4821) stdQV: 0.3249 (0.3237) time: 0.5216 data: 0.0003 max mem: 8962\n", - "Epoch: [5] [ 600/5005] eta: 0:36:50 lr: 0.0001 img/s: 494.24204166440353 loss: 0.9185 (0.9048) acc1: 68.3594 (68.9502) acc5: 87.1094 (87.0873) meanQV: 1.4785 (1.4821) stdQV: 0.3262 (0.3237) time: 0.5209 data: 0.0004 max mem: 8962\n", - "Epoch: [5] [ 700/5005] eta: 0:36:12 lr: 0.0001 img/s: 490.8462248780818 loss: 0.9053 (0.9051) acc1: 68.3594 (68.9840) acc5: 85.9375 (87.0927) meanQV: 1.4781 (1.4823) stdQV: 0.3240 (0.3236) time: 0.5222 data: 0.0003 max mem: 8962\n", - "Epoch: [5] [ 800/5005] eta: 0:35:30 lr: 0.0001 img/s: 493.49674301159627 loss: 0.8652 (0.9059) acc1: 70.3125 (68.9992) acc5: 87.5000 (87.0879) meanQV: 1.4918 (1.4824) stdQV: 0.3202 (0.3235) time: 0.5209 data: 0.0004 max mem: 8962\n", - "Epoch: [5] [ 900/5005] eta: 0:34:46 lr: 0.0001 img/s: 492.24203409939537 loss: 0.8904 (0.9059) acc1: 69.1406 (68.9659) acc5: 87.5000 (87.0920) meanQV: 1.4840 (1.4822) stdQV: 0.3224 (0.3236) time: 0.5210 data: 0.0003 max mem: 8962\n", - "Epoch: [5] [1000/5005] eta: 0:34:00 lr: 0.0001 img/s: 493.5423367172554 loss: 0.9307 (0.9063) acc1: 69.1406 (68.9541) acc5: 85.9375 (87.0731) meanQV: 1.4840 (1.4821) stdQV: 0.3228 (0.3236) time: 0.5217 data: 0.0004 max mem: 8962\n", - "Epoch: [5] [1100/5005] eta: 0:33:14 lr: 0.0001 img/s: 489.2419819520163 loss: 0.8769 (0.9064) acc1: 69.9219 (68.9540) acc5: 87.1094 (87.0718) meanQV: 1.4891 (1.4821) stdQV: 0.3212 (0.3236) time: 0.5218 data: 0.0003 max mem: 8962\n", - "Epoch: [5] [1200/5005] eta: 0:32:26 lr: 0.0001 img/s: 499.6392909558434 loss: 0.8655 (0.9056) acc1: 67.9688 (68.9608) acc5: 86.7188 (87.0716) meanQV: 1.4746 (1.4821) stdQV: 0.3257 (0.3236) time: 0.5207 data: 0.0003 max mem: 8962\n", - "Epoch: [5] [1300/5005] eta: 0:31:37 lr: 0.0001 img/s: 493.8163030755964 loss: 0.9314 (0.9051) acc1: 67.5781 (68.9860) acc5: 86.7188 (87.0968) meanQV: 1.4727 (1.4823) stdQV: 0.3280 (0.3235) time: 0.5206 data: 0.0004 max mem: 8962\n", - "Epoch: [5] [1400/5005] eta: 0:30:49 lr: 0.0001 img/s: 494.9975815812936 loss: 0.8642 (0.9041) acc1: 69.5312 (69.0129) acc5: 87.8906 (87.0929) meanQV: 1.4844 (1.4825) stdQV: 0.3216 (0.3234) time: 0.5214 data: 0.0003 max mem: 8962\n", - "Epoch: [5] [1500/5005] eta: 0:29:59 lr: 0.0001 img/s: 488.4738953923197 loss: 0.9535 (0.9047) acc1: 67.9688 (68.9798) acc5: 86.3281 (87.0698) meanQV: 1.4758 (1.4823) stdQV: 0.3262 (0.3235) time: 0.5208 data: 0.0004 max mem: 8962\n", - "Epoch: [5] [1600/5005] eta: 0:29:09 lr: 0.0001 img/s: 489.3487835801782 loss: 0.8809 (0.9042) acc1: 69.1406 (69.0108) acc5: 87.1094 (87.0811) meanQV: 1.4828 (1.4825) stdQV: 0.3237 (0.3235) time: 0.5214 data: 0.0004 max mem: 8962\n", - "Epoch: [5] [1700/5005] eta: 0:28:19 lr: 0.0001 img/s: 489.42685183665805 loss: 0.8938 (0.9043) acc1: 68.3594 (68.9845) acc5: 85.9375 (87.0756) meanQV: 1.4785 (1.4823) stdQV: 0.3262 (0.3235) time: 0.5206 data: 0.0003 max mem: 8962\n", - "Epoch: [5] [1800/5005] eta: 0:27:29 lr: 0.0001 img/s: 489.34900659686224 loss: 0.8771 (0.9038) acc1: 69.9219 (69.0081) acc5: 89.0625 (87.1057) meanQV: 1.4867 (1.4825) stdQV: 0.3192 (0.3235) time: 0.5219 data: 0.0004 max mem: 8962\n", - "Epoch: [5] [1900/5005] eta: 0:26:39 lr: 0.0001 img/s: 488.80233479720977 loss: 0.9229 (0.9041) acc1: 67.9688 (69.0192) acc5: 86.7188 (87.1049) meanQV: 1.4746 (1.4825) stdQV: 0.3236 (0.3234) time: 0.5216 data: 0.0004 max mem: 8962\n", - "Epoch: [5] [2000/5005] eta: 0:25:48 lr: 0.0001 img/s: 492.5014558430955 loss: 0.9266 (0.9045) acc1: 67.9688 (69.0048) acc5: 86.7188 (87.0990) meanQV: 1.4758 (1.4824) stdQV: 0.3262 (0.3235) time: 0.5205 data: 0.0004 max mem: 8962\n" - ] - }, - { - "ename": "KeyboardInterrupt", - "evalue": "", - "output_type": "error", - "traceback": [ - "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m", - "\u001b[0;31mKeyboardInterrupt\u001b[0m Traceback (most recent call last)", - "Cell \u001b[0;32mIn[8], line 27\u001b[0m\n\u001b[1;32m 1\u001b[0m \u001b[38;5;28;01mfrom\u001b[39;00m\u001b[38;5;250m \u001b[39m\u001b[38;5;21;01mtypes\u001b[39;00m\u001b[38;5;250m \u001b[39m\u001b[38;5;28;01mimport\u001b[39;00m SimpleNamespace\n\u001b[1;32m 3\u001b[0m args \u001b[38;5;241m=\u001b[39m SimpleNamespace(\n\u001b[1;32m 4\u001b[0m data_path\u001b[38;5;241m=\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m/home/cs/Documents/datasets/imagenet\u001b[39m\u001b[38;5;124m\"\u001b[39m,\n\u001b[1;32m 5\u001b[0m model\u001b[38;5;241m=\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mresnet18\u001b[39m\u001b[38;5;124m\"\u001b[39m,\n\u001b[0;32m (...)\u001b[0m\n\u001b[1;32m 24\u001b[0m trp_lambdas\u001b[38;5;241m=\u001b[39m[\u001b[38;5;241m0.4\u001b[39m, \u001b[38;5;241m0.2\u001b[39m, \u001b[38;5;241m0.1\u001b[39m],\n\u001b[1;32m 25\u001b[0m )\n\u001b[0;32m---> 27\u001b[0m \u001b[43mmain\u001b[49m\u001b[43m(\u001b[49m\u001b[43margs\u001b[49m\u001b[43m)\u001b[49m\n", - "Cell \u001b[0;32mIn[7], line 151\u001b[0m, in \u001b[0;36mmain\u001b[0;34m(args)\u001b[0m\n\u001b[1;32m 149\u001b[0m start_time \u001b[38;5;241m=\u001b[39m time\u001b[38;5;241m.\u001b[39mtime()\n\u001b[1;32m 150\u001b[0m \u001b[38;5;28;01mfor\u001b[39;00m epoch \u001b[38;5;129;01min\u001b[39;00m \u001b[38;5;28mrange\u001b[39m(args\u001b[38;5;241m.\u001b[39mepochs):\n\u001b[0;32m--> 151\u001b[0m \u001b[43mtrain_one_epoch\u001b[49m\u001b[43m(\u001b[49m\u001b[43mmodel\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43moptimizer\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mdata_loader\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mdevice\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mepoch\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43margs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m 152\u001b[0m lr_scheduler\u001b[38;5;241m.\u001b[39mstep()\n\u001b[1;32m 153\u001b[0m evaluate(model, nn\u001b[38;5;241m.\u001b[39mCrossEntropyLoss(), data_loader_test, device\u001b[38;5;241m=\u001b[39mdevice)\n", - "Cell \u001b[0;32mIn[7], line 61\u001b[0m, in \u001b[0;36mtrain_one_epoch\u001b[0;34m(model, optimizer, data_loader, device, epoch, args)\u001b[0m\n\u001b[1;32m 59\u001b[0m acc1, acc5 \u001b[38;5;241m=\u001b[39m utils\u001b[38;5;241m.\u001b[39maccuracy(output, target, topk\u001b[38;5;241m=\u001b[39m(\u001b[38;5;241m1\u001b[39m, \u001b[38;5;241m5\u001b[39m))\n\u001b[1;32m 60\u001b[0m batch_size \u001b[38;5;241m=\u001b[39m image\u001b[38;5;241m.\u001b[39mshape[\u001b[38;5;241m0\u001b[39m]\n\u001b[0;32m---> 61\u001b[0m metric_logger\u001b[38;5;241m.\u001b[39mupdate(loss\u001b[38;5;241m=\u001b[39m\u001b[43mloss\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mitem\u001b[49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m, lr\u001b[38;5;241m=\u001b[39moptimizer\u001b[38;5;241m.\u001b[39mparam_groups[\u001b[38;5;241m0\u001b[39m][\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mlr\u001b[39m\u001b[38;5;124m\"\u001b[39m])\n\u001b[1;32m 62\u001b[0m metric_logger\u001b[38;5;241m.\u001b[39mmeters[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124macc1\u001b[39m\u001b[38;5;124m\"\u001b[39m]\u001b[38;5;241m.\u001b[39mupdate(acc1\u001b[38;5;241m.\u001b[39mitem(), n\u001b[38;5;241m=\u001b[39mbatch_size)\n\u001b[1;32m 63\u001b[0m metric_logger\u001b[38;5;241m.\u001b[39mmeters[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124macc5\u001b[39m\u001b[38;5;124m\"\u001b[39m]\u001b[38;5;241m.\u001b[39mupdate(acc5\u001b[38;5;241m.\u001b[39mitem(), n\u001b[38;5;241m=\u001b[39mbatch_size)\n", - "\u001b[0;31mKeyboardInterrupt\u001b[0m: " - ] - } - ], - "source": [ - "from types import SimpleNamespace\n", - "\n", - "args = SimpleNamespace(\n", - " data_path=\"/home/cs/Documents/datasets/imagenet\",\n", - " model=\"resnet18\",\n", - " device=\"cuda\",\n", - " batch_size=256,\n", - " epochs=10,\n", - " opt=\"sgd\", # default unless overridden\n", - " lr=0.0004,\n", - " momentum=0.9,\n", - " weight_decay=1e-4,\n", - " lr_warmup_epochs=1,\n", - " lr_warmup_decay=0.0,\n", - " lr_step_size=2,\n", - " lr_gamma=0.5,\n", - " print_freq=100,\n", - " output_dir=\"resnet18\",\n", - " use_deterministic_algorithms=False,\n", - " weights=\"ResNet18_Weights.IMAGENET1K_V1\",\n", - " apply_trp=True,\n", - " trp_depths=[1, 1, 1],\n", - " trp_planes=256,\n", - " trp_lambdas=[0.4, 0.2, 0.1],\n", - ")\n", - "\n", - "main(args)" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "a3KD3WXU3l-O" - }, - "source": [ - "# Fine-tuning a language model" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "JAscNNUD3l-P" - }, - "source": [ - "In this notebook, we'll see how to fine-tune one of the [🤗 Transformers](https://github.com/huggingface/transformers) model on a language modeling tasks. We will cover two types of language modeling tasks which are:\n", - "\n", - "- Causal language modeling: the model has to predict the next token in the sentence (so the labels are the same as the inputs shifted to the right). To make sure the model does not cheat, it gets an attention mask that will prevent it to access the tokens after token i when trying to predict the token i+1 in the sentence.\n", - "\n", - "![Widget inference representing the causal language modeling task](https://github.com/huggingface/notebooks/blob/main/examples/images/causal_language_modeling.png?raw=1)\n", - "\n", - "- Masked language modeling: the model has to predict some tokens that are masked in the input. It still has access to the whole sentence, so it can use the tokens before and after the tokens masked to predict their value.\n", - "\n", - "![Widget inference representing the masked language modeling task](https://github.com/huggingface/notebooks/blob/main/examples/images/masked_language_modeling.png?raw=1)\n", - "\n", - "We will see how to easily load and preprocess the dataset for each one of those tasks, and how to use the `Trainer` API to fine-tune a model on it.\n", - "\n", - "A script version of this notebook you can directly run on a distributed environment or on TPU is available in our [examples folder](https://github.com/huggingface/transformers/tree/master/examples)." - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "1r_n9OWV3l-Q" - }, - "source": [ - "## Preparing the dataset" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "kswRMhPc3l-Q" - }, - "source": [ - "For each of those tasks, we will use the [Wikitext 2]() dataset as an example. You can load it very easily with the 🤗 Datasets library." - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "n2ZRs1cL3l-R" - }, - "outputs": [], - "source": [ - "from datasets import load_dataset\n", - "datasets = load_dataset('wikitext', 'wikitext-2-raw-v1')" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "f1-9jepM3l-W" - }, - "source": [ - "You can replace the dataset above with any dataset hosted on [the hub](https://huggingface.co/datasets) or use your own files. Just uncomment the following cell and replace the paths with values that will lead to your files:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "uxSaGa_l3l-W" - }, - "outputs": [], - "source": [ - "# datasets = load_dataset(\"text\", data_files={\"train\": path_to_train.txt, \"validation\": path_to_validation.txt}" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "jY1SwIrY3l-a" - }, - "source": [ - "You can also load datasets from a csv or a JSON file, see the [full documentation](https://huggingface.co/docs/datasets/loading_datasets.html#from-local-files) for more information." - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "u3EtYfeHIrIz" - }, - "source": [ - "To access an actual element, you need to select a split first, then give an index:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "X6HrpprwIrIz" - }, - "outputs": [], - "source": [ - "datasets[\"train\"][10]" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "WHUmphG3IrI3" - }, - "source": [ - "To get a sense of what the data looks like, the following function will show some examples picked randomly in the dataset." - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "ur5sNUcZ3l-g" - }, - "outputs": [], - "source": [ - "from datasets import ClassLabel\n", - "import random\n", - "import pandas as pd\n", - "from IPython.display import display, HTML\n", - "\n", - "def show_random_elements(dataset, num_examples=10):\n", - " assert num_examples <= len(dataset), \"Can't pick more elements than there are in the dataset.\"\n", - " picks = []\n", - " for _ in range(num_examples):\n", - " pick = random.randint(0, len(dataset)-1)\n", - " while pick in picks:\n", - " pick = random.randint(0, len(dataset)-1)\n", - " picks.append(pick)\n", - "\n", - " df = pd.DataFrame(dataset[picks])\n", - " for column, typ in dataset.features.items():\n", - " if isinstance(typ, ClassLabel):\n", - " df[column] = df[column].transform(lambda i: typ.names[i])\n", - " display(HTML(df.to_html()))" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "1Uk8NROQ3l-k" - }, - "outputs": [], - "source": [ - "show_random_elements(datasets[\"train\"])" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "CKerdF353l-o" - }, - "source": [ - "As we can see, some of the texts are a full paragraph of a Wikipedia article while others are just titles or empty lines." - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "JEA1ju653l-p" - }, - "source": [ - "## Causal Language modeling" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "v5GTGKZS3l-q" - }, - "source": [ - "For causal language modeling (CLM) we are going to take all the texts in our dataset and concatenate them after they are tokenized. Then we will split them in examples of a certain sequence length. This way the model will receive chunks of contiguous text that may look like:\n", - "```\n", - "part of text 1\n", - "```\n", - "or\n", - "```\n", - "end of text 1 [BOS_TOKEN] beginning of text 2\n", - "```\n", - "depending on whether they span over several of the original texts in the dataset or not. The labels will be the same as the inputs, shifted to the left.\n", - "\n", - "We will use the [`distilgpt2`](https://huggingface.co/distilgpt2) model for this example. You can pick any of the checkpoints listed [here](https://huggingface.co/models?filter=causal-lm) instead:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "-WGBCO343l-q" - }, - "outputs": [], - "source": [ - "model_checkpoint = \"distilgpt2\"" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "5io6fY_d3l-u" - }, - "source": [ - "To tokenize all our texts with the same vocabulary that was used when training the model, we have to download a pretrained tokenizer. This is all done by the `AutoTokenizer` class:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "iAYlS40Z3l-v" - }, - "outputs": [], - "source": [ - "from transformers import AutoTokenizer\n", - "\n", - "tokenizer = AutoTokenizer.from_pretrained(model_checkpoint, use_fast=True)" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "rpOiBrJ13l-y" - }, - "source": [ - "We can now call the tokenizer on all our texts. This is very simple, using the [`map`](https://huggingface.co/docs/datasets/package_reference/main_classes.html#datasets.Dataset.map) method from the Datasets library. First we define a function that call the tokenizer on our texts:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "lS2m25YM3l-z" - }, - "outputs": [], - "source": [ - "def tokenize_function(examples):\n", - " return tokenizer(examples[\"text\"])" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "M9xVAa3s3l-2" - }, - "source": [ - "Then we apply it to all the splits in our `datasets` object, using `batched=True` and 4 processes to speed up the preprocessing. We won't need the `text` column afterward, so we discard it." - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "NVAO0H8u3l-3" - }, - "outputs": [], - "source": [ - "tokenized_datasets = datasets.map(tokenize_function, batched=True, num_proc=4, remove_columns=[\"text\"])" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "8qik3J_C3l-7" - }, - "source": [ - "If we now look at an element of our datasets, we will see the text have been replaced by the `input_ids` the model will need:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "nYv_mcKk3l-7" - }, - "outputs": [], - "source": [ - "tokenized_datasets[\"train\"][1]" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "obvgcXda3l--" - }, - "source": [ - "Now for the harder part: we need to concatenate all our texts together then split the result in small chunks of a certain `block_size`. To do this, we will use the `map` method again, with the option `batched=True`. This option actually lets us change the number of examples in the datasets by returning a different number of examples than we got. This way, we can create our new samples from a batch of examples.\n", - "\n", - "First, we grab the maximum length our model was pretrained with. This might be a big too big to fit in your GPU RAM, so here we take a bit less at just 128." - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "DVHs5aCA3l-_" - }, - "outputs": [], - "source": [ - "# block_size = tokenizer.model_max_length\n", - "block_size = 128" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "RpNfGiMw3l_A" - }, - "source": [ - "Then we write the preprocessing function that will group our texts:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "iaAJy5Hu3l_B" - }, - "outputs": [], - "source": [ - "def group_texts(examples):\n", - " # Concatenate all texts.\n", - " concatenated_examples = {k: sum(examples[k], []) for k in examples.keys()}\n", - " total_length = len(concatenated_examples[list(examples.keys())[0]])\n", - " # We drop the small remainder, we could add padding if the model supported it instead of this drop, you can\n", - " # customize this part to your needs.\n", - " total_length = (total_length // block_size) * block_size\n", - " # Split by chunks of max_len.\n", - " result = {\n", - " k: [t[i : i + block_size] for i in range(0, total_length, block_size)]\n", - " for k, t in concatenated_examples.items()\n", - " }\n", - " result[\"labels\"] = result[\"input_ids\"].copy()\n", - " return result" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "LGJWXtNv3l_C" - }, - "source": [ - "First note that we duplicate the inputs for our labels. This is because the model of the 🤗 Transformers library apply the shifting to the right, so we don't need to do it manually.\n", - "\n", - "Also note that by default, the `map` method will send a batch of 1,000 examples to be treated by the preprocessing function. So here, we will drop the remainder to make the concatenated tokenized texts a multiple of `block_size` every 1,000 examples. You can adjust this behavior by passing a higher batch size (which will also be processed slower). You can also speed-up the preprocessing by using multiprocessing:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "gXUSfBrq3l_C" - }, - "outputs": [], - "source": [ - "lm_datasets = tokenized_datasets.map(\n", - " group_texts,\n", - " batched=True,\n", - " batch_size=1000,\n", - " num_proc=4,\n", - ")" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "6n84V8Gc3l_G" - }, - "source": [ - "And we can check our datasets have changed: now the samples contain chunks of `block_size` contiguous tokens, potentially spanning over several of our original texts." - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "hTeGCLl_3l_G" - }, - "outputs": [], - "source": [ - "tokenizer.decode(lm_datasets[\"train\"][1][\"input_ids\"])" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "iEmeQ7Xm3l_H" - }, - "source": [ - "Now that the data has been cleaned, we're ready to instantiate our `Trainer`. We will a model:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "sPqQA3TT3l_I" - }, - "outputs": [], - "source": [ - "from transformers import AutoModelForCausalLM\n", - "model = AutoModelForCausalLM.from_pretrained(model_checkpoint)" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "VyPQTOF_3l_J" - }, - "source": [ - "And some `TrainingArguments`:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "jElf8LJ33l_K" - }, - "outputs": [], - "source": [ - "from transformers import Trainer, TrainingArguments" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "YbSwEhQ63l_L" - }, - "outputs": [], - "source": [ - "model_name = model_checkpoint.split(\"/\")[-1]\n", - "training_args = TrainingArguments(\n", - " f\"{model_name}-finetuned-wikitext2\",\n", - " eval_strategy = \"epoch\",\n", - " learning_rate=2e-5,\n", - " weight_decay=0.01,\n", - " push_to_hub=True,\n", - ")" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "dsx-8ebng3FI" - }, - "source": [ - "The last argument to setup everything so we can push the model to the [Hub](https://huggingface.co/models) regularly during training. Remove it if you didn't follow the installation steps at the top of the notebook. If you want to save your model locally in a name that is different than the name of the repository it will be pushed, or if you want to push your model under an organization and not your name space, use the `hub_model_id` argument to set the repo name (it needs to be the full name, including your namespace: for instance `\"sgugger/gpt-finetuned-wikitext2\"` or `\"huggingface/gpt-finetuned-wikitext2\"`)." - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "sZRbT9ui3l_N" - }, - "source": [ - "We pass along all of those to the `Trainer` class:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "OEuqwIra3l_N" - }, - "outputs": [], - "source": [ - "trainer = Trainer(\n", - " model=model,\n", - " args=training_args,\n", - " train_dataset=lm_datasets[\"train\"],\n", - " eval_dataset=lm_datasets[\"validation\"],\n", - ")" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "6Vvz34Td3l_O" - }, - "source": [ - "And we can train our model:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "NyZvu_MF3l_P" - }, - "outputs": [], - "source": [ - "trainer.train()" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "3APq-vUc3l_R" - }, - "source": [ - "Once the training is completed, we can evaluate our model and get its perplexity on the validation set like this:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "diKZnB1I3l_R" - }, - "outputs": [], - "source": [ - "import math\n", - "eval_results = trainer.evaluate()\n", - "print(f\"Perplexity: {math.exp(eval_results['eval_loss']):.2f}\")" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "wY82caEX3l_i" - }, - "source": [ - "You can now upload the result of the training to the Hub, just execute this instruction:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "KjTnsZXwg3FI" - }, - "outputs": [], - "source": [ - "trainer.push_to_hub()" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "lKIc9Jr8g3FI" - }, - "source": [ - "You can now share this model with all your friends, family, favorite pets: they can all load it with the identifier `\"your-username/the-name-you-picked\"` so for instance:\n", - "\n", - "```python\n", - "from transformers import AutoModelForCausalLM\n", - "\n", - "model = AutoModelForCausalLM.from_pretrained(\"sgugger/my-awesome-model\")\n", - "```" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "q-EIELH43l_T" - }, - "source": [ - "## Masked language modeling" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "LWk97-Ny3l_T" - }, - "source": [ - "For masked language modeling (MLM) we are going to use the same preprocessing as before for our dataset with one additional step: we will randomly mask some tokens (by replacing them by `[MASK]`) and the labels will be adjusted to only include the masked tokens (we don't have to predict the non-masked tokens).\n", - "\n", - "We will use the [`distilroberta-base`](https://huggingface.co/distilroberta-base) model for this example. You can pick any of the checkpoints listed [here](https://huggingface.co/models?filter=masked-lm) instead:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "QRTpmyCc3l_T" - }, - "outputs": [], - "source": [ - "model_checkpoint = \"distilroberta-base\"" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "12F1ulgT3l_V" - }, - "source": [ - "We can apply the same tokenization function as before, we just need to update our tokenizer to use the checkpoint we just picked:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "h8RCYcvr3l_V" - }, - "outputs": [], - "source": [ - "tokenizer = AutoTokenizer.from_pretrained(model_checkpoint, use_fast=True)\n", - "tokenized_datasets = datasets.map(tokenize_function, batched=True, num_proc=4, remove_columns=[\"text\"])" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "MTuy8UUs3l_X" - }, - "source": [ - "And like before, we group texts together and chunk them in samples of length `block_size`. You can skip that step if your dataset is composed of individual sentences." - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "LVYPMwEs3l_X" - }, - "outputs": [], - "source": [ - "lm_datasets = tokenized_datasets.map(\n", - " group_texts,\n", - " batched=True,\n", - " batch_size=1000,\n", - " num_proc=4,\n", - ")" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "nFJ49iHJ3l_Z" - }, - "source": [ - "The rest is very similar to what we had, with two exceptions. First we use a model suitable for masked LM:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "PM10A9Za3l_Z" - }, - "outputs": [], - "source": [ - "from transformers import AutoModelForMaskedLM\n", - "model = AutoModelForMaskedLM.from_pretrained(model_checkpoint)" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "emlofW7Yg3FJ" - }, - "source": [ - "We redefine our `TrainingArguments`:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "Hgd9wpjHg3FJ" - }, - "outputs": [], - "source": [ - "model_name = model_checkpoint.split(\"/\")[-1]\n", - "training_args = TrainingArguments(\n", - " f\"{model_name}-finetuned-wikitext2\",\n", - " eval_strategy = \"epoch\",\n", - " learning_rate=2e-5,\n", - " weight_decay=0.01,\n", - " push_to_hub=True,\n", - ")" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "3tP4SZ3rg3FJ" - }, - "source": [ - "Like before, the last argument to setup everything so we can push the model to the [Hub](https://huggingface.co/models) regularly during training. Remove it if you didn't follow the installation steps at the top of the notebook. If you want to save your model locally in a name that is different than the name of the repository it will be pushed, or if you want to push your model under an organization and not your name space, use the `hub_model_id` argument to set the repo name (it needs to be the full name, including your namespace: for instance `\"sgugger/bert-finetuned-wikitext2\"` or `\"huggingface/bert-finetuned-wikitext2\"`)." - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "z6uuUnvz3l_b" - }, - "source": [ - "Finally, we use a special `data_collator`. The `data_collator` is a function that is responsible of taking the samples and batching them in tensors. In the previous example, we had nothing special to do, so we just used the default for this argument. Here we want to do the random-masking. We could do it as a pre-processing step (like the tokenization) but then the tokens would always be masked the same way at each epoch. By doing this step inside the `data_collator`, we ensure this random masking is done in a new way each time we go over the data.\n", - "\n", - "To do this masking for us, the library provides a `DataCollatorForLanguageModeling`. We can adjust the probability of the masking:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "nRZ-5v_P3l_b" - }, - "outputs": [], - "source": [ - "from transformers import DataCollatorForLanguageModeling\n", - "data_collator = DataCollatorForLanguageModeling(tokenizer=tokenizer, mlm_probability=0.15)" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "bqHnWcYC3l_d" - }, - "source": [ - "Then we just have to pass everything to `Trainer` and begin training:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "V-Y3gNqV3l_d" - }, - "outputs": [], - "source": [ - "trainer = Trainer(\n", - " model=model,\n", - " args=training_args,\n", - " train_dataset=lm_datasets[\"train\"],\n", - " eval_dataset=lm_datasets[\"validation\"],\n", - " data_collator=data_collator,\n", - ")" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "Y9TFqDG_3l_e" - }, - "outputs": [], - "source": [ - "trainer.train()" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "KDBi0reX3l_g" - }, - "source": [ - "Like before, we can evaluate our model on the validation set. The perplexity is much lower than for the CLM objective because for the MLM objective, we only have to make predictions for the masked tokens (which represent 15% of the total here) while having access to the rest of the tokens. It's thus an easier task for the model." - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "4hSaANqj3l_g" - }, - "outputs": [], - "source": [ - "eval_results = trainer.evaluate()\n", - "print(f\"Perplexity: {math.exp(eval_results['eval_loss']):.2f}\")" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "TDmITtEYg3FK" - }, - "source": [ - "You can now upload the result of the training to the Hub, just execute this instruction:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "-3CFIYUrg3FK" - }, - "outputs": [], - "source": [ - "trainer.push_to_hub()" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "Da10yeZNg3FK" - }, - "source": [ - "You can now share this model with all your friends, family, favorite pets: they can all load it with the identifier `\"your-username/the-name-you-picked\"` so for instance:\n", - "\n", - "```python\n", - "from transformers import AutoModelForMaskedLM\n", - "\n", - "model = AutoModelForMaskedLM.from_pretrained(\"sgugger/my-awesome-model\")\n", - "```" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "RqqSgjeSg3FK" - }, - "outputs": [], - "source": [] - } - ], - "metadata": { - "colab": { - "name": "Fine-tune a language model", - "provenance": [] - }, - "kernelspec": { - "display_name": "Python 3 (ipykernel)", - "language": "python", - "name": "python3" - }, - "language_info": { - "codemirror_mode": { - "name": "ipython", - "version": 3 - }, - "file_extension": ".py", - "mimetype": "text/x-python", - "name": "python", - "nbconvert_exporter": "python", - "pygments_lexer": "ipython3", - "version": "3.9.21" - } - }, - "nbformat": 4, - "nbformat_minor": 0 -}