├── distbelief ├── __init__.py ├── utils │ ├── __init__.py │ ├── serialization.py │ └── messaging.py ├── optim │ ├── __init__.py │ └── downpour_sgd.py └── server.py ├── docs ├── no_min_lr │ ├── server.log │ ├── first.log │ └── second.log ├── diagram.jpg ├── test_time.png ├── train_time.png └── experiment │ ├── process_0.png │ ├── process_1.png │ └── process_2.png ├── requirements-dev.txt ├── requirements.txt ├── .gitignore ├── setup.py ├── Makefile ├── example ├── graph.py ├── models.py └── main.py ├── README.md └── LICENSE.md /distbelief/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /distbelief/utils/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /docs/no_min_lr/server.log: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /distbelief/optim/__init__.py: -------------------------------------------------------------------------------- 1 | from .downpour_sgd import DownpourSGD 2 | -------------------------------------------------------------------------------- /requirements-dev.txt: -------------------------------------------------------------------------------- 1 | twine==1.11.0 2 | setuptools==40.2.0 3 | wheel==0.31.0 4 | -------------------------------------------------------------------------------- /docs/diagram.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ucla-labx/distbelief/HEAD/docs/diagram.jpg -------------------------------------------------------------------------------- /docs/test_time.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ucla-labx/distbelief/HEAD/docs/test_time.png -------------------------------------------------------------------------------- /docs/train_time.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ucla-labx/distbelief/HEAD/docs/train_time.png -------------------------------------------------------------------------------- /docs/experiment/process_0.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ucla-labx/distbelief/HEAD/docs/experiment/process_0.png -------------------------------------------------------------------------------- /docs/experiment/process_1.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ucla-labx/distbelief/HEAD/docs/experiment/process_1.png -------------------------------------------------------------------------------- /docs/experiment/process_2.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ucla-labx/distbelief/HEAD/docs/experiment/process_2.png -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- 1 | numpy==1.14.5 2 | six==1.11.0 3 | torch==0.4.0 4 | torchvision==0.2.1 5 | matplotlib 6 | pandas 7 | sklearn 8 | scipy 9 | -------------------------------------------------------------------------------- /.gitignore: -------------------------------------------------------------------------------- 1 | data 2 | __pycache__/ 3 | .pyc* 4 | train.log 5 | env/ 6 | venv/ 7 | .idea/ 8 | .ipynb_checkpoints/* 9 | log/* 10 | build/* 11 | dist/* 12 | pytorch_distbelief.egg-info/* 13 | -------------------------------------------------------------------------------- /setup.py: -------------------------------------------------------------------------------- 1 | import setuptools 2 | 3 | with open("README.md", "r") as fh: 4 | long_description = fh.read() 5 | 6 | setuptools.setup( 7 | name="pytorch-distbelief", 8 | version="0.1.0", 9 | author="Jesse Cai", 10 | author_email="jcjessecai@gmail.com", 11 | description="Distributed training for pytorch", 12 | long_description=long_description, 13 | long_description_content_type="text/markdown", 14 | url="https://github.com/ucla-labx/distbelief", 15 | packages=setuptools.find_packages(), 16 | classifiers=( 17 | "Programming Language :: Python :: 3", 18 | "License :: OSI Approved :: MIT License", 19 | "Operating System :: OS Independent", 20 | ), 21 | ) 22 | -------------------------------------------------------------------------------- /Makefile: -------------------------------------------------------------------------------- 1 | setup: 2 | -sudo apt-get -y virtualenv 3 | virtualenv -p python3 venv 4 | . venv/bin/activate && pip install -r requirements.txt && pip install . 5 | 6 | install: 7 | pip install . 8 | 9 | graph: 10 | python example/graph.py 11 | mv train_time.png test_time.png docs 12 | 13 | first: 14 | python example/main.py --rank 1 --world-size 3 15 | 16 | second: 17 | python example/main.py --rank 2 --world-size 3 18 | 19 | server: 20 | python example/main.py --rank 0 --world-size 3 --server 21 | 22 | single: 23 | python example/main.py --no-distributed 24 | 25 | gpu: 26 | python example/main.py --no-distributed --cuda 27 | 28 | dist: 29 | python3 setup.py sdist bdist_wheel 30 | 31 | upload: dist 32 | twine upload dist/* 33 | 34 | upload-test: dist 35 | twine upload --repository-url https://test.pypi.org/legacy/ dist/* 36 | 37 | install-test: 38 | python3 -m pip install --index-url https://test.pypi.org/simple/ pytorch-distbelief 39 | 40 | -------------------------------------------------------------------------------- /distbelief/utils/serialization.py: -------------------------------------------------------------------------------- 1 | import torch 2 | 3 | 4 | def ravel_model_params(model, grads=False): 5 | """ 6 | Squash model parameters or gradients into a single tensor. 7 | """ 8 | m_parameter = torch.Tensor([0]) 9 | for parameter in list(model.parameters()): 10 | if grads: 11 | m_parameter = torch.cat((m_parameter, parameter.grad.view(-1))) 12 | else: 13 | m_parameter = torch.cat((m_parameter, parameter.data.view(-1))) 14 | return m_parameter[1:] 15 | 16 | 17 | def unravel_model_params(model, parameter_update): 18 | """ 19 | Assigns parameter_update params to model.parameters. 20 | This is done by iterating through model.parameters() and assigning the relevant params in parameter_update. 21 | NOTE: this function manipulates model.parameters. 22 | """ 23 | current_index = 0 # keep track of where to read from parameter_update 24 | for parameter in model.parameters(): 25 | numel = parameter.data.numel() 26 | size = parameter.data.size() 27 | parameter.data.copy_(parameter_update[current_index:current_index+numel].view(size)) 28 | current_index += numel 29 | 30 | -------------------------------------------------------------------------------- /distbelief/server.py: -------------------------------------------------------------------------------- 1 | # 2 | """ 3 | Parameter server for distbelief 4 | """ 5 | import logging 6 | import torch 7 | import torch.optim 8 | from distbelief.utils.messaging import MessageCode, MessageListener, send_message 9 | from distbelief.utils.serialization import ravel_model_params, unravel_model_params 10 | 11 | _LOGGER = logging.getLogger(__name__) 12 | 13 | class ParameterServer(MessageListener): 14 | """ParameterServer""" 15 | def __init__(self, model): 16 | _LOGGER.info("Creating ParameterServer") 17 | self.parameter_shard = torch.rand(ravel_model_params(model).numel()) 18 | self.model = model 19 | #init superclass 20 | super().__init__(model) 21 | 22 | def receive(self, sender, message_code, parameter): 23 | print("Processing message: {} from sender {}".format(message_code.name, sender)) 24 | 25 | if message_code == MessageCode.ParameterUpdate: 26 | #be sure to clone here 27 | self.parameter_shard = parameter.clone() 28 | 29 | elif message_code == MessageCode.ParameterRequest: 30 | send_message(MessageCode.ParameterUpdate, self.parameter_shard, dst=sender) 31 | 32 | elif message_code == MessageCode.GradientUpdate: 33 | self.parameter_shard.add_(parameter) 34 | -------------------------------------------------------------------------------- /example/graph.py: -------------------------------------------------------------------------------- 1 | """ 2 | plots accuracy (test and train) vs. time 3 | """ 4 | import matplotlib as mpl 5 | mpl.use('TkAgg') 6 | 7 | import matplotlib.pyplot as plt 8 | import pandas as pd 9 | 10 | colors = ['blue', 'green', 'red', 'orange', 'magenta'] 11 | files_to_read = ['log/single.csv', 'log/gpu.csv', 'log/node1.csv', 'log/node2.csv', 'log/node3.csv'] 12 | log_dataframes = list(map(pd.read_csv, files_to_read)) 13 | 14 | for df in log_dataframes: 15 | df['timestamp'] = pd.to_datetime(df['timestamp']) 16 | df['timestamp'] -= df['timestamp'].min() 17 | 18 | 19 | def plot_train(df, label, color): 20 | plt.plot(df['timestamp'].dt.seconds / 3600.0, 21 | df['training_accuracy'].rolling(50).mean(), 22 | label=label, 23 | color=color) 24 | 25 | def plot_test(df, label, color): 26 | plt.plot(df.dropna()['timestamp'].dt.seconds / 3600.0, 27 | df.dropna()['test_accuracy'].rolling(5).mean(), 28 | label=label, 29 | color=color) 30 | 31 | 32 | fig1 = plt.figure(figsize=(20, 10)) 33 | 34 | for color, filename, df in zip(colors, files_to_read, log_dataframes): 35 | plot_train(df, filename, color) 36 | 37 | plt.ylabel('Training Accuracy') 38 | plt.xlabel('Time (hours)') 39 | plt.legend() 40 | plt.title("Training Accuracy vs. Time (50 iteration rolling average, freq: 3, lr: 0.1)") 41 | plt.savefig('train_time.png') 42 | 43 | fig = plt.figure(figsize=(20, 10)) 44 | 45 | for color, filename, df in zip(colors, files_to_read, log_dataframes): 46 | plot_test(df, filename, color) 47 | 48 | plt.ylabel('Test Accuracy') 49 | plt.xlabel('Time (hours)') 50 | plt.legend() 51 | plt.title("Test Accuracy vs. Time (5 iteration rolling average, freq: 3, lr: 0.1)") 52 | plt.savefig('test_time.png') 53 | -------------------------------------------------------------------------------- /example/models.py: -------------------------------------------------------------------------------- 1 | import torch 2 | import torch.nn as nn 3 | import torch.nn.functional as F 4 | 5 | class LeNet(nn.Module): 6 | def __init__(self): 7 | super(LeNet, self).__init__() 8 | self.conv1 = nn.Conv2d(3, 6, kernel_size=5) 9 | self.conv2 = nn.Conv2d(6, 16, kernel_size=5) 10 | self.conv2_drop = nn.Dropout2d() 11 | self.fc1 = nn.Linear(16 * 5 * 5, 120) 12 | self.fc2 = nn.Linear(120, 84) 13 | self.fc3 = nn.Linear(84, 10) 14 | 15 | def forward(self, x): 16 | x = F.relu(F.max_pool2d(self.conv1(x), 2)) 17 | x = F.relu(F.max_pool2d(self.conv2_drop(self.conv2(x)), 2)) 18 | x = x.view(-1, 16* 5* 5) 19 | x = F.relu(self.fc1(x)) 20 | x = F.dropout(x, training=self.training) 21 | x = F.relu(self.fc2(x)) 22 | x = self.fc3(x) 23 | return x 24 | 25 | class AlexNet(nn.Module): 26 | def __init__(self, num_classes=10): 27 | super(AlexNet, self).__init__() 28 | self.features = nn.Sequential( 29 | nn.Conv2d(3, 64, kernel_size=11, stride=4, padding=5), 30 | nn.ReLU(inplace=True), 31 | nn.MaxPool2d(kernel_size=2, stride=2), 32 | nn.Conv2d(64, 192, kernel_size=5, padding=2), 33 | nn.ReLU(inplace=True), 34 | nn.MaxPool2d(kernel_size=2, stride=2), 35 | nn.Conv2d(192, 384, kernel_size=3, padding=1), 36 | nn.ReLU(inplace=True), 37 | nn.Conv2d(384, 256, kernel_size=3, padding=1), 38 | nn.ReLU(inplace=True), 39 | nn.Conv2d(256, 256, kernel_size=3, padding=1), 40 | nn.ReLU(inplace=True), 41 | nn.MaxPool2d(kernel_size=2, stride=2), 42 | ) 43 | self.classifier = nn.Linear(256, num_classes) 44 | 45 | def forward(self, x): 46 | x = self.features(x) 47 | x = x.view(x.size(0), -1) 48 | x = self.classifier(x) 49 | return x 50 | -------------------------------------------------------------------------------- /distbelief/utils/messaging.py: -------------------------------------------------------------------------------- 1 | from enum import Enum 2 | import logging 3 | import torch 4 | import torch.distributed as dist 5 | from threading import Thread 6 | from distbelief.utils.serialization import ravel_model_params 7 | 8 | _LOGGER = logging.getLogger(__name__) 9 | 10 | 11 | class MessageCode(Enum): 12 | """Different types of messages between client and server that we support go here.""" 13 | ParameterRequest = 0 14 | GradientUpdate = 1 15 | ParameterUpdate = 2 16 | EvaluateParams = 3 17 | 18 | 19 | class MessageListener(Thread): 20 | """MessageListener 21 | 22 | base class for message listeners, extends pythons threading Thread 23 | """ 24 | def __init__(self, model): 25 | """__init__ 26 | 27 | :param model: nn.Module to be defined by the user 28 | """ 29 | self.model = model 30 | _LOGGER.info("Setting m_parameter") 31 | self.m_parameter = torch.zeros(ravel_model_params(model).numel() + 2) 32 | super(MessageListener, self).__init__() 33 | 34 | def receive(self, sender, message_code, parameter): 35 | """receive 36 | 37 | :param sender: rank id of the sender 38 | :param message_code: Enum code 39 | :param parameter: the data payload 40 | """ 41 | raise NotImplementedError() 42 | 43 | def run(self): 44 | _LOGGER.info("Started Running!") 45 | self.running = True 46 | while self.running: 47 | _LOGGER.info("Polling for message...") 48 | dist.recv(tensor=self.m_parameter) 49 | self.receive(int(self.m_parameter[0].item()), 50 | MessageCode(self.m_parameter[1].item()), 51 | self.m_parameter[2:]) 52 | 53 | 54 | def send_message(message_code, payload, dst=0): 55 | """Sends a message to a destination 56 | Concatenates both the message code and destination with the payload into a single tensor and then sends that as a tensor 57 | """ 58 | _LOGGER.info("SENDING MESSAGE: {} RANK: {}".format(message_code, dist.get_rank())) 59 | m_parameter = torch.Tensor([dist.get_rank(), message_code.value]) 60 | m_parameter = torch.cat((m_parameter, payload)) 61 | dist.isend(tensor=m_parameter, dst=dst) 62 | 63 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # distbelief 2 | Implementing Google's DistBelief paper. 3 | 4 | Check out the [blog post](https://jcaip.github.io/Distbelief/)! 5 | ## Installation/Development instructions 6 | 7 | To install the latest stable version (pytorch-distbelief 0.1.0), run `pip install pytorch-distbelief` 8 | 9 | Otherwise, you can build and run the latest master with the instructions below. 10 | 11 | You'll want to create a python3 virtualenv first by running `make setup`, after which, you should run `make install`. 12 | 13 | You'll then be able to use distbelief by importing distbelief 14 | ```python 15 | 16 | from distbelief.optim import DownpourSGD 17 | 18 | optimizer = DownpourSGD(net.parameters(), lr=0.1, n_push=5, n_pull=5, model=net) 19 | 20 | ``` 21 | 22 | As an example, you can see our implementation running by using the script provided in `example/main.py`. 23 | 24 | To run a 2-training node setup locally, open up three terminal windows, source the `venv` and then run `make first`, `make second`, and `make server`. 25 | This will begin training AlexNet on CIFAR10 locally with all default params. 26 | 27 | ## Benchmarking 28 | 29 | **NOTE:** we graph the train/test accuracy of each node, hence node1, node2, node3. A better comparison would be to evaluate the parameter server's params and use that value. 30 | However we can see that the accuracy between the three nodes is fairly consistent, and adding an evaluator might put too much stress on our server. 31 | 32 | We scale the learning rate of the nodes to be learning_rate/freq (.03) . 33 | 34 | ![train](/docs/train_time.png) 35 | 36 | ![test](/docs/test_time.png) 37 | 38 | We used AWS c4.xlarge instances to compare the CPU runs, and a GTX 1060 for the GPU run. 39 | 40 | ## DownpourSGD for PyTorch 41 | 42 | ### Diagram 43 | 44 | 45 | 46 | Here **2** and **3** happen concurrently. 47 | 48 | You can read more about our implementation [here](https://jcaip.github.io/Distbelief/). 49 | 50 | ### References 51 | - [Pytorch distributed tutorial](http://pytorch.org/tutorials/intermediate/dist_tuto.html) 52 | - [Akka implementation of distbelief](http://alexminnaar.com/implementing-the-distbelief-deep-neural-network-training-framework-with-akka.html) 53 | - [gevent actor tutorial](http://sdiehl.github.io/gevent-tutorial/#actors) 54 | - [DistBelief paper](https://static.googleusercontent.com/media/research.google.com/en//archive/large_deep_networks_nips2012.pdf) 55 | - [Analysis of delayed grad problem](https://openreview.net/pdf?id=BJLSGcywG) 56 | -------------------------------------------------------------------------------- /distbelief/optim/downpour_sgd.py: -------------------------------------------------------------------------------- 1 | import logging 2 | import torch 3 | from torch.optim.optimizer import Optimizer, required 4 | from distbelief.utils.serialization import ravel_model_params, unravel_model_params 5 | from distbelief.utils.messaging import MessageCode, MessageListener, send_message 6 | 7 | _LOGGER = logging.getLogger(__name__) 8 | 9 | class DownpourListener(MessageListener): 10 | """DownpourListener""" 11 | def __init__(self, model): 12 | super().__init__(model) 13 | 14 | def receive(self, sender, message_code, parameter): 15 | """receive parameter updates from the server and reflect them into the client's model.""" 16 | _LOGGER.info("Processing message: {}".format(message_code.name)) 17 | if message_code == MessageCode.ParameterUpdate: 18 | unravel_model_params(self.model, parameter) 19 | 20 | class DownpourSGD(Optimizer): 21 | """DownpourSGD""" 22 | 23 | def __init__(self, params, lr=required, n_push=required, n_pull=required, model=required): 24 | """__init__ 25 | 26 | :param params: 27 | :param lr: 28 | :param freq: 29 | :param model: 30 | """ 31 | if lr is not required and lr < 0.0: 32 | raise ValueError("Invalid learning rate: {}".format(lr)) 33 | 34 | defaults = dict(lr=lr,) 35 | self.accumulated_gradients = torch.zeros(ravel_model_params(model).size()) 36 | self.n_pull = n_pull 37 | self.n_push = n_push 38 | 39 | self.model = model 40 | # this sets the initial model parameters 41 | send_message(MessageCode.ParameterUpdate, ravel_model_params(self.model)) 42 | self.idx = 0 43 | 44 | listener = DownpourListener(self.model) 45 | listener.start() 46 | 47 | super(DownpourSGD, self).__init__(params, defaults) 48 | 49 | def step(self, closure=None): 50 | """Performs a single optimization step. 51 | 52 | Arguments: 53 | closure (callable, optional): A closure that reevaluates the model 54 | and returns the loss. 55 | """ 56 | loss = None 57 | if closure is not None: 58 | loss = closure() 59 | 60 | # send parameter request every N iterations 61 | if self.idx % self.n_pull == 0: 62 | send_message(MessageCode.ParameterRequest, self.accumulated_gradients) # dummy val 63 | 64 | #get the lr 65 | lr = self.param_groups[0]['lr'] 66 | # keep track of accumulated gradients so that we can send 67 | gradients = ravel_model_params(self.model, grads=True) 68 | self.accumulated_gradients.add_(-lr, gradients) 69 | 70 | # send gradient update every N iterations 71 | if self.idx % self.n_push == 0: 72 | send_message(MessageCode.GradientUpdate, self.accumulated_gradients) # send gradients to the server 73 | self.accumulated_gradients.zero_() 74 | 75 | # internal sgd update 76 | for group in self.param_groups: 77 | for p in group['params']: 78 | if p.grad is None: 79 | continue 80 | d_p = p.grad.data 81 | p.data.add_(-group['lr'], d_p) 82 | 83 | self.idx += 1 84 | return loss 85 | -------------------------------------------------------------------------------- /example/main.py: -------------------------------------------------------------------------------- 1 | import os 2 | import logging 3 | import argparse 4 | import csv 5 | import torch 6 | import torchvision 7 | import torchvision.transforms as transforms 8 | import numpy as np 9 | import torch.nn as nn 10 | import torch.nn.functional as F 11 | import torch.distributed as dist 12 | 13 | from datetime import datetime 14 | from models import LeNet, AlexNet 15 | from sklearn.metrics import classification_report, accuracy_score, confusion_matrix 16 | import pandas as pd 17 | 18 | import torch.optim as optim 19 | from distbelief.optim import DownpourSGD 20 | from distbelief.server import ParameterServer 21 | 22 | def get_dataset(args, transform): 23 | """ 24 | :param dataset_name: 25 | :param transform: 26 | :param batch_size: 27 | :return: iterators for the dataset 28 | """ 29 | if args.dataset == 'MNIST': 30 | trainset = torchvision.datasets.MNIST(root='./data', train=True, download=True, transform=transform) 31 | testset = torchvision.datasets.MNIST(root='./data', train=False, download=True, transform=transform) 32 | else: 33 | trainset = torchvision.datasets.CIFAR10(root='./data', train=True, download=True, transform=transform) 34 | testset = torchvision.datasets.CIFAR10(root='./data', train=False, download=True, transform=transform) 35 | 36 | trainloader = torch.utils.data.DataLoader(trainset, batch_size=args.batch_size, shuffle=True, num_workers=1) 37 | testloader = torch.utils.data.DataLoader(testset, batch_size=args.test_batch_size, shuffle=False, num_workers=1) 38 | return trainloader, testloader 39 | 40 | def main(args): 41 | 42 | logs = [] 43 | 44 | transform = transforms.Compose([ 45 | transforms.ToTensor(), 46 | transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)) 47 | ]) 48 | 49 | trainloader, testloader = get_dataset(args, transform) 50 | net = AlexNet() 51 | 52 | if args.no_distributed: 53 | optimizer = optim.SGD(net.parameters(), lr=args.lr, momentum=0.0) 54 | else: 55 | optimizer = DownpourSGD(net.parameters(), lr=args.lr, n_push=args.num_push, n_pull=args.num_pull, model=net) 56 | scheduler = optim.lr_scheduler.ReduceLROnPlateau(optimizer, patience=1, verbose=True, min_lr=1e-3) 57 | 58 | # train 59 | net.train() 60 | if args.cuda: 61 | net = net.cuda() 62 | 63 | for epoch in range(args.epochs): # loop over the dataset multiple times 64 | print("Training for epoch {}".format(epoch)) 65 | for i, data in enumerate(trainloader, 0): 66 | # get the inputs 67 | inputs, labels = data 68 | 69 | if args.cuda: 70 | inputs, labels = inputs.cuda(), labels.cuda() 71 | 72 | # zero the parameter gradients 73 | optimizer.zero_grad() 74 | # forward + backward + optimize 75 | outputs = net(inputs) 76 | loss = F.cross_entropy(outputs, labels) 77 | loss.backward() 78 | optimizer.step() 79 | 80 | _, predicted = torch.max(outputs, 1) 81 | accuracy = accuracy_score(predicted, labels) 82 | 83 | log_obj = { 84 | 'timestamp': datetime.now(), 85 | 'iteration': i, 86 | 'training_loss': loss.item(), 87 | 'training_accuracy': accuracy, 88 | } 89 | 90 | if i % args.log_interval == 0 and i > 0: # print every n mini-batches 91 | log_obj['test_loss'], log_obj['test_accuracy']= evaluate( net, testloader, args) 92 | print("Timestamp: {timestamp} | " 93 | "Iteration: {iteration:6} | " 94 | "Loss: {training_loss:6.4f} | " 95 | "Accuracy : {training_accuracy:6.4f} | " 96 | "Test Loss: {test_loss:6.4f} | " 97 | "Test Accuracy: {test_accuracy:6.4f}".format(**log_obj)) 98 | 99 | logs.append(log_obj) 100 | 101 | val_loss, val_accuracy = evaluate(net, testloader, args, verbose=True) 102 | scheduler.step(val_loss) 103 | 104 | df = pd.DataFrame(logs) 105 | print(df) 106 | if args.no_distributed: 107 | if args.cuda: 108 | df.to_csv('log/gpu.csv', index_label='index') 109 | else: 110 | df.to_csv('log/single.csv', index_label='index') 111 | else: 112 | df.to_csv('log/node{}.csv'.format(dist.get_rank()), index_label='index') 113 | 114 | print('Finished Training') 115 | 116 | 117 | def evaluate(net, testloader, args, verbose=False): 118 | if args.dataset == 'MNIST': 119 | classes = [str(i) for i in range(10)] 120 | else: 121 | classes = ('plane', 'car', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck') 122 | net.eval() 123 | 124 | test_loss = 0 125 | with torch.no_grad(): 126 | for data in testloader: 127 | images, labels = data 128 | 129 | if args.cuda: 130 | images, labels = images.cuda(), labels.cuda() 131 | 132 | outputs = net(images) 133 | _, predicted = torch.max(outputs, 1) 134 | test_loss += F.cross_entropy(outputs, labels).item() 135 | 136 | test_accuracy = accuracy_score(predicted, labels) 137 | if verbose: 138 | print('Loss: {:.3f}'.format(test_loss)) 139 | print('Accuracy: {:.3f}'.format(test_accuracy)) 140 | print(classification_report(predicted, labels, target_names=classes)) 141 | 142 | return test_loss, test_accuracy 143 | 144 | def init_server(): 145 | model = AlexNet() 146 | server = ParameterServer(model=model) 147 | server.run() 148 | 149 | if __name__ == "__main__": 150 | parser = argparse.ArgumentParser(description='Distbelief training example') 151 | parser.add_argument('--batch-size', type=int, default=64, metavar='N', help='input batch size for training (default: 64)') 152 | parser.add_argument('--test-batch-size', type=int, default=10000, metavar='N', help='input batch size for testing (default: 10000)') 153 | parser.add_argument('--epochs', type=int, default=20, metavar='N', help='number of epochs to train (default: 20)') 154 | parser.add_argument('--lr', type=float, default=0.003, metavar='LR', help='learning rate (default: 0.1)') 155 | parser.add_argument('--num-pull', type=int, default=5, metavar='N', help='how often to pull params (default: 5)') 156 | parser.add_argument('--num-push', type=int, default=5, metavar='N', help='how often to push grads (default: 5)') 157 | parser.add_argument('--cuda', action='store_true', default=False, help='use CUDA for training') 158 | parser.add_argument('--log-interval', type=int, default=20, metavar='N', help='how often to evaluate and print out') 159 | parser.add_argument('--no-distributed', action='store_true', default=False, help='whether to use DownpourSGD or normal SGD') 160 | parser.add_argument('--rank', type=int, metavar='N', help='rank of current process (0 is server, 1+ is training node)') 161 | parser.add_argument('--world-size', type=int, default=3, metavar='N', help='size of the world') 162 | parser.add_argument('--server', action='store_true', default=False, help='server node?') 163 | parser.add_argument('--dataset', type=str, default='CIFAR10', help='which dataset to train on') 164 | parser.add_argument('--master', type=str, default='localhost', help='ip address of the master (server) node') 165 | parser.add_argument('--port', type=str, default='29500', help='port on master node to communicate with') 166 | args = parser.parse_args() 167 | print(args) 168 | 169 | if not args.no_distributed: 170 | """ Initialize the distributed environment. 171 | Server and clients must call this as an entry point. 172 | """ 173 | os.environ['MASTER_ADDR'] = args.master 174 | os.environ['MASTER_PORT'] = args.port 175 | dist.init_process_group('tcp', rank=args.rank, world_size=args.world_size) 176 | if args.server: 177 | init_server() 178 | main(args) 179 | -------------------------------------------------------------------------------- /LICENSE.md: -------------------------------------------------------------------------------- 1 | GNU GENERAL PUBLIC LICENSE 2 | Version 3, 29 June 2007 3 | 4 | Copyright (C) 2007 Free Software Foundation, Inc. 5 | Everyone is permitted to copy and distribute verbatim copies 6 | of this license document, but changing it is not allowed. 7 | 8 | Preamble 9 | 10 | The GNU General Public License is a free, copyleft license for 11 | software and other kinds of works. 12 | 13 | The licenses for most software and other practical works are designed 14 | to take away your freedom to share and change the works. By contrast, 15 | the GNU General Public License is intended to guarantee your freedom to 16 | share and change all versions of a program--to make sure it remains free 17 | software for all its users. We, the Free Software Foundation, use the 18 | GNU General Public License for most of our software; it applies also to 19 | any other work released this way by its authors. You can apply it to 20 | your programs, too. 21 | 22 | When we speak of free software, we are referring to freedom, not 23 | price. Our General Public Licenses are designed to make sure that you 24 | have the freedom to distribute copies of free software (and charge for 25 | them if you wish), that you receive source code or can get it if you 26 | want it, that you can change the software or use pieces of it in new 27 | free programs, and that you know you can do these things. 28 | 29 | To protect your rights, we need to prevent others from denying you 30 | these rights or asking you to surrender the rights. Therefore, you have 31 | certain responsibilities if you distribute copies of the software, or if 32 | you modify it: responsibilities to respect the freedom of others. 33 | 34 | For example, if you distribute copies of such a program, whether 35 | gratis or for a fee, you must pass on to the recipients the same 36 | freedoms that you received. You must make sure that they, too, receive 37 | or can get the source code. And you must show them these terms so they 38 | know their rights. 39 | 40 | Developers that use the GNU GPL protect your rights with two steps: 41 | (1) assert copyright on the software, and (2) offer you this License 42 | giving you legal permission to copy, distribute and/or modify it. 43 | 44 | For the developers' and authors' protection, the GPL clearly explains 45 | that there is no warranty for this free software. For both users' and 46 | authors' sake, the GPL requires that modified versions be marked as 47 | changed, so that their problems will not be attributed erroneously to 48 | authors of previous versions. 49 | 50 | Some devices are designed to deny users access to install or run 51 | modified versions of the software inside them, although the manufacturer 52 | can do so. This is fundamentally incompatible with the aim of 53 | protecting users' freedom to change the software. The systematic 54 | pattern of such abuse occurs in the area of products for individuals to 55 | use, which is precisely where it is most unacceptable. Therefore, we 56 | have designed this version of the GPL to prohibit the practice for those 57 | products. If such problems arise substantially in other domains, we 58 | stand ready to extend this provision to those domains in future versions 59 | of the GPL, as needed to protect the freedom of users. 60 | 61 | Finally, every program is threatened constantly by software patents. 62 | States should not allow patents to restrict development and use of 63 | software on general-purpose computers, but in those that do, we wish to 64 | avoid the special danger that patents applied to a free program could 65 | make it effectively proprietary. To prevent this, the GPL assures that 66 | patents cannot be used to render the program non-free. 67 | 68 | The precise terms and conditions for copying, distribution and 69 | modification follow. 70 | 71 | TERMS AND CONDITIONS 72 | 73 | 0. Definitions. 74 | 75 | "This License" refers to version 3 of the GNU General Public License. 76 | 77 | "Copyright" also means copyright-like laws that apply to other kinds of 78 | works, such as semiconductor masks. 79 | 80 | "The Program" refers to any copyrightable work licensed under this 81 | License. Each licensee is addressed as "you". "Licensees" and 82 | "recipients" may be individuals or organizations. 83 | 84 | To "modify" a work means to copy from or adapt all or part of the work 85 | in a fashion requiring copyright permission, other than the making of an 86 | exact copy. The resulting work is called a "modified version" of the 87 | earlier work or a work "based on" the earlier work. 88 | 89 | A "covered work" means either the unmodified Program or a work based 90 | on the Program. 91 | 92 | To "propagate" a work means to do anything with it that, without 93 | permission, would make you directly or secondarily liable for 94 | infringement under applicable copyright law, except executing it on a 95 | computer or modifying a private copy. Propagation includes copying, 96 | distribution (with or without modification), making available to the 97 | public, and in some countries other activities as well. 98 | 99 | To "convey" a work means any kind of propagation that enables other 100 | parties to make or receive copies. Mere interaction with a user through 101 | a computer network, with no transfer of a copy, is not conveying. 102 | 103 | An interactive user interface displays "Appropriate Legal Notices" 104 | to the extent that it includes a convenient and prominently visible 105 | feature that (1) displays an appropriate copyright notice, and (2) 106 | tells the user that there is no warranty for the work (except to the 107 | extent that warranties are provided), that licensees may convey the 108 | work under this License, and how to view a copy of this License. If 109 | the interface presents a list of user commands or options, such as a 110 | menu, a prominent item in the list meets this criterion. 111 | 112 | 1. Source Code. 113 | 114 | The "source code" for a work means the preferred form of the work 115 | for making modifications to it. "Object code" means any non-source 116 | form of a work. 117 | 118 | A "Standard Interface" means an interface that either is an official 119 | standard defined by a recognized standards body, or, in the case of 120 | interfaces specified for a particular programming language, one that 121 | is widely used among developers working in that language. 122 | 123 | The "System Libraries" of an executable work include anything, other 124 | than the work as a whole, that (a) is included in the normal form of 125 | packaging a Major Component, but which is not part of that Major 126 | Component, and (b) serves only to enable use of the work with that 127 | Major Component, or to implement a Standard Interface for which an 128 | implementation is available to the public in source code form. A 129 | "Major Component", in this context, means a major essential component 130 | (kernel, window system, and so on) of the specific operating system 131 | (if any) on which the executable work runs, or a compiler used to 132 | produce the work, or an object code interpreter used to run it. 133 | 134 | The "Corresponding Source" for a work in object code form means all 135 | the source code needed to generate, install, and (for an executable 136 | work) run the object code and to modify the work, including scripts to 137 | control those activities. However, it does not include the work's 138 | System Libraries, or general-purpose tools or generally available free 139 | programs which are used unmodified in performing those activities but 140 | which are not part of the work. For example, Corresponding Source 141 | includes interface definition files associated with source files for 142 | the work, and the source code for shared libraries and dynamically 143 | linked subprograms that the work is specifically designed to require, 144 | such as by intimate data communication or control flow between those 145 | subprograms and other parts of the work. 146 | 147 | The Corresponding Source need not include anything that users 148 | can regenerate automatically from other parts of the Corresponding 149 | Source. 150 | 151 | The Corresponding Source for a work in source code form is that 152 | same work. 153 | 154 | 2. Basic Permissions. 155 | 156 | All rights granted under this License are granted for the term of 157 | copyright on the Program, and are irrevocable provided the stated 158 | conditions are met. This License explicitly affirms your unlimited 159 | permission to run the unmodified Program. The output from running a 160 | covered work is covered by this License only if the output, given its 161 | content, constitutes a covered work. This License acknowledges your 162 | rights of fair use or other equivalent, as provided by copyright law. 163 | 164 | You may make, run and propagate covered works that you do not 165 | convey, without conditions so long as your license otherwise remains 166 | in force. You may convey covered works to others for the sole purpose 167 | of having them make modifications exclusively for you, or provide you 168 | with facilities for running those works, provided that you comply with 169 | the terms of this License in conveying all material for which you do 170 | not control copyright. Those thus making or running the covered works 171 | for you must do so exclusively on your behalf, under your direction 172 | and control, on terms that prohibit them from making any copies of 173 | your copyrighted material outside their relationship with you. 174 | 175 | Conveying under any other circumstances is permitted solely under 176 | the conditions stated below. Sublicensing is not allowed; section 10 177 | makes it unnecessary. 178 | 179 | 3. Protecting Users' Legal Rights From Anti-Circumvention Law. 180 | 181 | No covered work shall be deemed part of an effective technological 182 | measure under any applicable law fulfilling obligations under article 183 | 11 of the WIPO copyright treaty adopted on 20 December 1996, or 184 | similar laws prohibiting or restricting circumvention of such 185 | measures. 186 | 187 | When you convey a covered work, you waive any legal power to forbid 188 | circumvention of technological measures to the extent such circumvention 189 | is effected by exercising rights under this License with respect to 190 | the covered work, and you disclaim any intention to limit operation or 191 | modification of the work as a means of enforcing, against the work's 192 | users, your or third parties' legal rights to forbid circumvention of 193 | technological measures. 194 | 195 | 4. Conveying Verbatim Copies. 196 | 197 | You may convey verbatim copies of the Program's source code as you 198 | receive it, in any medium, provided that you conspicuously and 199 | appropriately publish on each copy an appropriate copyright notice; 200 | keep intact all notices stating that this License and any 201 | non-permissive terms added in accord with section 7 apply to the code; 202 | keep intact all notices of the absence of any warranty; and give all 203 | recipients a copy of this License along with the Program. 204 | 205 | You may charge any price or no price for each copy that you convey, 206 | and you may offer support or warranty protection for a fee. 207 | 208 | 5. Conveying Modified Source Versions. 209 | 210 | You may convey a work based on the Program, or the modifications to 211 | produce it from the Program, in the form of source code under the 212 | terms of section 4, provided that you also meet all of these conditions: 213 | 214 | a) The work must carry prominent notices stating that you modified 215 | it, and giving a relevant date. 216 | 217 | b) The work must carry prominent notices stating that it is 218 | released under this License and any conditions added under section 219 | 7. This requirement modifies the requirement in section 4 to 220 | "keep intact all notices". 221 | 222 | c) You must license the entire work, as a whole, under this 223 | License to anyone who comes into possession of a copy. This 224 | License will therefore apply, along with any applicable section 7 225 | additional terms, to the whole of the work, and all its parts, 226 | regardless of how they are packaged. This License gives no 227 | permission to license the work in any other way, but it does not 228 | invalidate such permission if you have separately received it. 229 | 230 | d) If the work has interactive user interfaces, each must display 231 | Appropriate Legal Notices; however, if the Program has interactive 232 | interfaces that do not display Appropriate Legal Notices, your 233 | work need not make them do so. 234 | 235 | A compilation of a covered work with other separate and independent 236 | works, which are not by their nature extensions of the covered work, 237 | and which are not combined with it such as to form a larger program, 238 | in or on a volume of a storage or distribution medium, is called an 239 | "aggregate" if the compilation and its resulting copyright are not 240 | used to limit the access or legal rights of the compilation's users 241 | beyond what the individual works permit. Inclusion of a covered work 242 | in an aggregate does not cause this License to apply to the other 243 | parts of the aggregate. 244 | 245 | 6. Conveying Non-Source Forms. 246 | 247 | You may convey a covered work in object code form under the terms 248 | of sections 4 and 5, provided that you also convey the 249 | machine-readable Corresponding Source under the terms of this License, 250 | in one of these ways: 251 | 252 | a) Convey the object code in, or embodied in, a physical product 253 | (including a physical distribution medium), accompanied by the 254 | Corresponding Source fixed on a durable physical medium 255 | customarily used for software interchange. 256 | 257 | b) Convey the object code in, or embodied in, a physical product 258 | (including a physical distribution medium), accompanied by a 259 | written offer, valid for at least three years and valid for as 260 | long as you offer spare parts or customer support for that product 261 | model, to give anyone who possesses the object code either (1) a 262 | copy of the Corresponding Source for all the software in the 263 | product that is covered by this License, on a durable physical 264 | medium customarily used for software interchange, for a price no 265 | more than your reasonable cost of physically performing this 266 | conveying of source, or (2) access to copy the 267 | Corresponding Source from a network server at no charge. 268 | 269 | c) Convey individual copies of the object code with a copy of the 270 | written offer to provide the Corresponding Source. This 271 | alternative is allowed only occasionally and noncommercially, and 272 | only if you received the object code with such an offer, in accord 273 | with subsection 6b. 274 | 275 | d) Convey the object code by offering access from a designated 276 | place (gratis or for a charge), and offer equivalent access to the 277 | Corresponding Source in the same way through the same place at no 278 | further charge. You need not require recipients to copy the 279 | Corresponding Source along with the object code. If the place to 280 | copy the object code is a network server, the Corresponding Source 281 | may be on a different server (operated by you or a third party) 282 | that supports equivalent copying facilities, provided you maintain 283 | clear directions next to the object code saying where to find the 284 | Corresponding Source. Regardless of what server hosts the 285 | Corresponding Source, you remain obligated to ensure that it is 286 | available for as long as needed to satisfy these requirements. 287 | 288 | e) Convey the object code using peer-to-peer transmission, provided 289 | you inform other peers where the object code and Corresponding 290 | Source of the work are being offered to the general public at no 291 | charge under subsection 6d. 292 | 293 | A separable portion of the object code, whose source code is excluded 294 | from the Corresponding Source as a System Library, need not be 295 | included in conveying the object code work. 296 | 297 | A "User Product" is either (1) a "consumer product", which means any 298 | tangible personal property which is normally used for personal, family, 299 | or household purposes, or (2) anything designed or sold for incorporation 300 | into a dwelling. In determining whether a product is a consumer product, 301 | doubtful cases shall be resolved in favor of coverage. For a particular 302 | product received by a particular user, "normally used" refers to a 303 | typical or common use of that class of product, regardless of the status 304 | of the particular user or of the way in which the particular user 305 | actually uses, or expects or is expected to use, the product. A product 306 | is a consumer product regardless of whether the product has substantial 307 | commercial, industrial or non-consumer uses, unless such uses represent 308 | the only significant mode of use of the product. 309 | 310 | "Installation Information" for a User Product means any methods, 311 | procedures, authorization keys, or other information required to install 312 | and execute modified versions of a covered work in that User Product from 313 | a modified version of its Corresponding Source. The information must 314 | suffice to ensure that the continued functioning of the modified object 315 | code is in no case prevented or interfered with solely because 316 | modification has been made. 317 | 318 | If you convey an object code work under this section in, or with, or 319 | specifically for use in, a User Product, and the conveying occurs as 320 | part of a transaction in which the right of possession and use of the 321 | User Product is transferred to the recipient in perpetuity or for a 322 | fixed term (regardless of how the transaction is characterized), the 323 | Corresponding Source conveyed under this section must be accompanied 324 | by the Installation Information. But this requirement does not apply 325 | if neither you nor any third party retains the ability to install 326 | modified object code on the User Product (for example, the work has 327 | been installed in ROM). 328 | 329 | The requirement to provide Installation Information does not include a 330 | requirement to continue to provide support service, warranty, or updates 331 | for a work that has been modified or installed by the recipient, or for 332 | the User Product in which it has been modified or installed. Access to a 333 | network may be denied when the modification itself materially and 334 | adversely affects the operation of the network or violates the rules and 335 | protocols for communication across the network. 336 | 337 | Corresponding Source conveyed, and Installation Information provided, 338 | in accord with this section must be in a format that is publicly 339 | documented (and with an implementation available to the public in 340 | source code form), and must require no special password or key for 341 | unpacking, reading or copying. 342 | 343 | 7. Additional Terms. 344 | 345 | "Additional permissions" are terms that supplement the terms of this 346 | License by making exceptions from one or more of its conditions. 347 | Additional permissions that are applicable to the entire Program shall 348 | be treated as though they were included in this License, to the extent 349 | that they are valid under applicable law. If additional permissions 350 | apply only to part of the Program, that part may be used separately 351 | under those permissions, but the entire Program remains governed by 352 | this License without regard to the additional permissions. 353 | 354 | When you convey a copy of a covered work, you may at your option 355 | remove any additional permissions from that copy, or from any part of 356 | it. (Additional permissions may be written to require their own 357 | removal in certain cases when you modify the work.) You may place 358 | additional permissions on material, added by you to a covered work, 359 | for which you have or can give appropriate copyright permission. 360 | 361 | Notwithstanding any other provision of this License, for material you 362 | add to a covered work, you may (if authorized by the copyright holders of 363 | that material) supplement the terms of this License with terms: 364 | 365 | a) Disclaiming warranty or limiting liability differently from the 366 | terms of sections 15 and 16 of this License; or 367 | 368 | b) Requiring preservation of specified reasonable legal notices or 369 | author attributions in that material or in the Appropriate Legal 370 | Notices displayed by works containing it; or 371 | 372 | c) Prohibiting misrepresentation of the origin of that material, or 373 | requiring that modified versions of such material be marked in 374 | reasonable ways as different from the original version; or 375 | 376 | d) Limiting the use for publicity purposes of names of licensors or 377 | authors of the material; or 378 | 379 | e) Declining to grant rights under trademark law for use of some 380 | trade names, trademarks, or service marks; or 381 | 382 | f) Requiring indemnification of licensors and authors of that 383 | material by anyone who conveys the material (or modified versions of 384 | it) with contractual assumptions of liability to the recipient, for 385 | any liability that these contractual assumptions directly impose on 386 | those licensors and authors. 387 | 388 | All other non-permissive additional terms are considered "further 389 | restrictions" within the meaning of section 10. If the Program as you 390 | received it, or any part of it, contains a notice stating that it is 391 | governed by this License along with a term that is a further 392 | restriction, you may remove that term. If a license document contains 393 | a further restriction but permits relicensing or conveying under this 394 | License, you may add to a covered work material governed by the terms 395 | of that license document, provided that the further restriction does 396 | not survive such relicensing or conveying. 397 | 398 | If you add terms to a covered work in accord with this section, you 399 | must place, in the relevant source files, a statement of the 400 | additional terms that apply to those files, or a notice indicating 401 | where to find the applicable terms. 402 | 403 | Additional terms, permissive or non-permissive, may be stated in the 404 | form of a separately written license, or stated as exceptions; 405 | the above requirements apply either way. 406 | 407 | 8. Termination. 408 | 409 | You may not propagate or modify a covered work except as expressly 410 | provided under this License. Any attempt otherwise to propagate or 411 | modify it is void, and will automatically terminate your rights under 412 | this License (including any patent licenses granted under the third 413 | paragraph of section 11). 414 | 415 | However, if you cease all violation of this License, then your 416 | license from a particular copyright holder is reinstated (a) 417 | provisionally, unless and until the copyright holder explicitly and 418 | finally terminates your license, and (b) permanently, if the copyright 419 | holder fails to notify you of the violation by some reasonable means 420 | prior to 60 days after the cessation. 421 | 422 | Moreover, your license from a particular copyright holder is 423 | reinstated permanently if the copyright holder notifies you of the 424 | violation by some reasonable means, this is the first time you have 425 | received notice of violation of this License (for any work) from that 426 | copyright holder, and you cure the violation prior to 30 days after 427 | your receipt of the notice. 428 | 429 | Termination of your rights under this section does not terminate the 430 | licenses of parties who have received copies or rights from you under 431 | this License. If your rights have been terminated and not permanently 432 | reinstated, you do not qualify to receive new licenses for the same 433 | material under section 10. 434 | 435 | 9. Acceptance Not Required for Having Copies. 436 | 437 | You are not required to accept this License in order to receive or 438 | run a copy of the Program. Ancillary propagation of a covered work 439 | occurring solely as a consequence of using peer-to-peer transmission 440 | to receive a copy likewise does not require acceptance. However, 441 | nothing other than this License grants you permission to propagate or 442 | modify any covered work. These actions infringe copyright if you do 443 | not accept this License. Therefore, by modifying or propagating a 444 | covered work, you indicate your acceptance of this License to do so. 445 | 446 | 10. Automatic Licensing of Downstream Recipients. 447 | 448 | Each time you convey a covered work, the recipient automatically 449 | receives a license from the original licensors, to run, modify and 450 | propagate that work, subject to this License. You are not responsible 451 | for enforcing compliance by third parties with this License. 452 | 453 | An "entity transaction" is a transaction transferring control of an 454 | organization, or substantially all assets of one, or subdividing an 455 | organization, or merging organizations. If propagation of a covered 456 | work results from an entity transaction, each party to that 457 | transaction who receives a copy of the work also receives whatever 458 | licenses to the work the party's predecessor in interest had or could 459 | give under the previous paragraph, plus a right to possession of the 460 | Corresponding Source of the work from the predecessor in interest, if 461 | the predecessor has it or can get it with reasonable efforts. 462 | 463 | You may not impose any further restrictions on the exercise of the 464 | rights granted or affirmed under this License. For example, you may 465 | not impose a license fee, royalty, or other charge for exercise of 466 | rights granted under this License, and you may not initiate litigation 467 | (including a cross-claim or counterclaim in a lawsuit) alleging that 468 | any patent claim is infringed by making, using, selling, offering for 469 | sale, or importing the Program or any portion of it. 470 | 471 | 11. Patents. 472 | 473 | A "contributor" is a copyright holder who authorizes use under this 474 | License of the Program or a work on which the Program is based. The 475 | work thus licensed is called the contributor's "contributor version". 476 | 477 | A contributor's "essential patent claims" are all patent claims 478 | owned or controlled by the contributor, whether already acquired or 479 | hereafter acquired, that would be infringed by some manner, permitted 480 | by this License, of making, using, or selling its contributor version, 481 | but do not include claims that would be infringed only as a 482 | consequence of further modification of the contributor version. For 483 | purposes of this definition, "control" includes the right to grant 484 | patent sublicenses in a manner consistent with the requirements of 485 | this License. 486 | 487 | Each contributor grants you a non-exclusive, worldwide, royalty-free 488 | patent license under the contributor's essential patent claims, to 489 | make, use, sell, offer for sale, import and otherwise run, modify and 490 | propagate the contents of its contributor version. 491 | 492 | In the following three paragraphs, a "patent license" is any express 493 | agreement or commitment, however denominated, not to enforce a patent 494 | (such as an express permission to practice a patent or covenant not to 495 | sue for patent infringement). To "grant" such a patent license to a 496 | party means to make such an agreement or commitment not to enforce a 497 | patent against the party. 498 | 499 | If you convey a covered work, knowingly relying on a patent license, 500 | and the Corresponding Source of the work is not available for anyone 501 | to copy, free of charge and under the terms of this License, through a 502 | publicly available network server or other readily accessible means, 503 | then you must either (1) cause the Corresponding Source to be so 504 | available, or (2) arrange to deprive yourself of the benefit of the 505 | patent license for this particular work, or (3) arrange, in a manner 506 | consistent with the requirements of this License, to extend the patent 507 | license to downstream recipients. "Knowingly relying" means you have 508 | actual knowledge that, but for the patent license, your conveying the 509 | covered work in a country, or your recipient's use of the covered work 510 | in a country, would infringe one or more identifiable patents in that 511 | country that you have reason to believe are valid. 512 | 513 | If, pursuant to or in connection with a single transaction or 514 | arrangement, you convey, or propagate by procuring conveyance of, a 515 | covered work, and grant a patent license to some of the parties 516 | receiving the covered work authorizing them to use, propagate, modify 517 | or convey a specific copy of the covered work, then the patent license 518 | you grant is automatically extended to all recipients of the covered 519 | work and works based on it. 520 | 521 | A patent license is "discriminatory" if it does not include within 522 | the scope of its coverage, prohibits the exercise of, or is 523 | conditioned on the non-exercise of one or more of the rights that are 524 | specifically granted under this License. You may not convey a covered 525 | work if you are a party to an arrangement with a third party that is 526 | in the business of distributing software, under which you make payment 527 | to the third party based on the extent of your activity of conveying 528 | the work, and under which the third party grants, to any of the 529 | parties who would receive the covered work from you, a discriminatory 530 | patent license (a) in connection with copies of the covered work 531 | conveyed by you (or copies made from those copies), or (b) primarily 532 | for and in connection with specific products or compilations that 533 | contain the covered work, unless you entered into that arrangement, 534 | or that patent license was granted, prior to 28 March 2007. 535 | 536 | Nothing in this License shall be construed as excluding or limiting 537 | any implied license or other defenses to infringement that may 538 | otherwise be available to you under applicable patent law. 539 | 540 | 12. No Surrender of Others' Freedom. 541 | 542 | If conditions are imposed on you (whether by court order, agreement or 543 | otherwise) that contradict the conditions of this License, they do not 544 | excuse you from the conditions of this License. If you cannot convey a 545 | covered work so as to satisfy simultaneously your obligations under this 546 | License and any other pertinent obligations, then as a consequence you may 547 | not convey it at all. For example, if you agree to terms that obligate you 548 | to collect a royalty for further conveying from those to whom you convey 549 | the Program, the only way you could satisfy both those terms and this 550 | License would be to refrain entirely from conveying the Program. 551 | 552 | 13. Use with the GNU Affero General Public License. 553 | 554 | Notwithstanding any other provision of this License, you have 555 | permission to link or combine any covered work with a work licensed 556 | under version 3 of the GNU Affero General Public License into a single 557 | combined work, and to convey the resulting work. The terms of this 558 | License will continue to apply to the part which is the covered work, 559 | but the special requirements of the GNU Affero General Public License, 560 | section 13, concerning interaction through a network will apply to the 561 | combination as such. 562 | 563 | 14. Revised Versions of this License. 564 | 565 | The Free Software Foundation may publish revised and/or new versions of 566 | the GNU General Public License from time to time. Such new versions will 567 | be similar in spirit to the present version, but may differ in detail to 568 | address new problems or concerns. 569 | 570 | Each version is given a distinguishing version number. If the 571 | Program specifies that a certain numbered version of the GNU General 572 | Public License "or any later version" applies to it, you have the 573 | option of following the terms and conditions either of that numbered 574 | version or of any later version published by the Free Software 575 | Foundation. If the Program does not specify a version number of the 576 | GNU General Public License, you may choose any version ever published 577 | by the Free Software Foundation. 578 | 579 | If the Program specifies that a proxy can decide which future 580 | versions of the GNU General Public License can be used, that proxy's 581 | public statement of acceptance of a version permanently authorizes you 582 | to choose that version for the Program. 583 | 584 | Later license versions may give you additional or different 585 | permissions. However, no additional obligations are imposed on any 586 | author or copyright holder as a result of your choosing to follow a 587 | later version. 588 | 589 | 15. Disclaimer of Warranty. 590 | 591 | THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY 592 | APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT 593 | HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY 594 | OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, 595 | THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 596 | PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM 597 | IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF 598 | ALL NECESSARY SERVICING, REPAIR OR CORRECTION. 599 | 600 | 16. Limitation of Liability. 601 | 602 | IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING 603 | WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS 604 | THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY 605 | GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE 606 | USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF 607 | DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD 608 | PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), 609 | EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF 610 | SUCH DAMAGES. 611 | 612 | 17. Interpretation of Sections 15 and 16. 613 | 614 | If the disclaimer of warranty and limitation of liability provided 615 | above cannot be given local legal effect according to their terms, 616 | reviewing courts shall apply local law that most closely approximates 617 | an absolute waiver of all civil liability in connection with the 618 | Program, unless a warranty or assumption of liability accompanies a 619 | copy of the Program in return for a fee. 620 | 621 | END OF TERMS AND CONDITIONS 622 | 623 | How to Apply These Terms to Your New Programs 624 | 625 | If you develop a new program, and you want it to be of the greatest 626 | possible use to the public, the best way to achieve this is to make it 627 | free software which everyone can redistribute and change under these terms. 628 | 629 | To do so, attach the following notices to the program. It is safest 630 | to attach them to the start of each source file to most effectively 631 | state the exclusion of warranty; and each file should have at least 632 | the "copyright" line and a pointer to where the full notice is found. 633 | 634 | 635 | Copyright (C) 636 | 637 | This program is free software: you can redistribute it and/or modify 638 | it under the terms of the GNU General Public License as published by 639 | the Free Software Foundation, either version 3 of the License, or 640 | (at your option) any later version. 641 | 642 | This program is distributed in the hope that it will be useful, 643 | but WITHOUT ANY WARRANTY; without even the implied warranty of 644 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 645 | GNU General Public License for more details. 646 | 647 | You should have received a copy of the GNU General Public License 648 | along with this program. If not, see . 649 | 650 | Also add information on how to contact you by electronic and paper mail. 651 | 652 | If the program does terminal interaction, make it output a short 653 | notice like this when it starts in an interactive mode: 654 | 655 | Copyright (C) 656 | This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. 657 | This is free software, and you are welcome to redistribute it 658 | under certain conditions; type `show c' for details. 659 | 660 | The hypothetical commands `show w' and `show c' should show the appropriate 661 | parts of the General Public License. Of course, your program's commands 662 | might be different; for a GUI interface, you would use an "about box". 663 | 664 | You should also get your employer (if you work as a programmer) or school, 665 | if any, to sign a "copyright disclaimer" for the program, if necessary. 666 | For more information on this, and how to apply and follow the GNU GPL, see 667 | . 668 | 669 | The GNU General Public License does not permit incorporating your program 670 | into proprietary programs. If your program is a subroutine library, you 671 | may consider it more useful to permit linking proprietary applications with 672 | the library. If this is what you want to do, use the GNU Lesser General 673 | Public License instead of this License. But first, please read 674 | . 675 | -------------------------------------------------------------------------------- /docs/no_min_lr/first.log: -------------------------------------------------------------------------------- 1 | Files already downloaded and verified 2 | Files already downloaded and verified 3 | Epoch: 0, Iteration: 0 loss: 0.115 4 | Epoch: 0, Iteration: 20 loss: 2.303 5 | Epoch: 0, Iteration: 40 loss: 2.302 6 | Epoch: 0, Iteration: 60 loss: 2.302 7 | Epoch: 0, Iteration: 80 loss: 2.301 8 | Epoch: 0, Iteration: 100 loss: 2.301 9 | Epoch: 0, Iteration: 120 loss: 2.300 10 | Epoch: 0, Iteration: 140 loss: 2.298 11 | Epoch: 0, Iteration: 160 loss: 2.290 12 | Epoch: 0, Iteration: 180 loss: 2.263 13 | Epoch: 0, Iteration: 200 loss: 2.177 14 | Epoch: 0, Iteration: 220 loss: 2.158 15 | Epoch: 0, Iteration: 240 loss: 2.173 16 | Epoch: 0, Iteration: 260 loss: 2.147 17 | Epoch: 0, Iteration: 280 loss: 2.099 18 | Epoch: 0, Iteration: 300 loss: 2.102 19 | Epoch: 0, Iteration: 320 loss: 2.174 20 | Epoch: 0, Iteration: 340 loss: 2.217 21 | Epoch: 0, Iteration: 360 loss: 2.149 22 | Epoch: 0, Iteration: 380 loss: 2.102 23 | Epoch: 0, Iteration: 400 loss: 2.096 24 | Epoch: 0, Iteration: 420 loss: 2.212 25 | Epoch: 0, Iteration: 440 loss: 2.194 26 | Epoch: 0, Iteration: 460 loss: 2.116 27 | Epoch: 0, Iteration: 480 loss: 2.128 28 | Epoch: 0, Iteration: 500 loss: 2.103 29 | Epoch: 0, Iteration: 520 loss: 2.151 30 | Epoch: 0, Iteration: 540 loss: 2.161 31 | Epoch: 0, Iteration: 560 loss: 2.059 32 | Epoch: 0, Iteration: 580 loss: 2.080 33 | Epoch: 0, Iteration: 600 loss: 2.178 34 | Epoch: 0, Iteration: 620 loss: 2.140 35 | Epoch: 0, Iteration: 640 loss: 2.012 36 | Epoch: 0, Iteration: 660 loss: 2.040 37 | Epoch: 0, Iteration: 680 loss: 1.978 38 | Epoch: 0, Iteration: 700 loss: 2.053 39 | Epoch: 0, Iteration: 720 loss: 2.068 40 | Epoch: 0, Iteration: 740 loss: 2.082 41 | Epoch: 0, Iteration: 760 loss: 2.138 42 | Epoch: 0, Iteration: 780 loss: 1.964 43 | Loss: 2.256 44 | Accuracy of the network on the 10000 test images: 22 % 45 | Accuracy of plane : 37 % 46 | Accuracy of car : 0 % 47 | Accuracy of bird : 24 % 48 | Accuracy of cat : 0 % 49 | Accuracy of deer : 0 % 50 | Accuracy of dog : 79 % 51 | Accuracy of frog : 0 % 52 | Accuracy of horse : 31 % 53 | Accuracy of ship : 41 % 54 | Accuracy of truck : 16 % 55 | Epoch: 1, Iteration: 0 loss: 0.113 56 | Epoch: 1, Iteration: 20 loss: 1.982 57 | Epoch: 1, Iteration: 40 loss: 1.984 58 | Epoch: 1, Iteration: 60 loss: 2.033 59 | Epoch: 1, Iteration: 80 loss: 2.000 60 | Epoch: 1, Iteration: 100 loss: 1.986 61 | Epoch: 1, Iteration: 120 loss: 1.988 62 | Epoch: 1, Iteration: 140 loss: 2.003 63 | Epoch: 1, Iteration: 160 loss: 1.976 64 | Epoch: 1, Iteration: 180 loss: 1.894 65 | Epoch: 1, Iteration: 200 loss: 1.952 66 | Epoch: 1, Iteration: 220 loss: 1.953 67 | Epoch: 1, Iteration: 240 loss: 1.972 68 | Epoch: 1, Iteration: 260 loss: 1.927 69 | Epoch: 1, Iteration: 280 loss: 1.937 70 | Epoch: 1, Iteration: 300 loss: 1.919 71 | Epoch: 1, Iteration: 320 loss: 1.935 72 | Epoch: 1, Iteration: 340 loss: 1.918 73 | Epoch: 1, Iteration: 360 loss: 1.879 74 | Epoch: 1, Iteration: 380 loss: 1.897 75 | Epoch: 1, Iteration: 400 loss: 1.944 76 | Epoch: 1, Iteration: 420 loss: 1.929 77 | Epoch: 1, Iteration: 440 loss: 1.913 78 | Epoch: 1, Iteration: 460 loss: 1.943 79 | Epoch: 1, Iteration: 480 loss: 1.937 80 | Epoch: 1, Iteration: 500 loss: 1.896 81 | Epoch: 1, Iteration: 520 loss: 1.858 82 | Epoch: 1, Iteration: 540 loss: 1.808 83 | Epoch: 1, Iteration: 560 loss: 1.883 84 | Epoch: 1, Iteration: 580 loss: 1.853 85 | Epoch: 1, Iteration: 600 loss: 1.849 86 | Epoch: 1, Iteration: 620 loss: 1.845 87 | Epoch: 1, Iteration: 640 loss: 1.933 88 | Epoch: 1, Iteration: 660 loss: 1.794 89 | Epoch: 1, Iteration: 680 loss: 1.871 90 | Epoch: 1, Iteration: 700 loss: 1.837 91 | Epoch: 1, Iteration: 720 loss: 1.822 92 | Epoch: 1, Iteration: 740 loss: 1.820 93 | Epoch: 1, Iteration: 760 loss: 1.752 94 | Epoch: 1, Iteration: 780 loss: 1.907 95 | Loss: 1.773 96 | Accuracy of the network on the 10000 test images: 32 % 97 | Accuracy of plane : 8 % 98 | Accuracy of car : 72 % 99 | Accuracy of bird : 0 % 100 | Accuracy of cat : 45 % 101 | Accuracy of deer : 16 % 102 | Accuracy of dog : 35 % 103 | Accuracy of frog : 14 % 104 | Accuracy of horse : 48 % 105 | Accuracy of ship : 74 % 106 | Accuracy of truck : 21 % 107 | Epoch: 2, Iteration: 0 loss: 0.092 108 | Epoch: 2, Iteration: 20 loss: 1.783 109 | Epoch: 2, Iteration: 40 loss: 1.830 110 | Epoch: 2, Iteration: 60 loss: 1.820 111 | Epoch: 2, Iteration: 80 loss: 1.793 112 | Epoch: 2, Iteration: 100 loss: 1.822 113 | Epoch: 2, Iteration: 120 loss: 1.797 114 | Epoch: 2, Iteration: 140 loss: 1.897 115 | Epoch: 2, Iteration: 160 loss: 1.761 116 | Epoch: 2, Iteration: 180 loss: 1.880 117 | Epoch: 2, Iteration: 200 loss: 1.827 118 | Epoch: 2, Iteration: 220 loss: 1.903 119 | Epoch: 2, Iteration: 240 loss: 1.948 120 | Epoch: 2, Iteration: 260 loss: 1.830 121 | Epoch: 2, Iteration: 280 loss: 1.780 122 | Epoch: 2, Iteration: 300 loss: 1.839 123 | Epoch: 2, Iteration: 320 loss: 1.845 124 | Epoch: 2, Iteration: 340 loss: 1.858 125 | Epoch: 2, Iteration: 360 loss: 1.799 126 | Epoch: 2, Iteration: 380 loss: 1.750 127 | Epoch: 2, Iteration: 400 loss: 1.805 128 | Epoch: 2, Iteration: 420 loss: 1.807 129 | Epoch: 2, Iteration: 440 loss: 1.712 130 | Epoch: 2, Iteration: 460 loss: 1.677 131 | Epoch: 2, Iteration: 480 loss: 1.672 132 | Epoch: 2, Iteration: 500 loss: 1.696 133 | Epoch: 2, Iteration: 520 loss: 1.676 134 | Epoch: 2, Iteration: 540 loss: 1.654 135 | Epoch: 2, Iteration: 560 loss: 1.626 136 | Epoch: 2, Iteration: 580 loss: 1.710 137 | Epoch: 2, Iteration: 600 loss: 1.691 138 | Epoch: 2, Iteration: 620 loss: 1.696 139 | Epoch: 2, Iteration: 640 loss: 1.682 140 | Epoch: 2, Iteration: 660 loss: 1.696 141 | Epoch: 2, Iteration: 680 loss: 1.651 142 | Epoch: 2, Iteration: 700 loss: 1.732 143 | Epoch: 2, Iteration: 720 loss: 1.723 144 | Epoch: 2, Iteration: 740 loss: 1.690 145 | Epoch: 2, Iteration: 760 loss: 1.676 146 | Epoch: 2, Iteration: 780 loss: 1.678 147 | Loss: 1.605 148 | Accuracy of the network on the 10000 test images: 42 % 149 | Accuracy of plane : 64 % 150 | Accuracy of car : 24 % 151 | Accuracy of bird : 27 % 152 | Accuracy of cat : 23 % 153 | Accuracy of deer : 23 % 154 | Accuracy of dog : 50 % 155 | Accuracy of frog : 55 % 156 | Accuracy of horse : 60 % 157 | Accuracy of ship : 58 % 158 | Accuracy of truck : 38 % 159 | Epoch: 3, Iteration: 0 loss: 0.082 160 | Epoch: 3, Iteration: 20 loss: 1.708 161 | Epoch: 3, Iteration: 40 loss: 1.733 162 | Epoch: 3, Iteration: 60 loss: 1.810 163 | Epoch: 3, Iteration: 80 loss: 1.855 164 | Epoch: 3, Iteration: 100 loss: 1.678 165 | Epoch: 3, Iteration: 120 loss: 1.794 166 | Epoch: 3, Iteration: 140 loss: 1.848 167 | Epoch: 3, Iteration: 160 loss: 1.771 168 | Epoch: 3, Iteration: 180 loss: 1.675 169 | Epoch: 3, Iteration: 200 loss: 1.618 170 | Epoch: 3, Iteration: 220 loss: 1.613 171 | Epoch: 3, Iteration: 240 loss: 1.705 172 | Epoch: 3, Iteration: 260 loss: 1.635 173 | Epoch: 3, Iteration: 280 loss: 1.595 174 | Epoch: 3, Iteration: 300 loss: 1.576 175 | Epoch: 3, Iteration: 320 loss: 1.593 176 | Epoch: 3, Iteration: 340 loss: 1.564 177 | Epoch: 3, Iteration: 360 loss: 1.517 178 | Epoch: 3, Iteration: 380 loss: 1.608 179 | Epoch: 3, Iteration: 400 loss: 1.602 180 | Epoch: 3, Iteration: 420 loss: 1.584 181 | Epoch: 3, Iteration: 440 loss: 1.598 182 | Epoch: 3, Iteration: 460 loss: 1.559 183 | Epoch: 3, Iteration: 480 loss: 1.496 184 | Epoch: 3, Iteration: 500 loss: 1.542 185 | Epoch: 3, Iteration: 520 loss: 1.538 186 | Epoch: 3, Iteration: 540 loss: 1.429 187 | Epoch: 3, Iteration: 560 loss: 1.507 188 | Epoch: 3, Iteration: 580 loss: 1.528 189 | Epoch: 3, Iteration: 600 loss: 1.527 190 | Epoch: 3, Iteration: 620 loss: 1.519 191 | Epoch: 3, Iteration: 640 loss: 1.547 192 | Epoch: 3, Iteration: 660 loss: 1.530 193 | Epoch: 3, Iteration: 680 loss: 1.580 194 | Epoch: 3, Iteration: 700 loss: 1.542 195 | Epoch: 3, Iteration: 720 loss: 1.655 196 | Epoch: 3, Iteration: 740 loss: 1.523 197 | Epoch: 3, Iteration: 760 loss: 1.617 198 | Epoch: 3, Iteration: 780 loss: 1.591 199 | Loss: 1.621 200 | Accuracy of the network on the 10000 test images: 37 % 201 | Accuracy of plane : 57 % 202 | Accuracy of car : 64 % 203 | Accuracy of bird : 25 % 204 | Accuracy of cat : 50 % 205 | Accuracy of deer : 27 % 206 | Accuracy of dog : 54 % 207 | Accuracy of frog : 12 % 208 | Accuracy of horse : 9 % 209 | Accuracy of ship : 84 % 210 | Accuracy of truck : 5 % 211 | Epoch: 4, Iteration: 0 loss: 0.078 212 | Epoch: 4, Iteration: 20 loss: 1.631 213 | Epoch: 4, Iteration: 40 loss: 1.547 214 | Epoch: 4, Iteration: 60 loss: 1.434 215 | Epoch: 4, Iteration: 80 loss: 1.406 216 | Epoch: 4, Iteration: 100 loss: 1.409 217 | Epoch: 4, Iteration: 120 loss: 1.450 218 | Epoch: 4, Iteration: 140 loss: 1.577 219 | Epoch: 4, Iteration: 160 loss: 1.458 220 | Epoch: 4, Iteration: 180 loss: 1.472 221 | Epoch: 4, Iteration: 200 loss: 1.391 222 | Epoch: 4, Iteration: 220 loss: 1.436 223 | Epoch: 4, Iteration: 240 loss: 1.487 224 | Epoch: 4, Iteration: 260 loss: 1.442 225 | Epoch: 4, Iteration: 280 loss: 1.449 226 | Epoch: 4, Iteration: 300 loss: 1.420 227 | Epoch: 4, Iteration: 320 loss: 1.410 228 | Epoch: 4, Iteration: 340 loss: 1.403 229 | Epoch: 4, Iteration: 360 loss: 1.427 230 | Epoch: 4, Iteration: 380 loss: 1.448 231 | Epoch: 4, Iteration: 400 loss: 1.437 232 | Epoch: 4, Iteration: 420 loss: 1.403 233 | Epoch: 4, Iteration: 440 loss: 1.403 234 | Epoch: 4, Iteration: 460 loss: 1.364 235 | Epoch: 4, Iteration: 480 loss: 1.357 236 | Epoch: 4, Iteration: 500 loss: 1.399 237 | Epoch: 4, Iteration: 520 loss: 1.460 238 | Epoch: 4, Iteration: 540 loss: 1.385 239 | Epoch: 4, Iteration: 560 loss: 1.384 240 | Epoch: 4, Iteration: 580 loss: 1.358 241 | Epoch: 4, Iteration: 600 loss: 1.477 242 | Epoch: 4, Iteration: 620 loss: 1.477 243 | Epoch: 4, Iteration: 640 loss: 1.447 244 | Epoch: 4, Iteration: 660 loss: 1.343 245 | Epoch: 4, Iteration: 680 loss: 1.318 246 | Epoch: 4, Iteration: 700 loss: 1.351 247 | Epoch: 4, Iteration: 720 loss: 1.392 248 | Epoch: 4, Iteration: 740 loss: 1.380 249 | Epoch: 4, Iteration: 760 loss: 1.366 250 | Epoch: 4, Iteration: 780 loss: 1.317 251 | Loss: 1.293 252 | Accuracy of the network on the 10000 test images: 53 % 253 | Accuracy of plane : 53 % 254 | Accuracy of car : 68 % 255 | Accuracy of bird : 41 % 256 | Accuracy of cat : 43 % 257 | Accuracy of deer : 25 % 258 | Accuracy of dog : 35 % 259 | Accuracy of frog : 80 % 260 | Accuracy of horse : 59 % 261 | Accuracy of ship : 68 % 262 | Accuracy of truck : 66 % 263 | Epoch: 5, Iteration: 0 loss: 0.067 264 | Epoch: 5, Iteration: 20 loss: 1.354 265 | Epoch: 5, Iteration: 40 loss: 1.315 266 | Epoch: 5, Iteration: 60 loss: 1.290 267 | Epoch: 5, Iteration: 80 loss: 1.347 268 | Epoch: 5, Iteration: 100 loss: 1.354 269 | Epoch: 5, Iteration: 120 loss: 1.428 270 | Epoch: 5, Iteration: 140 loss: 1.356 271 | Epoch: 5, Iteration: 160 loss: 1.363 272 | Epoch: 5, Iteration: 180 loss: 1.389 273 | Epoch: 5, Iteration: 200 loss: 1.408 274 | Epoch: 5, Iteration: 220 loss: 1.365 275 | Epoch: 5, Iteration: 240 loss: 1.329 276 | Epoch: 5, Iteration: 260 loss: 1.309 277 | Epoch: 5, Iteration: 280 loss: 1.393 278 | Epoch: 5, Iteration: 300 loss: 1.313 279 | Epoch: 5, Iteration: 320 loss: 1.298 280 | Epoch: 5, Iteration: 340 loss: 1.395 281 | Epoch: 5, Iteration: 360 loss: 1.329 282 | Epoch: 5, Iteration: 380 loss: 1.240 283 | Epoch: 5, Iteration: 400 loss: 1.295 284 | Epoch: 5, Iteration: 420 loss: 1.237 285 | Epoch: 5, Iteration: 440 loss: 1.267 286 | Epoch: 5, Iteration: 460 loss: 1.318 287 | Epoch: 5, Iteration: 480 loss: 1.175 288 | Epoch: 5, Iteration: 500 loss: 1.256 289 | Epoch: 5, Iteration: 520 loss: 1.243 290 | Epoch: 5, Iteration: 540 loss: 1.319 291 | Epoch: 5, Iteration: 560 loss: 1.240 292 | Epoch: 5, Iteration: 580 loss: 1.259 293 | Epoch: 5, Iteration: 600 loss: 1.267 294 | Epoch: 5, Iteration: 620 loss: 1.321 295 | Epoch: 5, Iteration: 640 loss: 1.351 296 | Epoch: 5, Iteration: 660 loss: 1.307 297 | Epoch: 5, Iteration: 680 loss: 1.290 298 | Epoch: 5, Iteration: 700 loss: 1.241 299 | Epoch: 5, Iteration: 720 loss: 1.270 300 | Epoch: 5, Iteration: 740 loss: 1.241 301 | Epoch: 5, Iteration: 760 loss: 1.276 302 | Epoch: 5, Iteration: 780 loss: 1.239 303 | Loss: 1.539 304 | Accuracy of the network on the 10000 test images: 45 % 305 | Accuracy of plane : 51 % 306 | Accuracy of car : 8 % 307 | Accuracy of bird : 45 % 308 | Accuracy of cat : 6 % 309 | Accuracy of deer : 32 % 310 | Accuracy of dog : 66 % 311 | Accuracy of frog : 16 % 312 | Accuracy of horse : 71 % 313 | Accuracy of ship : 82 % 314 | Accuracy of truck : 65 % 315 | Epoch: 6, Iteration: 0 loss: 0.077 316 | Epoch: 6, Iteration: 20 loss: 1.317 317 | Epoch: 6, Iteration: 40 loss: 1.283 318 | Epoch: 6, Iteration: 60 loss: 1.202 319 | Epoch: 6, Iteration: 80 loss: 1.204 320 | Epoch: 6, Iteration: 100 loss: 1.209 321 | Epoch: 6, Iteration: 120 loss: 1.199 322 | Epoch: 6, Iteration: 140 loss: 1.304 323 | Epoch: 6, Iteration: 160 loss: 1.231 324 | Epoch: 6, Iteration: 180 loss: 1.139 325 | Epoch: 6, Iteration: 200 loss: 1.185 326 | Epoch: 6, Iteration: 220 loss: 1.230 327 | Epoch: 6, Iteration: 240 loss: 1.217 328 | Epoch: 6, Iteration: 260 loss: 1.163 329 | Epoch: 6, Iteration: 280 loss: 1.159 330 | Epoch: 6, Iteration: 300 loss: 1.159 331 | Epoch: 6, Iteration: 320 loss: 1.156 332 | Epoch: 6, Iteration: 340 loss: 1.247 333 | Epoch: 6, Iteration: 360 loss: 1.252 334 | Epoch: 6, Iteration: 380 loss: 1.182 335 | Epoch: 6, Iteration: 400 loss: 1.182 336 | Epoch: 6, Iteration: 420 loss: 1.146 337 | Epoch: 6, Iteration: 440 loss: 1.188 338 | Epoch: 6, Iteration: 460 loss: 1.223 339 | Epoch: 6, Iteration: 480 loss: 1.165 340 | Epoch: 6, Iteration: 500 loss: 1.206 341 | Epoch: 6, Iteration: 520 loss: 1.163 342 | Epoch: 6, Iteration: 540 loss: 1.116 343 | Epoch: 6, Iteration: 560 loss: 1.162 344 | Epoch: 6, Iteration: 580 loss: 1.183 345 | Epoch: 6, Iteration: 600 loss: 1.170 346 | Epoch: 6, Iteration: 620 loss: 1.189 347 | Epoch: 6, Iteration: 640 loss: 1.188 348 | Epoch: 6, Iteration: 660 loss: 1.143 349 | Epoch: 6, Iteration: 680 loss: 1.124 350 | Epoch: 6, Iteration: 700 loss: 1.079 351 | Epoch: 6, Iteration: 720 loss: 1.169 352 | Epoch: 6, Iteration: 740 loss: 1.164 353 | Epoch: 6, Iteration: 760 loss: 1.083 354 | Epoch: 6, Iteration: 780 loss: 1.124 355 | Loss: 1.336 356 | Accuracy of the network on the 10000 test images: 51 % 357 | Accuracy of plane : 85 % 358 | Accuracy of car : 74 % 359 | Accuracy of bird : 31 % 360 | Accuracy of cat : 42 % 361 | Accuracy of deer : 16 % 362 | Accuracy of dog : 52 % 363 | Accuracy of frog : 39 % 364 | Accuracy of horse : 81 % 365 | Accuracy of ship : 37 % 366 | Accuracy of truck : 62 % 367 | Epoch 6: reducing learning rate of group 0 to 1.0000e-02. 368 | Epoch: 7, Iteration: 0 loss: 0.060 369 | Epoch: 7, Iteration: 20 loss: 1.291 370 | Epoch: 7, Iteration: 40 loss: 1.127 371 | Epoch: 7, Iteration: 60 loss: 0.935 372 | Epoch: 7, Iteration: 80 loss: 0.923 373 | Epoch: 7, Iteration: 100 loss: 0.917 374 | Epoch: 7, Iteration: 120 loss: 0.916 375 | Epoch: 7, Iteration: 140 loss: 0.835 376 | Epoch: 7, Iteration: 160 loss: 0.850 377 | Epoch: 7, Iteration: 180 loss: 0.871 378 | Epoch: 7, Iteration: 200 loss: 0.861 379 | Epoch: 7, Iteration: 220 loss: 0.843 380 | Epoch: 7, Iteration: 240 loss: 0.877 381 | Epoch: 7, Iteration: 260 loss: 0.817 382 | Epoch: 7, Iteration: 280 loss: 0.832 383 | Epoch: 7, Iteration: 300 loss: 0.775 384 | Epoch: 7, Iteration: 320 loss: 0.830 385 | Epoch: 7, Iteration: 340 loss: 0.795 386 | Epoch: 7, Iteration: 360 loss: 0.794 387 | Epoch: 7, Iteration: 380 loss: 0.789 388 | Epoch: 7, Iteration: 400 loss: 0.860 389 | Epoch: 7, Iteration: 420 loss: 0.790 390 | Epoch: 7, Iteration: 440 loss: 0.781 391 | Epoch: 7, Iteration: 460 loss: 0.812 392 | Epoch: 7, Iteration: 480 loss: 0.759 393 | Epoch: 7, Iteration: 500 loss: 0.742 394 | Epoch: 7, Iteration: 520 loss: 0.832 395 | Epoch: 7, Iteration: 540 loss: 0.815 396 | Epoch: 7, Iteration: 560 loss: 0.809 397 | Epoch: 7, Iteration: 580 loss: 0.747 398 | Epoch: 7, Iteration: 600 loss: 0.819 399 | Epoch: 7, Iteration: 620 loss: 0.794 400 | Epoch: 7, Iteration: 640 loss: 0.800 401 | Epoch: 7, Iteration: 660 loss: 0.793 402 | Epoch: 7, Iteration: 680 loss: 0.732 403 | Epoch: 7, Iteration: 700 loss: 0.779 404 | Epoch: 7, Iteration: 720 loss: 0.771 405 | Epoch: 7, Iteration: 740 loss: 0.736 406 | Epoch: 7, Iteration: 760 loss: 0.801 407 | Epoch: 7, Iteration: 780 loss: 0.774 408 | Loss: 1.079 409 | Accuracy of the network on the 10000 test images: 61 % 410 | Accuracy of plane : 73 % 411 | Accuracy of car : 74 % 412 | Accuracy of bird : 55 % 413 | Accuracy of cat : 49 % 414 | Accuracy of deer : 41 % 415 | Accuracy of dog : 49 % 416 | Accuracy of frog : 75 % 417 | Accuracy of horse : 62 % 418 | Accuracy of ship : 72 % 419 | Accuracy of truck : 69 % 420 | Epoch: 8, Iteration: 0 loss: 0.044 421 | Epoch: 8, Iteration: 20 loss: 0.751 422 | Epoch: 8, Iteration: 40 loss: 0.745 423 | Epoch: 8, Iteration: 60 loss: 0.745 424 | Epoch: 8, Iteration: 80 loss: 0.722 425 | Epoch: 8, Iteration: 100 loss: 0.756 426 | Epoch: 8, Iteration: 120 loss: 0.757 427 | Epoch: 8, Iteration: 140 loss: 0.721 428 | Epoch: 8, Iteration: 160 loss: 0.725 429 | Epoch: 8, Iteration: 180 loss: 0.721 430 | Epoch: 8, Iteration: 200 loss: 0.747 431 | Epoch: 8, Iteration: 220 loss: 0.699 432 | Epoch: 8, Iteration: 240 loss: 0.684 433 | Epoch: 8, Iteration: 260 loss: 0.665 434 | Epoch: 8, Iteration: 280 loss: 0.731 435 | Epoch: 8, Iteration: 300 loss: 0.703 436 | Epoch: 8, Iteration: 320 loss: 0.717 437 | Epoch: 8, Iteration: 340 loss: 0.727 438 | Epoch: 8, Iteration: 360 loss: 0.743 439 | Epoch: 8, Iteration: 380 loss: 0.775 440 | Epoch: 8, Iteration: 400 loss: 0.648 441 | Epoch: 8, Iteration: 420 loss: 0.720 442 | Epoch: 8, Iteration: 440 loss: 0.750 443 | Epoch: 8, Iteration: 460 loss: 0.739 444 | Epoch: 8, Iteration: 480 loss: 0.716 445 | Epoch: 8, Iteration: 500 loss: 0.643 446 | Epoch: 8, Iteration: 520 loss: 0.721 447 | Epoch: 8, Iteration: 540 loss: 0.663 448 | Epoch: 8, Iteration: 560 loss: 0.756 449 | Epoch: 8, Iteration: 580 loss: 0.748 450 | Epoch: 8, Iteration: 600 loss: 0.704 451 | Epoch: 8, Iteration: 620 loss: 0.687 452 | Epoch: 8, Iteration: 640 loss: 0.666 453 | Epoch: 8, Iteration: 660 loss: 0.666 454 | Epoch: 8, Iteration: 680 loss: 0.701 455 | Epoch: 8, Iteration: 700 loss: 0.717 456 | Epoch: 8, Iteration: 720 loss: 0.741 457 | Epoch: 8, Iteration: 740 loss: 0.727 458 | Epoch: 8, Iteration: 760 loss: 0.694 459 | Epoch: 8, Iteration: 780 loss: 0.720 460 | Loss: 1.085 461 | Accuracy of the network on the 10000 test images: 64 % 462 | Accuracy of plane : 75 % 463 | Accuracy of car : 78 % 464 | Accuracy of bird : 60 % 465 | Accuracy of cat : 45 % 466 | Accuracy of deer : 49 % 467 | Accuracy of dog : 55 % 468 | Accuracy of frog : 76 % 469 | Accuracy of horse : 62 % 470 | Accuracy of ship : 79 % 471 | Accuracy of truck : 65 % 472 | Epoch: 9, Iteration: 0 loss: 0.040 473 | Epoch: 9, Iteration: 20 loss: 0.636 474 | Epoch: 9, Iteration: 40 loss: 0.729 475 | Epoch: 9, Iteration: 60 loss: 0.639 476 | Epoch: 9, Iteration: 80 loss: 0.625 477 | Epoch: 9, Iteration: 100 loss: 0.628 478 | Epoch: 9, Iteration: 120 loss: 0.583 479 | Epoch: 9, Iteration: 140 loss: 0.628 480 | Epoch: 9, Iteration: 160 loss: 0.652 481 | Epoch: 9, Iteration: 180 loss: 0.638 482 | Epoch: 9, Iteration: 200 loss: 0.678 483 | Epoch: 9, Iteration: 220 loss: 0.656 484 | Epoch: 9, Iteration: 240 loss: 0.687 485 | Epoch: 9, Iteration: 260 loss: 0.580 486 | Epoch: 9, Iteration: 280 loss: 0.670 487 | Epoch: 9, Iteration: 300 loss: 0.636 488 | Epoch: 9, Iteration: 320 loss: 0.675 489 | Epoch: 9, Iteration: 340 loss: 0.633 490 | Epoch: 9, Iteration: 360 loss: 0.650 491 | Epoch: 9, Iteration: 380 loss: 0.649 492 | Epoch: 9, Iteration: 400 loss: 0.636 493 | Epoch: 9, Iteration: 420 loss: 0.658 494 | Epoch: 9, Iteration: 440 loss: 0.612 495 | Epoch: 9, Iteration: 460 loss: 0.643 496 | Epoch: 9, Iteration: 480 loss: 0.622 497 | Epoch: 9, Iteration: 500 loss: 0.631 498 | Epoch: 9, Iteration: 520 loss: 0.616 499 | Epoch: 9, Iteration: 540 loss: 0.618 500 | Epoch: 9, Iteration: 560 loss: 0.656 501 | Epoch: 9, Iteration: 580 loss: 0.623 502 | Epoch: 9, Iteration: 600 loss: 0.670 503 | Epoch: 9, Iteration: 620 loss: 0.627 504 | Epoch: 9, Iteration: 640 loss: 0.613 505 | Epoch: 9, Iteration: 660 loss: 0.613 506 | Epoch: 9, Iteration: 680 loss: 0.577 507 | Epoch: 9, Iteration: 700 loss: 0.695 508 | Epoch: 9, Iteration: 720 loss: 0.625 509 | Epoch: 9, Iteration: 740 loss: 0.662 510 | Epoch: 9, Iteration: 760 loss: 0.679 511 | Epoch: 9, Iteration: 780 loss: 0.627 512 | Loss: 1.094 513 | Accuracy of the network on the 10000 test images: 62 % 514 | Accuracy of plane : 73 % 515 | Accuracy of car : 74 % 516 | Accuracy of bird : 50 % 517 | Accuracy of cat : 49 % 518 | Accuracy of deer : 50 % 519 | Accuracy of dog : 52 % 520 | Accuracy of frog : 71 % 521 | Accuracy of horse : 64 % 522 | Accuracy of ship : 79 % 523 | Accuracy of truck : 64 % 524 | Epoch 9: reducing learning rate of group 0 to 1.0000e-03. 525 | Epoch: 10, Iteration: 0 loss: 0.030 526 | Epoch: 10, Iteration: 20 loss: 0.587 527 | Epoch: 10, Iteration: 40 loss: 0.563 528 | Epoch: 10, Iteration: 60 loss: 0.563 529 | Epoch: 10, Iteration: 80 loss: 0.593 530 | Epoch: 10, Iteration: 100 loss: 0.597 531 | Epoch: 10, Iteration: 120 loss: 0.568 532 | Epoch: 10, Iteration: 140 loss: 0.597 533 | Epoch: 10, Iteration: 160 loss: 0.559 534 | Epoch: 10, Iteration: 180 loss: 0.564 535 | Epoch: 10, Iteration: 200 loss: 0.614 536 | Epoch: 10, Iteration: 220 loss: 0.579 537 | Epoch: 10, Iteration: 240 loss: 0.577 538 | Epoch: 10, Iteration: 260 loss: 0.553 539 | Epoch: 10, Iteration: 280 loss: 0.574 540 | Epoch: 10, Iteration: 300 loss: 0.544 541 | Epoch: 10, Iteration: 320 loss: 0.592 542 | Epoch: 10, Iteration: 340 loss: 0.525 543 | Epoch: 10, Iteration: 360 loss: 0.556 544 | Epoch: 10, Iteration: 380 loss: 0.608 545 | Epoch: 10, Iteration: 400 loss: 0.581 546 | Epoch: 10, Iteration: 420 loss: 0.549 547 | Epoch: 10, Iteration: 440 loss: 0.525 548 | Epoch: 10, Iteration: 460 loss: 0.606 549 | Epoch: 10, Iteration: 480 loss: 0.577 550 | Epoch: 10, Iteration: 500 loss: 0.534 551 | Epoch: 10, Iteration: 520 loss: 0.616 552 | Epoch: 10, Iteration: 540 loss: 0.554 553 | Epoch: 10, Iteration: 560 loss: 0.600 554 | Epoch: 10, Iteration: 580 loss: 0.542 555 | Epoch: 10, Iteration: 600 loss: 0.596 556 | Epoch: 10, Iteration: 620 loss: 0.599 557 | Epoch: 10, Iteration: 640 loss: 0.563 558 | Epoch: 10, Iteration: 660 loss: 0.566 559 | Epoch: 10, Iteration: 680 loss: 0.590 560 | Epoch: 10, Iteration: 700 loss: 0.558 561 | Epoch: 10, Iteration: 720 loss: 0.565 562 | Epoch: 10, Iteration: 740 loss: 0.592 563 | Epoch: 10, Iteration: 760 loss: 0.645 564 | Epoch: 10, Iteration: 780 loss: 0.546 565 | Loss: 1.101 566 | Accuracy of the network on the 10000 test images: 63 % 567 | Accuracy of plane : 75 % 568 | Accuracy of car : 76 % 569 | Accuracy of bird : 55 % 570 | Accuracy of cat : 49 % 571 | Accuracy of deer : 49 % 572 | Accuracy of dog : 52 % 573 | Accuracy of frog : 73 % 574 | Accuracy of horse : 60 % 575 | Accuracy of ship : 81 % 576 | Accuracy of truck : 65 % 577 | Epoch: 11, Iteration: 0 loss: 0.032 578 | Epoch: 11, Iteration: 20 loss: 0.578 579 | Epoch: 11, Iteration: 40 loss: 0.523 580 | Epoch: 11, Iteration: 60 loss: 0.597 581 | Epoch: 11, Iteration: 80 loss: 0.560 582 | Epoch: 11, Iteration: 100 loss: 0.592 583 | Epoch: 11, Iteration: 120 loss: 0.547 584 | Epoch: 11, Iteration: 140 loss: 0.552 585 | Epoch: 11, Iteration: 160 loss: 0.574 586 | Epoch: 11, Iteration: 180 loss: 0.564 587 | Epoch: 11, Iteration: 200 loss: 0.558 588 | Epoch: 11, Iteration: 220 loss: 0.610 589 | Epoch: 11, Iteration: 240 loss: 0.515 590 | Epoch: 11, Iteration: 260 loss: 0.570 591 | Epoch: 11, Iteration: 280 loss: 0.595 592 | Epoch: 11, Iteration: 300 loss: 0.539 593 | Epoch: 11, Iteration: 320 loss: 0.533 594 | Epoch: 11, Iteration: 340 loss: 0.607 595 | Epoch: 11, Iteration: 360 loss: 0.536 596 | Epoch: 11, Iteration: 380 loss: 0.536 597 | Epoch: 11, Iteration: 400 loss: 0.561 598 | Epoch: 11, Iteration: 420 loss: 0.545 599 | Epoch: 11, Iteration: 440 loss: 0.557 600 | Epoch: 11, Iteration: 460 loss: 0.559 601 | Epoch: 11, Iteration: 480 loss: 0.583 602 | Epoch: 11, Iteration: 500 loss: 0.563 603 | Epoch: 11, Iteration: 520 loss: 0.545 604 | Epoch: 11, Iteration: 540 loss: 0.553 605 | Epoch: 11, Iteration: 560 loss: 0.608 606 | Epoch: 11, Iteration: 580 loss: 0.574 607 | Epoch: 11, Iteration: 600 loss: 0.577 608 | Epoch: 11, Iteration: 620 loss: 0.588 609 | Epoch: 11, Iteration: 640 loss: 0.565 610 | Epoch: 11, Iteration: 660 loss: 0.565 611 | Epoch: 11, Iteration: 680 loss: 0.566 612 | Epoch: 11, Iteration: 700 loss: 0.554 613 | Epoch: 11, Iteration: 720 loss: 0.547 614 | Epoch: 11, Iteration: 740 loss: 0.583 615 | Epoch: 11, Iteration: 760 loss: 0.541 616 | Epoch: 11, Iteration: 780 loss: 0.539 617 | Loss: 1.107 618 | Accuracy of the network on the 10000 test images: 63 % 619 | Accuracy of plane : 75 % 620 | Accuracy of car : 76 % 621 | Accuracy of bird : 55 % 622 | Accuracy of cat : 49 % 623 | Accuracy of deer : 50 % 624 | Accuracy of dog : 55 % 625 | Accuracy of frog : 73 % 626 | Accuracy of horse : 60 % 627 | Accuracy of ship : 81 % 628 | Accuracy of truck : 65 % 629 | Epoch 11: reducing learning rate of group 0 to 1.0000e-04. 630 | Epoch: 12, Iteration: 0 loss: 0.023 631 | Epoch: 12, Iteration: 20 loss: 0.578 632 | Epoch: 12, Iteration: 40 loss: 0.558 633 | Epoch: 12, Iteration: 60 loss: 0.571 634 | Epoch: 12, Iteration: 80 loss: 0.551 635 | Epoch: 12, Iteration: 100 loss: 0.583 636 | Epoch: 12, Iteration: 120 loss: 0.565 637 | Epoch: 12, Iteration: 140 loss: 0.584 638 | Epoch: 12, Iteration: 160 loss: 0.549 639 | Epoch: 12, Iteration: 180 loss: 0.618 640 | Epoch: 12, Iteration: 200 loss: 0.569 641 | Epoch: 12, Iteration: 220 loss: 0.553 642 | Epoch: 12, Iteration: 240 loss: 0.539 643 | Epoch: 12, Iteration: 260 loss: 0.554 644 | Epoch: 12, Iteration: 280 loss: 0.538 645 | Epoch: 12, Iteration: 300 loss: 0.535 646 | Epoch: 12, Iteration: 320 loss: 0.573 647 | Epoch: 12, Iteration: 340 loss: 0.531 648 | Epoch: 12, Iteration: 360 loss: 0.571 649 | Epoch: 12, Iteration: 380 loss: 0.542 650 | Epoch: 12, Iteration: 400 loss: 0.545 651 | Epoch: 12, Iteration: 420 loss: 0.518 652 | Epoch: 12, Iteration: 440 loss: 0.579 653 | Epoch: 12, Iteration: 460 loss: 0.532 654 | Epoch: 12, Iteration: 480 loss: 0.520 655 | Epoch: 12, Iteration: 500 loss: 0.534 656 | Epoch: 12, Iteration: 520 loss: 0.541 657 | Epoch: 12, Iteration: 540 loss: 0.554 658 | Epoch: 12, Iteration: 560 loss: 0.564 659 | Epoch: 12, Iteration: 580 loss: 0.550 660 | Epoch: 12, Iteration: 600 loss: 0.544 661 | Epoch: 12, Iteration: 620 loss: 0.529 662 | Epoch: 12, Iteration: 640 loss: 0.535 663 | Epoch: 12, Iteration: 660 loss: 0.565 664 | Epoch: 12, Iteration: 680 loss: 0.582 665 | Epoch: 12, Iteration: 700 loss: 0.584 666 | Epoch: 12, Iteration: 720 loss: 0.542 667 | Epoch: 12, Iteration: 740 loss: 0.596 668 | Epoch: 12, Iteration: 760 loss: 0.537 669 | Epoch: 12, Iteration: 780 loss: 0.532 670 | Loss: 1.108 671 | Accuracy of the network on the 10000 test images: 63 % 672 | Accuracy of plane : 75 % 673 | Accuracy of car : 76 % 674 | Accuracy of bird : 55 % 675 | Accuracy of cat : 49 % 676 | Accuracy of deer : 50 % 677 | Accuracy of dog : 54 % 678 | Accuracy of frog : 73 % 679 | Accuracy of horse : 60 % 680 | Accuracy of ship : 81 % 681 | Accuracy of truck : 65 % 682 | Epoch: 13, Iteration: 0 loss: 0.019 683 | Epoch: 13, Iteration: 20 loss: 0.547 684 | Epoch: 13, Iteration: 40 loss: 0.577 685 | Epoch: 13, Iteration: 60 loss: 0.551 686 | Epoch: 13, Iteration: 80 loss: 0.570 687 | Epoch: 13, Iteration: 100 loss: 0.547 688 | Epoch: 13, Iteration: 120 loss: 0.516 689 | Epoch: 13, Iteration: 140 loss: 0.553 690 | Epoch: 13, Iteration: 160 loss: 0.547 691 | Epoch: 13, Iteration: 180 loss: 0.538 692 | Epoch: 13, Iteration: 200 loss: 0.550 693 | Epoch: 13, Iteration: 220 loss: 0.557 694 | Epoch: 13, Iteration: 240 loss: 0.533 695 | Epoch: 13, Iteration: 260 loss: 0.555 696 | Epoch: 13, Iteration: 280 loss: 0.543 697 | Epoch: 13, Iteration: 300 loss: 0.578 698 | Epoch: 13, Iteration: 320 loss: 0.565 699 | Epoch: 13, Iteration: 340 loss: 0.560 700 | Epoch: 13, Iteration: 360 loss: 0.531 701 | Epoch: 13, Iteration: 380 loss: 0.560 702 | Epoch: 13, Iteration: 400 loss: 0.548 703 | Epoch: 13, Iteration: 420 loss: 0.552 704 | Epoch: 13, Iteration: 440 loss: 0.560 705 | Epoch: 13, Iteration: 460 loss: 0.571 706 | Epoch: 13, Iteration: 480 loss: 0.524 707 | Epoch: 13, Iteration: 500 loss: 0.554 708 | Epoch: 13, Iteration: 520 loss: 0.512 709 | Epoch: 13, Iteration: 540 loss: 0.550 710 | Epoch: 13, Iteration: 560 loss: 0.587 711 | Epoch: 13, Iteration: 580 loss: 0.567 712 | Epoch: 13, Iteration: 600 loss: 0.554 713 | Epoch: 13, Iteration: 620 loss: 0.539 714 | Epoch: 13, Iteration: 640 loss: 0.594 715 | Epoch: 13, Iteration: 660 loss: 0.562 716 | Epoch: 13, Iteration: 680 loss: 0.539 717 | Epoch: 13, Iteration: 700 loss: 0.529 718 | Epoch: 13, Iteration: 720 loss: 0.597 719 | Epoch: 13, Iteration: 740 loss: 0.577 720 | Epoch: 13, Iteration: 760 loss: 0.569 721 | Epoch: 13, Iteration: 780 loss: 0.554 722 | Loss: 1.108 723 | Accuracy of the network on the 10000 test images: 63 % 724 | Accuracy of plane : 75 % 725 | Accuracy of car : 76 % 726 | Accuracy of bird : 55 % 727 | Accuracy of cat : 49 % 728 | Accuracy of deer : 50 % 729 | Accuracy of dog : 54 % 730 | Accuracy of frog : 73 % 731 | Accuracy of horse : 60 % 732 | Accuracy of ship : 81 % 733 | Accuracy of truck : 65 % 734 | Epoch 13: reducing learning rate of group 0 to 1.0000e-05. 735 | Epoch: 14, Iteration: 0 loss: 0.036 736 | Epoch: 14, Iteration: 20 loss: 0.511 737 | Epoch: 14, Iteration: 40 loss: 0.577 738 | Epoch: 14, Iteration: 60 loss: 0.588 739 | Epoch: 14, Iteration: 80 loss: 0.582 740 | Epoch: 14, Iteration: 100 loss: 0.573 741 | Epoch: 14, Iteration: 120 loss: 0.560 742 | Epoch: 14, Iteration: 140 loss: 0.539 743 | Epoch: 14, Iteration: 160 loss: 0.528 744 | Epoch: 14, Iteration: 180 loss: 0.572 745 | Epoch: 14, Iteration: 200 loss: 0.580 746 | Epoch: 14, Iteration: 220 loss: 0.518 747 | Epoch: 14, Iteration: 240 loss: 0.568 748 | Epoch: 14, Iteration: 260 loss: 0.546 749 | Epoch: 14, Iteration: 280 loss: 0.564 750 | Epoch: 14, Iteration: 300 loss: 0.545 751 | Epoch: 14, Iteration: 320 loss: 0.531 752 | Epoch: 14, Iteration: 340 loss: 0.544 753 | Epoch: 14, Iteration: 360 loss: 0.608 754 | Epoch: 14, Iteration: 380 loss: 0.549 755 | Epoch: 14, Iteration: 400 loss: 0.549 756 | Epoch: 14, Iteration: 420 loss: 0.504 757 | Epoch: 14, Iteration: 440 loss: 0.537 758 | Epoch: 14, Iteration: 460 loss: 0.538 759 | Epoch: 14, Iteration: 480 loss: 0.544 760 | Epoch: 14, Iteration: 500 loss: 0.561 761 | Epoch: 14, Iteration: 520 loss: 0.585 762 | Epoch: 14, Iteration: 540 loss: 0.586 763 | Epoch: 14, Iteration: 560 loss: 0.519 764 | Epoch: 14, Iteration: 580 loss: 0.610 765 | Epoch: 14, Iteration: 600 loss: 0.550 766 | Epoch: 14, Iteration: 620 loss: 0.534 767 | Epoch: 14, Iteration: 640 loss: 0.529 768 | Epoch: 14, Iteration: 660 loss: 0.570 769 | Epoch: 14, Iteration: 680 loss: 0.497 770 | Epoch: 14, Iteration: 700 loss: 0.555 771 | Epoch: 14, Iteration: 720 loss: 0.525 772 | Epoch: 14, Iteration: 740 loss: 0.589 773 | Epoch: 14, Iteration: 760 loss: 0.573 774 | Epoch: 14, Iteration: 780 loss: 0.524 775 | Loss: 1.108 776 | Accuracy of the network on the 10000 test images: 63 % 777 | Accuracy of plane : 75 % 778 | Accuracy of car : 76 % 779 | Accuracy of bird : 55 % 780 | Accuracy of cat : 49 % 781 | Accuracy of deer : 50 % 782 | Accuracy of dog : 54 % 783 | Accuracy of frog : 73 % 784 | Accuracy of horse : 60 % 785 | Accuracy of ship : 81 % 786 | Accuracy of truck : 65 % 787 | Epoch: 15, Iteration: 0 loss: 0.027 788 | Epoch: 15, Iteration: 20 loss: 0.528 789 | Epoch: 15, Iteration: 40 loss: 0.561 790 | Epoch: 15, Iteration: 60 loss: 0.546 791 | Epoch: 15, Iteration: 80 loss: 0.553 792 | Epoch: 15, Iteration: 100 loss: 0.569 793 | Epoch: 15, Iteration: 120 loss: 0.590 794 | Epoch: 15, Iteration: 140 loss: 0.538 795 | Epoch: 15, Iteration: 160 loss: 0.516 796 | Epoch: 15, Iteration: 180 loss: 0.569 797 | Epoch: 15, Iteration: 200 loss: 0.531 798 | Epoch: 15, Iteration: 220 loss: 0.565 799 | Epoch: 15, Iteration: 240 loss: 0.559 800 | Epoch: 15, Iteration: 260 loss: 0.552 801 | Epoch: 15, Iteration: 280 loss: 0.506 802 | Epoch: 15, Iteration: 300 loss: 0.564 803 | Epoch: 15, Iteration: 320 loss: 0.529 804 | Epoch: 15, Iteration: 340 loss: 0.527 805 | Epoch: 15, Iteration: 360 loss: 0.575 806 | Epoch: 15, Iteration: 380 loss: 0.543 807 | Epoch: 15, Iteration: 400 loss: 0.513 808 | Epoch: 15, Iteration: 420 loss: 0.583 809 | Epoch: 15, Iteration: 440 loss: 0.552 810 | Epoch: 15, Iteration: 460 loss: 0.540 811 | Epoch: 15, Iteration: 480 loss: 0.563 812 | Epoch: 15, Iteration: 500 loss: 0.534 813 | Epoch: 15, Iteration: 520 loss: 0.544 814 | Epoch: 15, Iteration: 540 loss: 0.544 815 | Epoch: 15, Iteration: 560 loss: 0.568 816 | Epoch: 15, Iteration: 580 loss: 0.563 817 | Epoch: 15, Iteration: 600 loss: 0.587 818 | Epoch: 15, Iteration: 620 loss: 0.530 819 | Epoch: 15, Iteration: 640 loss: 0.563 820 | Epoch: 15, Iteration: 660 loss: 0.573 821 | Epoch: 15, Iteration: 680 loss: 0.560 822 | Epoch: 15, Iteration: 700 loss: 0.536 823 | Epoch: 15, Iteration: 720 loss: 0.552 824 | Epoch: 15, Iteration: 740 loss: 0.570 825 | Epoch: 15, Iteration: 760 loss: 0.588 826 | Epoch: 15, Iteration: 780 loss: 0.578 827 | Loss: 1.108 828 | Accuracy of the network on the 10000 test images: 63 % 829 | Accuracy of plane : 75 % 830 | Accuracy of car : 76 % 831 | Accuracy of bird : 55 % 832 | Accuracy of cat : 49 % 833 | Accuracy of deer : 50 % 834 | Accuracy of dog : 54 % 835 | Accuracy of frog : 73 % 836 | Accuracy of horse : 60 % 837 | Accuracy of ship : 81 % 838 | Accuracy of truck : 65 % 839 | Epoch 15: reducing learning rate of group 0 to 1.0000e-06. 840 | Epoch: 16, Iteration: 0 loss: 0.027 841 | Epoch: 16, Iteration: 20 loss: 0.584 842 | Epoch: 16, Iteration: 40 loss: 0.580 843 | Epoch: 16, Iteration: 60 loss: 0.542 844 | Epoch: 16, Iteration: 80 loss: 0.563 845 | Epoch: 16, Iteration: 100 loss: 0.535 846 | Epoch: 16, Iteration: 120 loss: 0.544 847 | Epoch: 16, Iteration: 140 loss: 0.534 848 | Epoch: 16, Iteration: 160 loss: 0.564 849 | Epoch: 16, Iteration: 180 loss: 0.556 850 | Epoch: 16, Iteration: 200 loss: 0.554 851 | Epoch: 16, Iteration: 220 loss: 0.540 852 | Epoch: 16, Iteration: 240 loss: 0.532 853 | Epoch: 16, Iteration: 260 loss: 0.564 854 | Epoch: 16, Iteration: 280 loss: 0.566 855 | Epoch: 16, Iteration: 300 loss: 0.508 856 | Epoch: 16, Iteration: 320 loss: 0.544 857 | Epoch: 16, Iteration: 340 loss: 0.561 858 | Epoch: 16, Iteration: 360 loss: 0.536 859 | Epoch: 16, Iteration: 380 loss: 0.560 860 | Epoch: 16, Iteration: 400 loss: 0.579 861 | Epoch: 16, Iteration: 420 loss: 0.510 862 | Epoch: 16, Iteration: 440 loss: 0.549 863 | Epoch: 16, Iteration: 460 loss: 0.575 864 | Epoch: 16, Iteration: 480 loss: 0.562 865 | Epoch: 16, Iteration: 500 loss: 0.584 866 | Epoch: 16, Iteration: 520 loss: 0.520 867 | Epoch: 16, Iteration: 540 loss: 0.566 868 | Epoch: 16, Iteration: 560 loss: 0.587 869 | Epoch: 16, Iteration: 580 loss: 0.543 870 | Epoch: 16, Iteration: 600 loss: 0.518 871 | Epoch: 16, Iteration: 620 loss: 0.533 872 | Epoch: 16, Iteration: 640 loss: 0.555 873 | Epoch: 16, Iteration: 660 loss: 0.562 874 | Epoch: 16, Iteration: 680 loss: 0.537 875 | Epoch: 16, Iteration: 700 loss: 0.553 876 | Epoch: 16, Iteration: 720 loss: 0.558 877 | Epoch: 16, Iteration: 740 loss: 0.564 878 | Epoch: 16, Iteration: 760 loss: 0.548 879 | Epoch: 16, Iteration: 780 loss: 0.597 880 | Loss: 1.108 881 | Accuracy of the network on the 10000 test images: 63 % 882 | Accuracy of plane : 75 % 883 | Accuracy of car : 76 % 884 | Accuracy of bird : 55 % 885 | Accuracy of cat : 49 % 886 | Accuracy of deer : 50 % 887 | Accuracy of dog : 54 % 888 | Accuracy of frog : 73 % 889 | Accuracy of horse : 60 % 890 | Accuracy of ship : 81 % 891 | Accuracy of truck : 65 % 892 | Epoch: 17, Iteration: 0 loss: 0.020 893 | Epoch: 17, Iteration: 20 loss: 0.555 894 | Epoch: 17, Iteration: 40 loss: 0.559 895 | Epoch: 17, Iteration: 60 loss: 0.551 896 | Epoch: 17, Iteration: 80 loss: 0.581 897 | Epoch: 17, Iteration: 100 loss: 0.522 898 | Epoch: 17, Iteration: 120 loss: 0.559 899 | Epoch: 17, Iteration: 140 loss: 0.555 900 | Epoch: 17, Iteration: 160 loss: 0.549 901 | Epoch: 17, Iteration: 180 loss: 0.560 902 | Epoch: 17, Iteration: 200 loss: 0.561 903 | Epoch: 17, Iteration: 220 loss: 0.536 904 | Epoch: 17, Iteration: 240 loss: 0.541 905 | Epoch: 17, Iteration: 260 loss: 0.553 906 | Epoch: 17, Iteration: 280 loss: 0.508 907 | Epoch: 17, Iteration: 300 loss: 0.519 908 | Epoch: 17, Iteration: 320 loss: 0.547 909 | Epoch: 17, Iteration: 340 loss: 0.536 910 | Epoch: 17, Iteration: 360 loss: 0.550 911 | Epoch: 17, Iteration: 380 loss: 0.547 912 | Epoch: 17, Iteration: 400 loss: 0.540 913 | Epoch: 17, Iteration: 420 loss: 0.570 914 | Epoch: 17, Iteration: 440 loss: 0.541 915 | Epoch: 17, Iteration: 460 loss: 0.573 916 | Epoch: 17, Iteration: 480 loss: 0.556 917 | Epoch: 17, Iteration: 500 loss: 0.519 918 | Epoch: 17, Iteration: 520 loss: 0.588 919 | Epoch: 17, Iteration: 540 loss: 0.527 920 | Epoch: 17, Iteration: 560 loss: 0.544 921 | Epoch: 17, Iteration: 580 loss: 0.533 922 | Epoch: 17, Iteration: 600 loss: 0.587 923 | Epoch: 17, Iteration: 620 loss: 0.583 924 | Epoch: 17, Iteration: 640 loss: 0.576 925 | Epoch: 17, Iteration: 660 loss: 0.551 926 | Epoch: 17, Iteration: 680 loss: 0.584 927 | Epoch: 17, Iteration: 700 loss: 0.548 928 | Epoch: 17, Iteration: 720 loss: 0.567 929 | Epoch: 17, Iteration: 740 loss: 0.563 930 | Epoch: 17, Iteration: 760 loss: 0.576 931 | Epoch: 17, Iteration: 780 loss: 0.559 932 | Loss: 1.108 933 | Accuracy of the network on the 10000 test images: 63 % 934 | Accuracy of plane : 75 % 935 | Accuracy of car : 76 % 936 | Accuracy of bird : 55 % 937 | Accuracy of cat : 49 % 938 | Accuracy of deer : 50 % 939 | Accuracy of dog : 54 % 940 | Accuracy of frog : 73 % 941 | Accuracy of horse : 60 % 942 | Accuracy of ship : 81 % 943 | Accuracy of truck : 65 % 944 | Epoch 17: reducing learning rate of group 0 to 1.0000e-07. 945 | Epoch: 18, Iteration: 0 loss: 0.025 946 | Epoch: 18, Iteration: 20 loss: 0.549 947 | Epoch: 18, Iteration: 40 loss: 0.500 948 | Epoch: 18, Iteration: 60 loss: 0.538 949 | Epoch: 18, Iteration: 80 loss: 0.614 950 | Epoch: 18, Iteration: 100 loss: 0.529 951 | Epoch: 18, Iteration: 120 loss: 0.568 952 | Epoch: 18, Iteration: 140 loss: 0.557 953 | Epoch: 18, Iteration: 160 loss: 0.527 954 | Epoch: 18, Iteration: 180 loss: 0.588 955 | Epoch: 18, Iteration: 200 loss: 0.548 956 | Epoch: 18, Iteration: 220 loss: 0.571 957 | Epoch: 18, Iteration: 240 loss: 0.523 958 | Epoch: 18, Iteration: 260 loss: 0.592 959 | Epoch: 18, Iteration: 280 loss: 0.565 960 | Epoch: 18, Iteration: 300 loss: 0.526 961 | Epoch: 18, Iteration: 320 loss: 0.574 962 | Epoch: 18, Iteration: 340 loss: 0.524 963 | Epoch: 18, Iteration: 360 loss: 0.555 964 | Epoch: 18, Iteration: 380 loss: 0.567 965 | Epoch: 18, Iteration: 400 loss: 0.550 966 | Epoch: 18, Iteration: 420 loss: 0.529 967 | Epoch: 18, Iteration: 440 loss: 0.544 968 | Epoch: 18, Iteration: 460 loss: 0.568 969 | Epoch: 18, Iteration: 480 loss: 0.547 970 | Epoch: 18, Iteration: 500 loss: 0.554 971 | Epoch: 18, Iteration: 520 loss: 0.554 972 | Epoch: 18, Iteration: 540 loss: 0.536 973 | Epoch: 18, Iteration: 560 loss: 0.574 974 | Epoch: 18, Iteration: 580 loss: 0.579 975 | Epoch: 18, Iteration: 600 loss: 0.559 976 | Epoch: 18, Iteration: 620 loss: 0.537 977 | Epoch: 18, Iteration: 640 loss: 0.545 978 | Epoch: 18, Iteration: 660 loss: 0.565 979 | Epoch: 18, Iteration: 680 loss: 0.538 980 | Epoch: 18, Iteration: 700 loss: 0.571 981 | Epoch: 18, Iteration: 720 loss: 0.550 982 | Epoch: 18, Iteration: 740 loss: 0.578 983 | Epoch: 18, Iteration: 760 loss: 0.511 984 | Epoch: 18, Iteration: 780 loss: 0.564 985 | Loss: 1.108 986 | Accuracy of the network on the 10000 test images: 63 % 987 | Accuracy of plane : 75 % 988 | Accuracy of car : 76 % 989 | Accuracy of bird : 55 % 990 | Accuracy of cat : 49 % 991 | Accuracy of deer : 50 % 992 | Accuracy of dog : 54 % 993 | Accuracy of frog : 73 % 994 | Accuracy of horse : 60 % 995 | Accuracy of ship : 81 % 996 | Accuracy of truck : 65 % 997 | Epoch: 19, Iteration: 0 loss: 0.021 998 | Epoch: 19, Iteration: 20 loss: 0.527 999 | Epoch: 19, Iteration: 40 loss: 0.558 1000 | Epoch: 19, Iteration: 60 loss: 0.578 1001 | Epoch: 19, Iteration: 80 loss: 0.566 1002 | Epoch: 19, Iteration: 100 loss: 0.556 1003 | Epoch: 19, Iteration: 120 loss: 0.500 1004 | Epoch: 19, Iteration: 140 loss: 0.564 1005 | Epoch: 19, Iteration: 160 loss: 0.561 1006 | Epoch: 19, Iteration: 180 loss: 0.567 1007 | Epoch: 19, Iteration: 200 loss: 0.569 1008 | Epoch: 19, Iteration: 220 loss: 0.549 1009 | Epoch: 19, Iteration: 240 loss: 0.508 1010 | Epoch: 19, Iteration: 260 loss: 0.564 1011 | Epoch: 19, Iteration: 280 loss: 0.587 1012 | Epoch: 19, Iteration: 300 loss: 0.544 1013 | Epoch: 19, Iteration: 320 loss: 0.548 1014 | Epoch: 19, Iteration: 340 loss: 0.538 1015 | Epoch: 19, Iteration: 360 loss: 0.525 1016 | Epoch: 19, Iteration: 380 loss: 0.525 1017 | Epoch: 19, Iteration: 400 loss: 0.606 1018 | Epoch: 19, Iteration: 420 loss: 0.530 1019 | Epoch: 19, Iteration: 440 loss: 0.548 1020 | Epoch: 19, Iteration: 460 loss: 0.510 1021 | Epoch: 19, Iteration: 480 loss: 0.533 1022 | Epoch: 19, Iteration: 500 loss: 0.560 1023 | Epoch: 19, Iteration: 520 loss: 0.590 1024 | Epoch: 19, Iteration: 540 loss: 0.642 1025 | Epoch: 19, Iteration: 560 loss: 0.516 1026 | Epoch: 19, Iteration: 580 loss: 0.567 1027 | Epoch: 19, Iteration: 600 loss: 0.566 1028 | Epoch: 19, Iteration: 620 loss: 0.548 1029 | Epoch: 19, Iteration: 640 loss: 0.564 1030 | Epoch: 19, Iteration: 660 loss: 0.535 1031 | Epoch: 19, Iteration: 680 loss: 0.533 1032 | Epoch: 19, Iteration: 700 loss: 0.544 1033 | Epoch: 19, Iteration: 720 loss: 0.581 1034 | Epoch: 19, Iteration: 740 loss: 0.561 1035 | Epoch: 19, Iteration: 760 loss: 0.548 1036 | Epoch: 19, Iteration: 780 loss: 0.551 1037 | Loss: 1.108 1038 | Accuracy of the network on the 10000 test images: 63 % 1039 | Accuracy of plane : 75 % 1040 | Accuracy of car : 76 % 1041 | Accuracy of bird : 55 % 1042 | Accuracy of cat : 49 % 1043 | Accuracy of deer : 50 % 1044 | Accuracy of dog : 54 % 1045 | Accuracy of frog : 73 % 1046 | Accuracy of horse : 60 % 1047 | Accuracy of ship : 81 % 1048 | Accuracy of truck : 65 % 1049 | Epoch 19: reducing learning rate of group 0 to 1.0000e-08. 1050 | Epoch: 20, Iteration: 0 loss: 0.020 1051 | Epoch: 20, Iteration: 20 loss: 0.595 1052 | Epoch: 20, Iteration: 40 loss: 0.552 1053 | Epoch: 20, Iteration: 60 loss: 0.589 1054 | Epoch: 20, Iteration: 80 loss: 0.552 1055 | Epoch: 20, Iteration: 100 loss: 0.557 1056 | Epoch: 20, Iteration: 120 loss: 0.575 1057 | Epoch: 20, Iteration: 140 loss: 0.573 1058 | Epoch: 20, Iteration: 160 loss: 0.515 1059 | Epoch: 20, Iteration: 180 loss: 0.532 1060 | Epoch: 20, Iteration: 200 loss: 0.537 1061 | Epoch: 20, Iteration: 220 loss: 0.568 1062 | Epoch: 20, Iteration: 240 loss: 0.533 1063 | Epoch: 20, Iteration: 260 loss: 0.514 1064 | Epoch: 20, Iteration: 280 loss: 0.550 1065 | Epoch: 20, Iteration: 300 loss: 0.552 1066 | Epoch: 20, Iteration: 320 loss: 0.523 1067 | Epoch: 20, Iteration: 340 loss: 0.570 1068 | Epoch: 20, Iteration: 360 loss: 0.578 1069 | Epoch: 20, Iteration: 380 loss: 0.546 1070 | Epoch: 20, Iteration: 400 loss: 0.549 1071 | Epoch: 20, Iteration: 420 loss: 0.533 1072 | Epoch: 20, Iteration: 440 loss: 0.531 1073 | Epoch: 20, Iteration: 460 loss: 0.614 1074 | Epoch: 20, Iteration: 480 loss: 0.562 1075 | Epoch: 20, Iteration: 500 loss: 0.546 1076 | Epoch: 20, Iteration: 520 loss: 0.534 1077 | Epoch: 20, Iteration: 540 loss: 0.544 1078 | Epoch: 20, Iteration: 560 loss: 0.594 1079 | Epoch: 20, Iteration: 580 loss: 0.586 1080 | Epoch: 20, Iteration: 600 loss: 0.543 1081 | Epoch: 20, Iteration: 620 loss: 0.550 1082 | Epoch: 20, Iteration: 640 loss: 0.542 1083 | Epoch: 20, Iteration: 660 loss: 0.559 1084 | Epoch: 20, Iteration: 680 loss: 0.516 1085 | Epoch: 20, Iteration: 700 loss: 0.530 1086 | Epoch: 20, Iteration: 720 loss: 0.539 1087 | Epoch: 20, Iteration: 740 loss: 0.569 1088 | Epoch: 20, Iteration: 760 loss: 0.555 1089 | Epoch: 20, Iteration: 780 loss: 0.563 1090 | Loss: 1.108 1091 | Accuracy of the network on the 10000 test images: 63 % 1092 | Accuracy of plane : 75 % 1093 | Accuracy of car : 76 % 1094 | Accuracy of bird : 55 % 1095 | Accuracy of cat : 49 % 1096 | Accuracy of deer : 50 % 1097 | Accuracy of dog : 54 % 1098 | Accuracy of frog : 73 % 1099 | Accuracy of horse : 60 % 1100 | Accuracy of ship : 81 % 1101 | Accuracy of truck : 65 % 1102 | Epoch: 21, Iteration: 0 loss: 0.021 1103 | Epoch: 21, Iteration: 20 loss: 0.524 1104 | Epoch: 21, Iteration: 40 loss: 0.557 1105 | Epoch: 21, Iteration: 60 loss: 0.582 1106 | Epoch: 21, Iteration: 80 loss: 0.508 1107 | Epoch: 21, Iteration: 100 loss: 0.559 1108 | Epoch: 21, Iteration: 120 loss: 0.535 1109 | Epoch: 21, Iteration: 140 loss: 0.565 1110 | Epoch: 21, Iteration: 160 loss: 0.546 1111 | Epoch: 21, Iteration: 180 loss: 0.554 1112 | Epoch: 21, Iteration: 200 loss: 0.548 1113 | Epoch: 21, Iteration: 220 loss: 0.536 1114 | Epoch: 21, Iteration: 240 loss: 0.528 1115 | Epoch: 21, Iteration: 260 loss: 0.541 1116 | Epoch: 21, Iteration: 280 loss: 0.559 1117 | Epoch: 21, Iteration: 300 loss: 0.594 1118 | Epoch: 21, Iteration: 320 loss: 0.563 1119 | Epoch: 21, Iteration: 340 loss: 0.574 1120 | Epoch: 21, Iteration: 360 loss: 0.565 1121 | Epoch: 21, Iteration: 380 loss: 0.550 1122 | Epoch: 21, Iteration: 400 loss: 0.520 1123 | Epoch: 21, Iteration: 420 loss: 0.519 1124 | Epoch: 21, Iteration: 440 loss: 0.543 1125 | Epoch: 21, Iteration: 460 loss: 0.565 1126 | Epoch: 21, Iteration: 480 loss: 0.508 1127 | Epoch: 21, Iteration: 500 loss: 0.605 1128 | Epoch: 21, Iteration: 520 loss: 0.583 1129 | Epoch: 21, Iteration: 540 loss: 0.578 1130 | Epoch: 21, Iteration: 560 loss: 0.563 1131 | Epoch: 21, Iteration: 580 loss: 0.606 1132 | Epoch: 21, Iteration: 600 loss: 0.518 1133 | Epoch: 21, Iteration: 620 loss: 0.510 1134 | Epoch: 21, Iteration: 640 loss: 0.554 1135 | Epoch: 21, Iteration: 660 loss: 0.557 1136 | Epoch: 21, Iteration: 680 loss: 0.538 1137 | Epoch: 21, Iteration: 700 loss: 0.588 1138 | Epoch: 21, Iteration: 720 loss: 0.563 1139 | Epoch: 21, Iteration: 740 loss: 0.570 1140 | Epoch: 21, Iteration: 760 loss: 0.541 1141 | Epoch: 21, Iteration: 780 loss: 0.548 1142 | Loss: 1.108 1143 | Accuracy of the network on the 10000 test images: 63 % 1144 | Accuracy of plane : 75 % 1145 | Accuracy of car : 76 % 1146 | Accuracy of bird : 55 % 1147 | Accuracy of cat : 49 % 1148 | Accuracy of deer : 50 % 1149 | Accuracy of dog : 54 % 1150 | Accuracy of frog : 73 % 1151 | Accuracy of horse : 60 % 1152 | Accuracy of ship : 81 % 1153 | Accuracy of truck : 65 % 1154 | Epoch: 22, Iteration: 0 loss: 0.028 1155 | Epoch: 22, Iteration: 20 loss: 0.587 1156 | Epoch: 22, Iteration: 40 loss: 0.550 1157 | Epoch: 22, Iteration: 60 loss: 0.551 1158 | Epoch: 22, Iteration: 80 loss: 0.534 1159 | Epoch: 22, Iteration: 100 loss: 0.567 1160 | Epoch: 22, Iteration: 120 loss: 0.579 1161 | Epoch: 22, Iteration: 140 loss: 0.525 1162 | Epoch: 22, Iteration: 160 loss: 0.551 1163 | Epoch: 22, Iteration: 180 loss: 0.494 1164 | Epoch: 22, Iteration: 200 loss: 0.517 1165 | Epoch: 22, Iteration: 220 loss: 0.555 1166 | Epoch: 22, Iteration: 240 loss: 0.534 1167 | Epoch: 22, Iteration: 260 loss: 0.552 1168 | Epoch: 22, Iteration: 280 loss: 0.580 1169 | Epoch: 22, Iteration: 300 loss: 0.542 1170 | Epoch: 22, Iteration: 320 loss: 0.546 1171 | Epoch: 22, Iteration: 340 loss: 0.548 1172 | Epoch: 22, Iteration: 360 loss: 0.562 1173 | Epoch: 22, Iteration: 380 loss: 0.586 1174 | Epoch: 22, Iteration: 400 loss: 0.546 1175 | Epoch: 22, Iteration: 420 loss: 0.565 1176 | Epoch: 22, Iteration: 440 loss: 0.571 1177 | Epoch: 22, Iteration: 460 loss: 0.583 1178 | Epoch: 22, Iteration: 480 loss: 0.560 1179 | Epoch: 22, Iteration: 500 loss: 0.551 1180 | Epoch: 22, Iteration: 520 loss: 0.544 1181 | Epoch: 22, Iteration: 540 loss: 0.576 1182 | Epoch: 22, Iteration: 560 loss: 0.530 1183 | Epoch: 22, Iteration: 580 loss: 0.522 1184 | Epoch: 22, Iteration: 600 loss: 0.528 1185 | Epoch: 22, Iteration: 620 loss: 0.575 1186 | Epoch: 22, Iteration: 640 loss: 0.567 1187 | Epoch: 22, Iteration: 660 loss: 0.566 1188 | Epoch: 22, Iteration: 680 loss: 0.542 1189 | Epoch: 22, Iteration: 700 loss: 0.542 1190 | Epoch: 22, Iteration: 720 loss: 0.559 1191 | Epoch: 22, Iteration: 740 loss: 0.536 1192 | Epoch: 22, Iteration: 760 loss: 0.566 1193 | Epoch: 22, Iteration: 780 loss: 0.568 1194 | Loss: 1.108 1195 | Accuracy of the network on the 10000 test images: 63 % 1196 | Accuracy of plane : 75 % 1197 | Accuracy of car : 76 % 1198 | Accuracy of bird : 55 % 1199 | Accuracy of cat : 49 % 1200 | Accuracy of deer : 50 % 1201 | Accuracy of dog : 54 % 1202 | Accuracy of frog : 73 % 1203 | Accuracy of horse : 60 % 1204 | Accuracy of ship : 81 % 1205 | Accuracy of truck : 65 % 1206 | Epoch: 23, Iteration: 0 loss: 0.028 1207 | Epoch: 23, Iteration: 20 loss: 0.535 1208 | Epoch: 23, Iteration: 40 loss: 0.499 1209 | Epoch: 23, Iteration: 60 loss: 0.497 1210 | Epoch: 23, Iteration: 80 loss: 0.518 1211 | Epoch: 23, Iteration: 100 loss: 0.588 1212 | Epoch: 23, Iteration: 120 loss: 0.575 1213 | Epoch: 23, Iteration: 140 loss: 0.563 1214 | Epoch: 23, Iteration: 160 loss: 0.550 1215 | Epoch: 23, Iteration: 180 loss: 0.543 1216 | Epoch: 23, Iteration: 200 loss: 0.559 1217 | Epoch: 23, Iteration: 220 loss: 0.513 1218 | Epoch: 23, Iteration: 240 loss: 0.596 1219 | Epoch: 23, Iteration: 260 loss: 0.559 1220 | Epoch: 23, Iteration: 280 loss: 0.614 1221 | Epoch: 23, Iteration: 300 loss: 0.565 1222 | Epoch: 23, Iteration: 320 loss: 0.539 1223 | Epoch: 23, Iteration: 340 loss: 0.570 1224 | Epoch: 23, Iteration: 360 loss: 0.576 1225 | Epoch: 23, Iteration: 380 loss: 0.541 1226 | Epoch: 23, Iteration: 400 loss: 0.555 1227 | Epoch: 23, Iteration: 420 loss: 0.541 1228 | Epoch: 23, Iteration: 440 loss: 0.558 1229 | Epoch: 23, Iteration: 460 loss: 0.547 1230 | Epoch: 23, Iteration: 480 loss: 0.581 1231 | Epoch: 23, Iteration: 500 loss: 0.525 1232 | Epoch: 23, Iteration: 520 loss: 0.544 1233 | Epoch: 23, Iteration: 540 loss: 0.563 1234 | Epoch: 23, Iteration: 560 loss: 0.568 1235 | Epoch: 23, Iteration: 580 loss: 0.585 1236 | Epoch: 23, Iteration: 600 loss: 0.517 1237 | Epoch: 23, Iteration: 620 loss: 0.541 1238 | Epoch: 23, Iteration: 640 loss: 0.521 1239 | Epoch: 23, Iteration: 660 loss: 0.539 1240 | Epoch: 23, Iteration: 680 loss: 0.595 1241 | Epoch: 23, Iteration: 700 loss: 0.525 1242 | Epoch: 23, Iteration: 720 loss: 0.531 1243 | Epoch: 23, Iteration: 740 loss: 0.578 1244 | Epoch: 23, Iteration: 760 loss: 0.577 1245 | Epoch: 23, Iteration: 780 loss: 0.575 1246 | Loss: 1.108 1247 | Accuracy of the network on the 10000 test images: 63 % 1248 | Accuracy of plane : 75 % 1249 | Accuracy of car : 76 % 1250 | Accuracy of bird : 55 % 1251 | Accuracy of cat : 49 % 1252 | Accuracy of deer : 50 % 1253 | Accuracy of dog : 54 % 1254 | Accuracy of frog : 73 % 1255 | Accuracy of horse : 60 % 1256 | Accuracy of ship : 81 % 1257 | Accuracy of truck : 65 % 1258 | Epoch: 24, Iteration: 0 loss: 0.025 1259 | Epoch: 24, Iteration: 20 loss: 0.549 1260 | Epoch: 24, Iteration: 40 loss: 0.560 1261 | Epoch: 24, Iteration: 60 loss: 0.583 1262 | Epoch: 24, Iteration: 80 loss: 0.540 1263 | Epoch: 24, Iteration: 100 loss: 0.552 1264 | Epoch: 24, Iteration: 120 loss: 0.546 1265 | Epoch: 24, Iteration: 140 loss: 0.606 1266 | Epoch: 24, Iteration: 160 loss: 0.565 1267 | Epoch: 24, Iteration: 180 loss: 0.563 1268 | Epoch: 24, Iteration: 200 loss: 0.528 1269 | Epoch: 24, Iteration: 220 loss: 0.510 1270 | Epoch: 24, Iteration: 240 loss: 0.521 1271 | Epoch: 24, Iteration: 260 loss: 0.538 1272 | Epoch: 24, Iteration: 280 loss: 0.529 1273 | Epoch: 24, Iteration: 300 loss: 0.599 1274 | Epoch: 24, Iteration: 320 loss: 0.544 1275 | Epoch: 24, Iteration: 340 loss: 0.593 1276 | Epoch: 24, Iteration: 360 loss: 0.628 1277 | Epoch: 24, Iteration: 380 loss: 0.552 1278 | Epoch: 24, Iteration: 400 loss: 0.567 1279 | Epoch: 24, Iteration: 420 loss: 0.481 1280 | Epoch: 24, Iteration: 440 loss: 0.559 1281 | Epoch: 24, Iteration: 460 loss: 0.577 1282 | Epoch: 24, Iteration: 480 loss: 0.567 1283 | Epoch: 24, Iteration: 500 loss: 0.555 1284 | Epoch: 24, Iteration: 520 loss: 0.573 1285 | Epoch: 24, Iteration: 540 loss: 0.543 1286 | Epoch: 24, Iteration: 560 loss: 0.563 1287 | Epoch: 24, Iteration: 580 loss: 0.526 1288 | Epoch: 24, Iteration: 600 loss: 0.607 1289 | Epoch: 24, Iteration: 620 loss: 0.510 1290 | Epoch: 24, Iteration: 640 loss: 0.504 1291 | Epoch: 24, Iteration: 660 loss: 0.580 1292 | Epoch: 24, Iteration: 680 loss: 0.574 1293 | Epoch: 24, Iteration: 700 loss: 0.560 1294 | Epoch: 24, Iteration: 720 loss: 0.549 1295 | Epoch: 24, Iteration: 740 loss: 0.542 1296 | Epoch: 24, Iteration: 760 loss: 0.546 1297 | Epoch: 24, Iteration: 780 loss: 0.473 1298 | Loss: 1.108 1299 | Accuracy of the network on the 10000 test images: 63 % 1300 | Accuracy of plane : 75 % 1301 | Accuracy of car : 76 % 1302 | Accuracy of bird : 55 % 1303 | Accuracy of cat : 49 % 1304 | Accuracy of deer : 50 % 1305 | Accuracy of dog : 54 % 1306 | Accuracy of frog : 73 % 1307 | Accuracy of horse : 60 % 1308 | Accuracy of ship : 81 % 1309 | Accuracy of truck : 65 % 1310 | Finished Training 1311 | -------------------------------------------------------------------------------- /docs/no_min_lr/second.log: -------------------------------------------------------------------------------- 1 | Files already downloaded and verified 2 | Files already downloaded and verified 3 | Epoch: 0, Iteration: 0 loss: 0.115 4 | Epoch: 0, Iteration: 20 loss: 2.304 5 | Epoch: 0, Iteration: 40 loss: 2.304 6 | Epoch: 0, Iteration: 60 loss: 2.304 7 | Epoch: 0, Iteration: 80 loss: 2.302 8 | Epoch: 0, Iteration: 100 loss: 2.300 9 | Epoch: 0, Iteration: 120 loss: 2.297 10 | Epoch: 0, Iteration: 140 loss: 2.296 11 | Epoch: 0, Iteration: 160 loss: 2.284 12 | Epoch: 0, Iteration: 180 loss: 2.255 13 | Epoch: 0, Iteration: 200 loss: 2.146 14 | Epoch: 0, Iteration: 220 loss: 2.134 15 | Epoch: 0, Iteration: 240 loss: 2.136 16 | Epoch: 0, Iteration: 260 loss: 2.114 17 | Epoch: 0, Iteration: 280 loss: 2.096 18 | Epoch: 0, Iteration: 300 loss: 2.130 19 | Epoch: 0, Iteration: 320 loss: 2.167 20 | Epoch: 0, Iteration: 340 loss: 2.201 21 | Epoch: 0, Iteration: 360 loss: 2.111 22 | Epoch: 0, Iteration: 380 loss: 2.092 23 | Epoch: 0, Iteration: 400 loss: 2.054 24 | Epoch: 0, Iteration: 420 loss: 2.211 25 | Epoch: 0, Iteration: 440 loss: 2.192 26 | Epoch: 0, Iteration: 460 loss: 2.125 27 | Epoch: 0, Iteration: 480 loss: 2.062 28 | Epoch: 0, Iteration: 500 loss: 2.128 29 | Epoch: 0, Iteration: 520 loss: 2.152 30 | Epoch: 0, Iteration: 540 loss: 2.170 31 | Epoch: 0, Iteration: 560 loss: 2.132 32 | Epoch: 0, Iteration: 580 loss: 2.122 33 | Epoch: 0, Iteration: 600 loss: 2.147 34 | Epoch: 0, Iteration: 620 loss: 2.142 35 | Epoch: 0, Iteration: 640 loss: 2.000 36 | Epoch: 0, Iteration: 660 loss: 2.050 37 | Epoch: 0, Iteration: 680 loss: 1.988 38 | Epoch: 0, Iteration: 700 loss: 2.025 39 | Epoch: 0, Iteration: 720 loss: 2.076 40 | Epoch: 0, Iteration: 740 loss: 2.122 41 | Epoch: 0, Iteration: 760 loss: 2.139 42 | Epoch: 0, Iteration: 780 loss: 1.990 43 | Loss: 1.935 44 | Accuracy of the network on the 10000 test images: 24 % 45 | Accuracy of plane : 67 % 46 | Accuracy of car : 0 % 47 | Accuracy of bird : 0 % 48 | Accuracy of cat : 0 % 49 | Accuracy of deer : 20 % 50 | Accuracy of dog : 61 % 51 | Accuracy of frog : 32 % 52 | Accuracy of horse : 32 % 53 | Accuracy of ship : 32 % 54 | Accuracy of truck : 16 % 55 | Epoch: 1, Iteration: 0 loss: 0.094 56 | Epoch: 1, Iteration: 20 loss: 2.025 57 | Epoch: 1, Iteration: 40 loss: 1.955 58 | Epoch: 1, Iteration: 60 loss: 2.046 59 | Epoch: 1, Iteration: 80 loss: 1.964 60 | Epoch: 1, Iteration: 100 loss: 1.985 61 | Epoch: 1, Iteration: 120 loss: 2.057 62 | Epoch: 1, Iteration: 140 loss: 1.997 63 | Epoch: 1, Iteration: 160 loss: 1.959 64 | Epoch: 1, Iteration: 180 loss: 1.914 65 | Epoch: 1, Iteration: 200 loss: 1.912 66 | Epoch: 1, Iteration: 220 loss: 1.967 67 | Epoch: 1, Iteration: 240 loss: 1.995 68 | Epoch: 1, Iteration: 260 loss: 1.942 69 | Epoch: 1, Iteration: 280 loss: 1.981 70 | Epoch: 1, Iteration: 300 loss: 1.955 71 | Epoch: 1, Iteration: 320 loss: 1.899 72 | Epoch: 1, Iteration: 340 loss: 1.879 73 | Epoch: 1, Iteration: 360 loss: 1.880 74 | Epoch: 1, Iteration: 380 loss: 1.912 75 | Epoch: 1, Iteration: 400 loss: 1.908 76 | Epoch: 1, Iteration: 420 loss: 1.866 77 | Epoch: 1, Iteration: 440 loss: 1.970 78 | Epoch: 1, Iteration: 460 loss: 1.904 79 | Epoch: 1, Iteration: 480 loss: 1.931 80 | Epoch: 1, Iteration: 500 loss: 1.935 81 | Epoch: 1, Iteration: 520 loss: 1.811 82 | Epoch: 1, Iteration: 540 loss: 1.830 83 | Epoch: 1, Iteration: 560 loss: 1.846 84 | Epoch: 1, Iteration: 580 loss: 1.857 85 | Epoch: 1, Iteration: 600 loss: 1.866 86 | Epoch: 1, Iteration: 620 loss: 1.850 87 | Epoch: 1, Iteration: 640 loss: 1.834 88 | Epoch: 1, Iteration: 660 loss: 1.830 89 | Epoch: 1, Iteration: 680 loss: 1.879 90 | Epoch: 1, Iteration: 700 loss: 1.841 91 | Epoch: 1, Iteration: 720 loss: 1.811 92 | Epoch: 1, Iteration: 740 loss: 1.781 93 | Epoch: 1, Iteration: 760 loss: 1.831 94 | Epoch: 1, Iteration: 780 loss: 1.974 95 | Loss: 1.810 96 | Accuracy of the network on the 10000 test images: 29 % 97 | Accuracy of plane : 1 % 98 | Accuracy of car : 78 % 99 | Accuracy of bird : 0 % 100 | Accuracy of cat : 34 % 101 | Accuracy of deer : 20 % 102 | Accuracy of dog : 37 % 103 | Accuracy of frog : 28 % 104 | Accuracy of horse : 64 % 105 | Accuracy of ship : 25 % 106 | Accuracy of truck : 16 % 107 | Epoch: 2, Iteration: 0 loss: 0.096 108 | Epoch: 2, Iteration: 20 loss: 1.823 109 | Epoch: 2, Iteration: 40 loss: 1.829 110 | Epoch: 2, Iteration: 60 loss: 1.773 111 | Epoch: 2, Iteration: 80 loss: 1.772 112 | Epoch: 2, Iteration: 100 loss: 1.797 113 | Epoch: 2, Iteration: 120 loss: 1.843 114 | Epoch: 2, Iteration: 140 loss: 1.791 115 | Epoch: 2, Iteration: 160 loss: 1.811 116 | Epoch: 2, Iteration: 180 loss: 1.806 117 | Epoch: 2, Iteration: 200 loss: 1.851 118 | Epoch: 2, Iteration: 220 loss: 1.935 119 | Epoch: 2, Iteration: 240 loss: 1.929 120 | Epoch: 2, Iteration: 260 loss: 1.894 121 | Epoch: 2, Iteration: 280 loss: 1.816 122 | Epoch: 2, Iteration: 300 loss: 1.898 123 | Epoch: 2, Iteration: 320 loss: 1.915 124 | Epoch: 2, Iteration: 340 loss: 1.815 125 | Epoch: 2, Iteration: 360 loss: 1.843 126 | Epoch: 2, Iteration: 380 loss: 1.839 127 | Epoch: 2, Iteration: 400 loss: 1.697 128 | Epoch: 2, Iteration: 420 loss: 1.720 129 | Epoch: 2, Iteration: 440 loss: 1.679 130 | Epoch: 2, Iteration: 460 loss: 1.719 131 | Epoch: 2, Iteration: 480 loss: 1.679 132 | Epoch: 2, Iteration: 500 loss: 1.612 133 | Epoch: 2, Iteration: 520 loss: 1.672 134 | Epoch: 2, Iteration: 540 loss: 1.628 135 | Epoch: 2, Iteration: 560 loss: 1.637 136 | Epoch: 2, Iteration: 580 loss: 1.675 137 | Epoch: 2, Iteration: 600 loss: 1.773 138 | Epoch: 2, Iteration: 620 loss: 1.754 139 | Epoch: 2, Iteration: 640 loss: 1.664 140 | Epoch: 2, Iteration: 660 loss: 1.656 141 | Epoch: 2, Iteration: 680 loss: 1.677 142 | Epoch: 2, Iteration: 700 loss: 1.656 143 | Epoch: 2, Iteration: 720 loss: 1.708 144 | Epoch: 2, Iteration: 740 loss: 1.686 145 | Epoch: 2, Iteration: 760 loss: 1.656 146 | Epoch: 2, Iteration: 780 loss: 1.691 147 | Loss: 1.668 148 | Accuracy of the network on the 10000 test images: 38 % 149 | Accuracy of plane : 53 % 150 | Accuracy of car : 52 % 151 | Accuracy of bird : 12 % 152 | Accuracy of cat : 9 % 153 | Accuracy of deer : 7 % 154 | Accuracy of dog : 67 % 155 | Accuracy of frog : 10 % 156 | Accuracy of horse : 56 % 157 | Accuracy of ship : 63 % 158 | Accuracy of truck : 56 % 159 | Epoch: 3, Iteration: 0 loss: 0.074 160 | Epoch: 3, Iteration: 20 loss: 1.717 161 | Epoch: 3, Iteration: 40 loss: 1.806 162 | Epoch: 3, Iteration: 60 loss: 1.762 163 | Epoch: 3, Iteration: 80 loss: 1.798 164 | Epoch: 3, Iteration: 100 loss: 1.782 165 | Epoch: 3, Iteration: 120 loss: 1.871 166 | Epoch: 3, Iteration: 140 loss: 1.878 167 | Epoch: 3, Iteration: 160 loss: 1.674 168 | Epoch: 3, Iteration: 180 loss: 1.587 169 | Epoch: 3, Iteration: 200 loss: 1.631 170 | Epoch: 3, Iteration: 220 loss: 1.682 171 | Epoch: 3, Iteration: 240 loss: 1.673 172 | Epoch: 3, Iteration: 260 loss: 1.676 173 | Epoch: 3, Iteration: 280 loss: 1.550 174 | Epoch: 3, Iteration: 300 loss: 1.523 175 | Epoch: 3, Iteration: 320 loss: 1.669 176 | Epoch: 3, Iteration: 340 loss: 1.568 177 | Epoch: 3, Iteration: 360 loss: 1.510 178 | Epoch: 3, Iteration: 380 loss: 1.510 179 | Epoch: 3, Iteration: 400 loss: 1.625 180 | Epoch: 3, Iteration: 420 loss: 1.632 181 | Epoch: 3, Iteration: 440 loss: 1.573 182 | Epoch: 3, Iteration: 460 loss: 1.550 183 | Epoch: 3, Iteration: 480 loss: 1.491 184 | Epoch: 3, Iteration: 500 loss: 1.478 185 | Epoch: 3, Iteration: 520 loss: 1.505 186 | Epoch: 3, Iteration: 540 loss: 1.490 187 | Epoch: 3, Iteration: 560 loss: 1.453 188 | Epoch: 3, Iteration: 580 loss: 1.557 189 | Epoch: 3, Iteration: 600 loss: 1.544 190 | Epoch: 3, Iteration: 620 loss: 1.546 191 | Epoch: 3, Iteration: 640 loss: 1.548 192 | Epoch: 3, Iteration: 660 loss: 1.517 193 | Epoch: 3, Iteration: 680 loss: 1.571 194 | Epoch: 3, Iteration: 700 loss: 1.527 195 | Epoch: 3, Iteration: 720 loss: 1.594 196 | Epoch: 3, Iteration: 740 loss: 1.606 197 | Epoch: 3, Iteration: 760 loss: 1.576 198 | Epoch: 3, Iteration: 780 loss: 1.711 199 | Loss: 1.696 200 | Accuracy of the network on the 10000 test images: 38 % 201 | Accuracy of plane : 25 % 202 | Accuracy of car : 90 % 203 | Accuracy of bird : 40 % 204 | Accuracy of cat : 71 % 205 | Accuracy of deer : 1 % 206 | Accuracy of dog : 44 % 207 | Accuracy of frog : 0 % 208 | Accuracy of horse : 7 % 209 | Accuracy of ship : 63 % 210 | Accuracy of truck : 34 % 211 | Epoch: 4, Iteration: 0 loss: 0.080 212 | Epoch: 4, Iteration: 20 loss: 1.616 213 | Epoch: 4, Iteration: 40 loss: 1.483 214 | Epoch: 4, Iteration: 60 loss: 1.451 215 | Epoch: 4, Iteration: 80 loss: 1.427 216 | Epoch: 4, Iteration: 100 loss: 1.417 217 | Epoch: 4, Iteration: 120 loss: 1.487 218 | Epoch: 4, Iteration: 140 loss: 1.539 219 | Epoch: 4, Iteration: 160 loss: 1.535 220 | Epoch: 4, Iteration: 180 loss: 1.451 221 | Epoch: 4, Iteration: 200 loss: 1.387 222 | Epoch: 4, Iteration: 220 loss: 1.503 223 | Epoch: 4, Iteration: 240 loss: 1.449 224 | Epoch: 4, Iteration: 260 loss: 1.429 225 | Epoch: 4, Iteration: 280 loss: 1.446 226 | Epoch: 4, Iteration: 300 loss: 1.409 227 | Epoch: 4, Iteration: 320 loss: 1.362 228 | Epoch: 4, Iteration: 340 loss: 1.435 229 | Epoch: 4, Iteration: 360 loss: 1.388 230 | Epoch: 4, Iteration: 380 loss: 1.430 231 | Epoch: 4, Iteration: 400 loss: 1.441 232 | Epoch: 4, Iteration: 420 loss: 1.398 233 | Epoch: 4, Iteration: 440 loss: 1.358 234 | Epoch: 4, Iteration: 460 loss: 1.316 235 | Epoch: 4, Iteration: 480 loss: 1.384 236 | Epoch: 4, Iteration: 500 loss: 1.381 237 | Epoch: 4, Iteration: 520 loss: 1.469 238 | Epoch: 4, Iteration: 540 loss: 1.420 239 | Epoch: 4, Iteration: 560 loss: 1.465 240 | Epoch: 4, Iteration: 580 loss: 1.417 241 | Epoch: 4, Iteration: 600 loss: 1.362 242 | Epoch: 4, Iteration: 620 loss: 1.446 243 | Epoch: 4, Iteration: 640 loss: 1.436 244 | Epoch: 4, Iteration: 660 loss: 1.351 245 | Epoch: 4, Iteration: 680 loss: 1.275 246 | Epoch: 4, Iteration: 700 loss: 1.394 247 | Epoch: 4, Iteration: 720 loss: 1.361 248 | Epoch: 4, Iteration: 740 loss: 1.361 249 | Epoch: 4, Iteration: 760 loss: 1.385 250 | Epoch: 4, Iteration: 780 loss: 1.355 251 | Loss: 1.280 252 | Accuracy of the network on the 10000 test images: 57 % 253 | Accuracy of plane : 67 % 254 | Accuracy of car : 72 % 255 | Accuracy of bird : 45 % 256 | Accuracy of cat : 43 % 257 | Accuracy of deer : 23 % 258 | Accuracy of dog : 40 % 259 | Accuracy of frog : 78 % 260 | Accuracy of horse : 65 % 261 | Accuracy of ship : 68 % 262 | Accuracy of truck : 67 % 263 | Epoch: 5, Iteration: 0 loss: 0.057 264 | Epoch: 5, Iteration: 20 loss: 1.336 265 | Epoch: 5, Iteration: 40 loss: 1.338 266 | Epoch: 5, Iteration: 60 loss: 1.369 267 | Epoch: 5, Iteration: 80 loss: 1.334 268 | Epoch: 5, Iteration: 100 loss: 1.407 269 | Epoch: 5, Iteration: 120 loss: 1.410 270 | Epoch: 5, Iteration: 140 loss: 1.305 271 | Epoch: 5, Iteration: 160 loss: 1.365 272 | Epoch: 5, Iteration: 180 loss: 1.435 273 | Epoch: 5, Iteration: 200 loss: 1.422 274 | Epoch: 5, Iteration: 220 loss: 1.375 275 | Epoch: 5, Iteration: 240 loss: 1.332 276 | Epoch: 5, Iteration: 260 loss: 1.264 277 | Epoch: 5, Iteration: 280 loss: 1.334 278 | Epoch: 5, Iteration: 300 loss: 1.349 279 | Epoch: 5, Iteration: 320 loss: 1.323 280 | Epoch: 5, Iteration: 340 loss: 1.344 281 | Epoch: 5, Iteration: 360 loss: 1.317 282 | Epoch: 5, Iteration: 380 loss: 1.285 283 | Epoch: 5, Iteration: 400 loss: 1.227 284 | Epoch: 5, Iteration: 420 loss: 1.281 285 | Epoch: 5, Iteration: 440 loss: 1.268 286 | Epoch: 5, Iteration: 460 loss: 1.259 287 | Epoch: 5, Iteration: 480 loss: 1.246 288 | Epoch: 5, Iteration: 500 loss: 1.233 289 | Epoch: 5, Iteration: 520 loss: 1.243 290 | Epoch: 5, Iteration: 540 loss: 1.259 291 | Epoch: 5, Iteration: 560 loss: 1.257 292 | Epoch: 5, Iteration: 580 loss: 1.181 293 | Epoch: 5, Iteration: 600 loss: 1.328 294 | Epoch: 5, Iteration: 620 loss: 1.376 295 | Epoch: 5, Iteration: 640 loss: 1.312 296 | Epoch: 5, Iteration: 660 loss: 1.267 297 | Epoch: 5, Iteration: 680 loss: 1.255 298 | Epoch: 5, Iteration: 700 loss: 1.347 299 | Epoch: 5, Iteration: 720 loss: 1.300 300 | Epoch: 5, Iteration: 740 loss: 1.267 301 | Epoch: 5, Iteration: 760 loss: 1.293 302 | Epoch: 5, Iteration: 780 loss: 1.269 303 | Loss: 1.540 304 | Accuracy of the network on the 10000 test images: 44 % 305 | Accuracy of plane : 41 % 306 | Accuracy of car : 14 % 307 | Accuracy of bird : 37 % 308 | Accuracy of cat : 27 % 309 | Accuracy of deer : 16 % 310 | Accuracy of dog : 76 % 311 | Accuracy of frog : 32 % 312 | Accuracy of horse : 35 % 313 | Accuracy of ship : 82 % 314 | Accuracy of truck : 75 % 315 | Epoch: 6, Iteration: 0 loss: 0.076 316 | Epoch: 6, Iteration: 20 loss: 1.309 317 | Epoch: 6, Iteration: 40 loss: 1.249 318 | Epoch: 6, Iteration: 60 loss: 1.215 319 | Epoch: 6, Iteration: 80 loss: 1.251 320 | Epoch: 6, Iteration: 100 loss: 1.220 321 | Epoch: 6, Iteration: 120 loss: 1.200 322 | Epoch: 6, Iteration: 140 loss: 1.200 323 | Epoch: 6, Iteration: 160 loss: 1.244 324 | Epoch: 6, Iteration: 180 loss: 1.214 325 | Epoch: 6, Iteration: 200 loss: 1.157 326 | Epoch: 6, Iteration: 220 loss: 1.221 327 | Epoch: 6, Iteration: 240 loss: 1.190 328 | Epoch: 6, Iteration: 260 loss: 1.163 329 | Epoch: 6, Iteration: 280 loss: 1.143 330 | Epoch: 6, Iteration: 300 loss: 1.206 331 | Epoch: 6, Iteration: 320 loss: 1.193 332 | Epoch: 6, Iteration: 340 loss: 1.136 333 | Epoch: 6, Iteration: 360 loss: 1.229 334 | Epoch: 6, Iteration: 380 loss: 1.198 335 | Epoch: 6, Iteration: 400 loss: 1.188 336 | Epoch: 6, Iteration: 420 loss: 1.130 337 | Epoch: 6, Iteration: 440 loss: 1.199 338 | Epoch: 6, Iteration: 460 loss: 1.202 339 | Epoch: 6, Iteration: 480 loss: 1.165 340 | Epoch: 6, Iteration: 500 loss: 1.183 341 | Epoch: 6, Iteration: 520 loss: 1.202 342 | Epoch: 6, Iteration: 540 loss: 1.129 343 | Epoch: 6, Iteration: 560 loss: 1.171 344 | Epoch: 6, Iteration: 580 loss: 1.228 345 | Epoch: 6, Iteration: 600 loss: 1.171 346 | Epoch: 6, Iteration: 620 loss: 1.172 347 | Epoch: 6, Iteration: 640 loss: 1.138 348 | Epoch: 6, Iteration: 660 loss: 1.154 349 | Epoch: 6, Iteration: 680 loss: 1.111 350 | Epoch: 6, Iteration: 700 loss: 1.153 351 | Epoch: 6, Iteration: 720 loss: 1.119 352 | Epoch: 6, Iteration: 740 loss: 1.119 353 | Epoch: 6, Iteration: 760 loss: 1.116 354 | Epoch: 6, Iteration: 780 loss: 1.161 355 | Loss: 1.349 356 | Accuracy of the network on the 10000 test images: 50 % 357 | Accuracy of plane : 80 % 358 | Accuracy of car : 90 % 359 | Accuracy of bird : 34 % 360 | Accuracy of cat : 34 % 361 | Accuracy of deer : 9 % 362 | Accuracy of dog : 66 % 363 | Accuracy of frog : 50 % 364 | Accuracy of horse : 64 % 365 | Accuracy of ship : 55 % 366 | Accuracy of truck : 38 % 367 | Epoch 6: reducing learning rate of group 0 to 1.0000e-02. 368 | Epoch: 7, Iteration: 0 loss: 0.056 369 | Epoch: 7, Iteration: 20 loss: 1.191 370 | Epoch: 7, Iteration: 40 loss: 1.073 371 | Epoch: 7, Iteration: 60 loss: 0.949 372 | Epoch: 7, Iteration: 80 loss: 0.972 373 | Epoch: 7, Iteration: 100 loss: 0.923 374 | Epoch: 7, Iteration: 120 loss: 0.881 375 | Epoch: 7, Iteration: 140 loss: 0.890 376 | Epoch: 7, Iteration: 160 loss: 0.868 377 | Epoch: 7, Iteration: 180 loss: 0.899 378 | Epoch: 7, Iteration: 200 loss: 0.856 379 | Epoch: 7, Iteration: 220 loss: 0.926 380 | Epoch: 7, Iteration: 240 loss: 0.798 381 | Epoch: 7, Iteration: 260 loss: 0.808 382 | Epoch: 7, Iteration: 280 loss: 0.838 383 | Epoch: 7, Iteration: 300 loss: 0.770 384 | Epoch: 7, Iteration: 320 loss: 0.805 385 | Epoch: 7, Iteration: 340 loss: 0.778 386 | Epoch: 7, Iteration: 360 loss: 0.868 387 | Epoch: 7, Iteration: 380 loss: 0.840 388 | Epoch: 7, Iteration: 400 loss: 0.851 389 | Epoch: 7, Iteration: 420 loss: 0.811 390 | Epoch: 7, Iteration: 440 loss: 0.821 391 | Epoch: 7, Iteration: 460 loss: 0.740 392 | Epoch: 7, Iteration: 480 loss: 0.813 393 | Epoch: 7, Iteration: 500 loss: 0.787 394 | Epoch: 7, Iteration: 520 loss: 0.768 395 | Epoch: 7, Iteration: 540 loss: 0.775 396 | Epoch: 7, Iteration: 560 loss: 0.807 397 | Epoch: 7, Iteration: 580 loss: 0.831 398 | Epoch: 7, Iteration: 600 loss: 0.803 399 | Epoch: 7, Iteration: 620 loss: 0.778 400 | Epoch: 7, Iteration: 640 loss: 0.771 401 | Epoch: 7, Iteration: 660 loss: 0.728 402 | Epoch: 7, Iteration: 680 loss: 0.757 403 | Epoch: 7, Iteration: 700 loss: 0.727 404 | Epoch: 7, Iteration: 720 loss: 0.789 405 | Epoch: 7, Iteration: 740 loss: 0.774 406 | Epoch: 7, Iteration: 760 loss: 0.720 407 | Epoch: 7, Iteration: 780 loss: 0.773 408 | Loss: 1.080 409 | Accuracy of the network on the 10000 test images: 62 % 410 | Accuracy of plane : 75 % 411 | Accuracy of car : 76 % 412 | Accuracy of bird : 54 % 413 | Accuracy of cat : 49 % 414 | Accuracy of deer : 49 % 415 | Accuracy of dog : 50 % 416 | Accuracy of frog : 75 % 417 | Accuracy of horse : 62 % 418 | Accuracy of ship : 72 % 419 | Accuracy of truck : 67 % 420 | Epoch: 8, Iteration: 0 loss: 0.046 421 | Epoch: 8, Iteration: 20 loss: 0.764 422 | Epoch: 8, Iteration: 40 loss: 0.760 423 | Epoch: 8, Iteration: 60 loss: 0.747 424 | Epoch: 8, Iteration: 80 loss: 0.762 425 | Epoch: 8, Iteration: 100 loss: 0.725 426 | Epoch: 8, Iteration: 120 loss: 0.745 427 | Epoch: 8, Iteration: 140 loss: 0.695 428 | Epoch: 8, Iteration: 160 loss: 0.694 429 | Epoch: 8, Iteration: 180 loss: 0.743 430 | Epoch: 8, Iteration: 200 loss: 0.755 431 | Epoch: 8, Iteration: 220 loss: 0.715 432 | Epoch: 8, Iteration: 240 loss: 0.726 433 | Epoch: 8, Iteration: 260 loss: 0.697 434 | Epoch: 8, Iteration: 280 loss: 0.725 435 | Epoch: 8, Iteration: 300 loss: 0.674 436 | Epoch: 8, Iteration: 320 loss: 0.701 437 | Epoch: 8, Iteration: 340 loss: 0.734 438 | Epoch: 8, Iteration: 360 loss: 0.751 439 | Epoch: 8, Iteration: 380 loss: 0.736 440 | Epoch: 8, Iteration: 400 loss: 0.714 441 | Epoch: 8, Iteration: 420 loss: 0.706 442 | Epoch: 8, Iteration: 440 loss: 0.727 443 | Epoch: 8, Iteration: 460 loss: 0.735 444 | Epoch: 8, Iteration: 480 loss: 0.737 445 | Epoch: 8, Iteration: 500 loss: 0.721 446 | Epoch: 8, Iteration: 520 loss: 0.652 447 | Epoch: 8, Iteration: 540 loss: 0.661 448 | Epoch: 8, Iteration: 560 loss: 0.703 449 | Epoch: 8, Iteration: 580 loss: 0.697 450 | Epoch: 8, Iteration: 600 loss: 0.726 451 | Epoch: 8, Iteration: 620 loss: 0.735 452 | Epoch: 8, Iteration: 640 loss: 0.704 453 | Epoch: 8, Iteration: 660 loss: 0.716 454 | Epoch: 8, Iteration: 680 loss: 0.717 455 | Epoch: 8, Iteration: 700 loss: 0.700 456 | Epoch: 8, Iteration: 720 loss: 0.680 457 | Epoch: 8, Iteration: 740 loss: 0.731 458 | Epoch: 8, Iteration: 760 loss: 0.650 459 | Epoch: 8, Iteration: 780 loss: 0.675 460 | Loss: 1.085 461 | Accuracy of the network on the 10000 test images: 64 % 462 | Accuracy of plane : 75 % 463 | Accuracy of car : 76 % 464 | Accuracy of bird : 60 % 465 | Accuracy of cat : 41 % 466 | Accuracy of deer : 52 % 467 | Accuracy of dog : 55 % 468 | Accuracy of frog : 75 % 469 | Accuracy of horse : 64 % 470 | Accuracy of ship : 79 % 471 | Accuracy of truck : 69 % 472 | Epoch: 9, Iteration: 0 loss: 0.031 473 | Epoch: 9, Iteration: 20 loss: 0.627 474 | Epoch: 9, Iteration: 40 loss: 0.637 475 | Epoch: 9, Iteration: 60 loss: 0.711 476 | Epoch: 9, Iteration: 80 loss: 0.696 477 | Epoch: 9, Iteration: 100 loss: 0.674 478 | Epoch: 9, Iteration: 120 loss: 0.680 479 | Epoch: 9, Iteration: 140 loss: 0.596 480 | Epoch: 9, Iteration: 160 loss: 0.681 481 | Epoch: 9, Iteration: 180 loss: 0.648 482 | Epoch: 9, Iteration: 200 loss: 0.669 483 | Epoch: 9, Iteration: 220 loss: 0.580 484 | Epoch: 9, Iteration: 240 loss: 0.610 485 | Epoch: 9, Iteration: 260 loss: 0.688 486 | Epoch: 9, Iteration: 280 loss: 0.709 487 | Epoch: 9, Iteration: 300 loss: 0.675 488 | Epoch: 9, Iteration: 320 loss: 0.627 489 | Epoch: 9, Iteration: 340 loss: 0.596 490 | Epoch: 9, Iteration: 360 loss: 0.681 491 | Epoch: 9, Iteration: 380 loss: 0.660 492 | Epoch: 9, Iteration: 400 loss: 0.672 493 | Epoch: 9, Iteration: 420 loss: 0.628 494 | Epoch: 9, Iteration: 440 loss: 0.594 495 | Epoch: 9, Iteration: 460 loss: 0.638 496 | Epoch: 9, Iteration: 480 loss: 0.640 497 | Epoch: 9, Iteration: 500 loss: 0.626 498 | Epoch: 9, Iteration: 520 loss: 0.643 499 | Epoch: 9, Iteration: 540 loss: 0.625 500 | Epoch: 9, Iteration: 560 loss: 0.648 501 | Epoch: 9, Iteration: 580 loss: 0.610 502 | Epoch: 9, Iteration: 600 loss: 0.578 503 | Epoch: 9, Iteration: 620 loss: 0.638 504 | Epoch: 9, Iteration: 640 loss: 0.587 505 | Epoch: 9, Iteration: 660 loss: 0.628 506 | Epoch: 9, Iteration: 680 loss: 0.636 507 | Epoch: 9, Iteration: 700 loss: 0.649 508 | Epoch: 9, Iteration: 720 loss: 0.639 509 | Epoch: 9, Iteration: 740 loss: 0.640 510 | Epoch: 9, Iteration: 760 loss: 0.630 511 | Epoch: 9, Iteration: 780 loss: 0.598 512 | Loss: 1.100 513 | Accuracy of the network on the 10000 test images: 63 % 514 | Accuracy of plane : 73 % 515 | Accuracy of car : 72 % 516 | Accuracy of bird : 53 % 517 | Accuracy of cat : 53 % 518 | Accuracy of deer : 43 % 519 | Accuracy of dog : 52 % 520 | Accuracy of frog : 73 % 521 | Accuracy of horse : 65 % 522 | Accuracy of ship : 81 % 523 | Accuracy of truck : 69 % 524 | Epoch 9: reducing learning rate of group 0 to 1.0000e-03. 525 | Epoch: 10, Iteration: 0 loss: 0.035 526 | Epoch: 10, Iteration: 20 loss: 0.600 527 | Epoch: 10, Iteration: 40 loss: 0.549 528 | Epoch: 10, Iteration: 60 loss: 0.573 529 | Epoch: 10, Iteration: 80 loss: 0.581 530 | Epoch: 10, Iteration: 100 loss: 0.567 531 | Epoch: 10, Iteration: 120 loss: 0.609 532 | Epoch: 10, Iteration: 140 loss: 0.589 533 | Epoch: 10, Iteration: 160 loss: 0.555 534 | Epoch: 10, Iteration: 180 loss: 0.602 535 | Epoch: 10, Iteration: 200 loss: 0.550 536 | Epoch: 10, Iteration: 220 loss: 0.595 537 | Epoch: 10, Iteration: 240 loss: 0.581 538 | Epoch: 10, Iteration: 260 loss: 0.549 539 | Epoch: 10, Iteration: 280 loss: 0.531 540 | Epoch: 10, Iteration: 300 loss: 0.605 541 | Epoch: 10, Iteration: 320 loss: 0.607 542 | Epoch: 10, Iteration: 340 loss: 0.624 543 | Epoch: 10, Iteration: 360 loss: 0.584 544 | Epoch: 10, Iteration: 380 loss: 0.591 545 | Epoch: 10, Iteration: 400 loss: 0.552 546 | Epoch: 10, Iteration: 420 loss: 0.589 547 | Epoch: 10, Iteration: 440 loss: 0.564 548 | Epoch: 10, Iteration: 460 loss: 0.553 549 | Epoch: 10, Iteration: 480 loss: 0.581 550 | Epoch: 10, Iteration: 500 loss: 0.607 551 | Epoch: 10, Iteration: 520 loss: 0.524 552 | Epoch: 10, Iteration: 540 loss: 0.626 553 | Epoch: 10, Iteration: 560 loss: 0.557 554 | Epoch: 10, Iteration: 580 loss: 0.577 555 | Epoch: 10, Iteration: 600 loss: 0.587 556 | Epoch: 10, Iteration: 620 loss: 0.547 557 | Epoch: 10, Iteration: 640 loss: 0.588 558 | Epoch: 10, Iteration: 660 loss: 0.542 559 | Epoch: 10, Iteration: 680 loss: 0.603 560 | Epoch: 10, Iteration: 700 loss: 0.564 561 | Epoch: 10, Iteration: 720 loss: 0.576 562 | Epoch: 10, Iteration: 740 loss: 0.511 563 | Epoch: 10, Iteration: 760 loss: 0.573 564 | Epoch: 10, Iteration: 780 loss: 0.535 565 | Loss: 1.101 566 | Accuracy of the network on the 10000 test images: 62 % 567 | Accuracy of plane : 75 % 568 | Accuracy of car : 76 % 569 | Accuracy of bird : 54 % 570 | Accuracy of cat : 49 % 571 | Accuracy of deer : 49 % 572 | Accuracy of dog : 52 % 573 | Accuracy of frog : 73 % 574 | Accuracy of horse : 60 % 575 | Accuracy of ship : 81 % 576 | Accuracy of truck : 65 % 577 | Epoch: 11, Iteration: 0 loss: 0.024 578 | Epoch: 11, Iteration: 20 loss: 0.586 579 | Epoch: 11, Iteration: 40 loss: 0.552 580 | Epoch: 11, Iteration: 60 loss: 0.567 581 | Epoch: 11, Iteration: 80 loss: 0.550 582 | Epoch: 11, Iteration: 100 loss: 0.546 583 | Epoch: 11, Iteration: 120 loss: 0.555 584 | Epoch: 11, Iteration: 140 loss: 0.555 585 | Epoch: 11, Iteration: 160 loss: 0.596 586 | Epoch: 11, Iteration: 180 loss: 0.560 587 | Epoch: 11, Iteration: 200 loss: 0.589 588 | Epoch: 11, Iteration: 220 loss: 0.582 589 | Epoch: 11, Iteration: 240 loss: 0.610 590 | Epoch: 11, Iteration: 260 loss: 0.568 591 | Epoch: 11, Iteration: 280 loss: 0.560 592 | Epoch: 11, Iteration: 300 loss: 0.557 593 | Epoch: 11, Iteration: 320 loss: 0.548 594 | Epoch: 11, Iteration: 340 loss: 0.530 595 | Epoch: 11, Iteration: 360 loss: 0.553 596 | Epoch: 11, Iteration: 380 loss: 0.588 597 | Epoch: 11, Iteration: 400 loss: 0.584 598 | Epoch: 11, Iteration: 420 loss: 0.567 599 | Epoch: 11, Iteration: 440 loss: 0.565 600 | Epoch: 11, Iteration: 460 loss: 0.560 601 | Epoch: 11, Iteration: 480 loss: 0.504 602 | Epoch: 11, Iteration: 500 loss: 0.519 603 | Epoch: 11, Iteration: 520 loss: 0.537 604 | Epoch: 11, Iteration: 540 loss: 0.555 605 | Epoch: 11, Iteration: 560 loss: 0.542 606 | Epoch: 11, Iteration: 580 loss: 0.568 607 | Epoch: 11, Iteration: 600 loss: 0.553 608 | Epoch: 11, Iteration: 620 loss: 0.531 609 | Epoch: 11, Iteration: 640 loss: 0.571 610 | Epoch: 11, Iteration: 660 loss: 0.551 611 | Epoch: 11, Iteration: 680 loss: 0.561 612 | Epoch: 11, Iteration: 700 loss: 0.576 613 | Epoch: 11, Iteration: 720 loss: 0.605 614 | Epoch: 11, Iteration: 740 loss: 0.596 615 | Epoch: 11, Iteration: 760 loss: 0.593 616 | Epoch: 11, Iteration: 780 loss: 0.572 617 | Loss: 1.107 618 | Accuracy of the network on the 10000 test images: 63 % 619 | Accuracy of plane : 75 % 620 | Accuracy of car : 76 % 621 | Accuracy of bird : 55 % 622 | Accuracy of cat : 49 % 623 | Accuracy of deer : 50 % 624 | Accuracy of dog : 54 % 625 | Accuracy of frog : 73 % 626 | Accuracy of horse : 60 % 627 | Accuracy of ship : 81 % 628 | Accuracy of truck : 65 % 629 | Epoch 11: reducing learning rate of group 0 to 1.0000e-04. 630 | Epoch: 12, Iteration: 0 loss: 0.033 631 | Epoch: 12, Iteration: 20 loss: 0.583 632 | Epoch: 12, Iteration: 40 loss: 0.570 633 | Epoch: 12, Iteration: 60 loss: 0.556 634 | Epoch: 12, Iteration: 80 loss: 0.583 635 | Epoch: 12, Iteration: 100 loss: 0.577 636 | Epoch: 12, Iteration: 120 loss: 0.529 637 | Epoch: 12, Iteration: 140 loss: 0.523 638 | Epoch: 12, Iteration: 160 loss: 0.520 639 | Epoch: 12, Iteration: 180 loss: 0.528 640 | Epoch: 12, Iteration: 200 loss: 0.548 641 | Epoch: 12, Iteration: 220 loss: 0.567 642 | Epoch: 12, Iteration: 240 loss: 0.563 643 | Epoch: 12, Iteration: 260 loss: 0.548 644 | Epoch: 12, Iteration: 280 loss: 0.585 645 | Epoch: 12, Iteration: 300 loss: 0.551 646 | Epoch: 12, Iteration: 320 loss: 0.556 647 | Epoch: 12, Iteration: 340 loss: 0.526 648 | Epoch: 12, Iteration: 360 loss: 0.552 649 | Epoch: 12, Iteration: 380 loss: 0.541 650 | Epoch: 12, Iteration: 400 loss: 0.504 651 | Epoch: 12, Iteration: 420 loss: 0.524 652 | Epoch: 12, Iteration: 440 loss: 0.529 653 | Epoch: 12, Iteration: 460 loss: 0.544 654 | Epoch: 12, Iteration: 480 loss: 0.574 655 | Epoch: 12, Iteration: 500 loss: 0.556 656 | Epoch: 12, Iteration: 520 loss: 0.581 657 | Epoch: 12, Iteration: 540 loss: 0.573 658 | Epoch: 12, Iteration: 560 loss: 0.595 659 | Epoch: 12, Iteration: 580 loss: 0.556 660 | Epoch: 12, Iteration: 600 loss: 0.556 661 | Epoch: 12, Iteration: 620 loss: 0.521 662 | Epoch: 12, Iteration: 640 loss: 0.526 663 | Epoch: 12, Iteration: 660 loss: 0.572 664 | Epoch: 12, Iteration: 680 loss: 0.561 665 | Epoch: 12, Iteration: 700 loss: 0.560 666 | Epoch: 12, Iteration: 720 loss: 0.586 667 | Epoch: 12, Iteration: 740 loss: 0.594 668 | Epoch: 12, Iteration: 760 loss: 0.575 669 | Epoch: 12, Iteration: 780 loss: 0.547 670 | Loss: 1.108 671 | Accuracy of the network on the 10000 test images: 63 % 672 | Accuracy of plane : 75 % 673 | Accuracy of car : 76 % 674 | Accuracy of bird : 55 % 675 | Accuracy of cat : 49 % 676 | Accuracy of deer : 50 % 677 | Accuracy of dog : 54 % 678 | Accuracy of frog : 73 % 679 | Accuracy of horse : 60 % 680 | Accuracy of ship : 81 % 681 | Accuracy of truck : 65 % 682 | Epoch: 13, Iteration: 0 loss: 0.020 683 | Epoch: 13, Iteration: 20 loss: 0.549 684 | Epoch: 13, Iteration: 40 loss: 0.525 685 | Epoch: 13, Iteration: 60 loss: 0.541 686 | Epoch: 13, Iteration: 80 loss: 0.546 687 | Epoch: 13, Iteration: 100 loss: 0.536 688 | Epoch: 13, Iteration: 120 loss: 0.584 689 | Epoch: 13, Iteration: 140 loss: 0.588 690 | Epoch: 13, Iteration: 160 loss: 0.538 691 | Epoch: 13, Iteration: 180 loss: 0.580 692 | Epoch: 13, Iteration: 200 loss: 0.568 693 | Epoch: 13, Iteration: 220 loss: 0.565 694 | Epoch: 13, Iteration: 240 loss: 0.564 695 | Epoch: 13, Iteration: 260 loss: 0.553 696 | Epoch: 13, Iteration: 280 loss: 0.494 697 | Epoch: 13, Iteration: 300 loss: 0.571 698 | Epoch: 13, Iteration: 320 loss: 0.557 699 | Epoch: 13, Iteration: 340 loss: 0.579 700 | Epoch: 13, Iteration: 360 loss: 0.580 701 | Epoch: 13, Iteration: 380 loss: 0.544 702 | Epoch: 13, Iteration: 400 loss: 0.573 703 | Epoch: 13, Iteration: 420 loss: 0.501 704 | Epoch: 13, Iteration: 440 loss: 0.533 705 | Epoch: 13, Iteration: 460 loss: 0.581 706 | Epoch: 13, Iteration: 480 loss: 0.574 707 | Epoch: 13, Iteration: 500 loss: 0.555 708 | Epoch: 13, Iteration: 520 loss: 0.563 709 | Epoch: 13, Iteration: 540 loss: 0.551 710 | Epoch: 13, Iteration: 560 loss: 0.533 711 | Epoch: 13, Iteration: 580 loss: 0.621 712 | Epoch: 13, Iteration: 600 loss: 0.540 713 | Epoch: 13, Iteration: 620 loss: 0.538 714 | Epoch: 13, Iteration: 640 loss: 0.521 715 | Epoch: 13, Iteration: 660 loss: 0.567 716 | Epoch: 13, Iteration: 680 loss: 0.544 717 | Epoch: 13, Iteration: 700 loss: 0.521 718 | Epoch: 13, Iteration: 720 loss: 0.539 719 | Epoch: 13, Iteration: 740 loss: 0.585 720 | Epoch: 13, Iteration: 760 loss: 0.567 721 | Epoch: 13, Iteration: 780 loss: 0.543 722 | Loss: 1.108 723 | Accuracy of the network on the 10000 test images: 63 % 724 | Accuracy of plane : 75 % 725 | Accuracy of car : 76 % 726 | Accuracy of bird : 55 % 727 | Accuracy of cat : 49 % 728 | Accuracy of deer : 50 % 729 | Accuracy of dog : 54 % 730 | Accuracy of frog : 73 % 731 | Accuracy of horse : 60 % 732 | Accuracy of ship : 81 % 733 | Accuracy of truck : 65 % 734 | Epoch 13: reducing learning rate of group 0 to 1.0000e-05. 735 | Epoch: 14, Iteration: 0 loss: 0.027 736 | Epoch: 14, Iteration: 20 loss: 0.540 737 | Epoch: 14, Iteration: 40 loss: 0.544 738 | Epoch: 14, Iteration: 60 loss: 0.540 739 | Epoch: 14, Iteration: 80 loss: 0.541 740 | Epoch: 14, Iteration: 100 loss: 0.567 741 | Epoch: 14, Iteration: 120 loss: 0.550 742 | Epoch: 14, Iteration: 140 loss: 0.572 743 | Epoch: 14, Iteration: 160 loss: 0.514 744 | Epoch: 14, Iteration: 180 loss: 0.591 745 | Epoch: 14, Iteration: 200 loss: 0.530 746 | Epoch: 14, Iteration: 220 loss: 0.546 747 | Epoch: 14, Iteration: 240 loss: 0.599 748 | Epoch: 14, Iteration: 260 loss: 0.540 749 | Epoch: 14, Iteration: 280 loss: 0.545 750 | Epoch: 14, Iteration: 300 loss: 0.532 751 | Epoch: 14, Iteration: 320 loss: 0.520 752 | Epoch: 14, Iteration: 340 loss: 0.527 753 | Epoch: 14, Iteration: 360 loss: 0.538 754 | Epoch: 14, Iteration: 380 loss: 0.568 755 | Epoch: 14, Iteration: 400 loss: 0.567 756 | Epoch: 14, Iteration: 420 loss: 0.575 757 | Epoch: 14, Iteration: 440 loss: 0.563 758 | Epoch: 14, Iteration: 460 loss: 0.548 759 | Epoch: 14, Iteration: 480 loss: 0.575 760 | Epoch: 14, Iteration: 500 loss: 0.550 761 | Epoch: 14, Iteration: 520 loss: 0.575 762 | Epoch: 14, Iteration: 540 loss: 0.537 763 | Epoch: 14, Iteration: 560 loss: 0.539 764 | Epoch: 14, Iteration: 580 loss: 0.553 765 | Epoch: 14, Iteration: 600 loss: 0.567 766 | Epoch: 14, Iteration: 620 loss: 0.554 767 | Epoch: 14, Iteration: 640 loss: 0.544 768 | Epoch: 14, Iteration: 660 loss: 0.546 769 | Epoch: 14, Iteration: 680 loss: 0.570 770 | Epoch: 14, Iteration: 700 loss: 0.549 771 | Epoch: 14, Iteration: 720 loss: 0.609 772 | Epoch: 14, Iteration: 740 loss: 0.581 773 | Epoch: 14, Iteration: 760 loss: 0.565 774 | Epoch: 14, Iteration: 780 loss: 0.502 775 | Loss: 1.108 776 | Accuracy of the network on the 10000 test images: 63 % 777 | Accuracy of plane : 75 % 778 | Accuracy of car : 76 % 779 | Accuracy of bird : 55 % 780 | Accuracy of cat : 49 % 781 | Accuracy of deer : 50 % 782 | Accuracy of dog : 54 % 783 | Accuracy of frog : 73 % 784 | Accuracy of horse : 60 % 785 | Accuracy of ship : 81 % 786 | Accuracy of truck : 65 % 787 | Epoch: 15, Iteration: 0 loss: 0.026 788 | Epoch: 15, Iteration: 20 loss: 0.513 789 | Epoch: 15, Iteration: 40 loss: 0.566 790 | Epoch: 15, Iteration: 60 loss: 0.563 791 | Epoch: 15, Iteration: 80 loss: 0.531 792 | Epoch: 15, Iteration: 100 loss: 0.566 793 | Epoch: 15, Iteration: 120 loss: 0.529 794 | Epoch: 15, Iteration: 140 loss: 0.563 795 | Epoch: 15, Iteration: 160 loss: 0.579 796 | Epoch: 15, Iteration: 180 loss: 0.579 797 | Epoch: 15, Iteration: 200 loss: 0.544 798 | Epoch: 15, Iteration: 220 loss: 0.596 799 | Epoch: 15, Iteration: 240 loss: 0.561 800 | Epoch: 15, Iteration: 260 loss: 0.561 801 | Epoch: 15, Iteration: 280 loss: 0.578 802 | Epoch: 15, Iteration: 300 loss: 0.587 803 | Epoch: 15, Iteration: 320 loss: 0.586 804 | Epoch: 15, Iteration: 340 loss: 0.550 805 | Epoch: 15, Iteration: 360 loss: 0.573 806 | Epoch: 15, Iteration: 380 loss: 0.534 807 | Epoch: 15, Iteration: 400 loss: 0.555 808 | Epoch: 15, Iteration: 420 loss: 0.591 809 | Epoch: 15, Iteration: 440 loss: 0.555 810 | Epoch: 15, Iteration: 460 loss: 0.536 811 | Epoch: 15, Iteration: 480 loss: 0.550 812 | Epoch: 15, Iteration: 500 loss: 0.506 813 | Epoch: 15, Iteration: 520 loss: 0.552 814 | Epoch: 15, Iteration: 540 loss: 0.584 815 | Epoch: 15, Iteration: 560 loss: 0.522 816 | Epoch: 15, Iteration: 580 loss: 0.562 817 | Epoch: 15, Iteration: 600 loss: 0.549 818 | Epoch: 15, Iteration: 620 loss: 0.516 819 | Epoch: 15, Iteration: 640 loss: 0.539 820 | Epoch: 15, Iteration: 660 loss: 0.549 821 | Epoch: 15, Iteration: 680 loss: 0.531 822 | Epoch: 15, Iteration: 700 loss: 0.536 823 | Epoch: 15, Iteration: 720 loss: 0.556 824 | Epoch: 15, Iteration: 740 loss: 0.549 825 | Epoch: 15, Iteration: 760 loss: 0.533 826 | Epoch: 15, Iteration: 780 loss: 0.540 827 | Loss: 1.108 828 | Accuracy of the network on the 10000 test images: 63 % 829 | Accuracy of plane : 75 % 830 | Accuracy of car : 76 % 831 | Accuracy of bird : 55 % 832 | Accuracy of cat : 49 % 833 | Accuracy of deer : 50 % 834 | Accuracy of dog : 54 % 835 | Accuracy of frog : 73 % 836 | Accuracy of horse : 60 % 837 | Accuracy of ship : 81 % 838 | Accuracy of truck : 65 % 839 | Epoch 15: reducing learning rate of group 0 to 1.0000e-06. 840 | Epoch: 16, Iteration: 0 loss: 0.034 841 | Epoch: 16, Iteration: 20 loss: 0.551 842 | Epoch: 16, Iteration: 40 loss: 0.549 843 | Epoch: 16, Iteration: 60 loss: 0.516 844 | Epoch: 16, Iteration: 80 loss: 0.542 845 | Epoch: 16, Iteration: 100 loss: 0.576 846 | Epoch: 16, Iteration: 120 loss: 0.541 847 | Epoch: 16, Iteration: 140 loss: 0.525 848 | Epoch: 16, Iteration: 160 loss: 0.527 849 | Epoch: 16, Iteration: 180 loss: 0.550 850 | Epoch: 16, Iteration: 200 loss: 0.553 851 | Epoch: 16, Iteration: 220 loss: 0.595 852 | Epoch: 16, Iteration: 240 loss: 0.561 853 | Epoch: 16, Iteration: 260 loss: 0.572 854 | Epoch: 16, Iteration: 280 loss: 0.575 855 | Epoch: 16, Iteration: 300 loss: 0.533 856 | Epoch: 16, Iteration: 320 loss: 0.578 857 | Epoch: 16, Iteration: 340 loss: 0.511 858 | Epoch: 16, Iteration: 360 loss: 0.525 859 | Epoch: 16, Iteration: 380 loss: 0.504 860 | Epoch: 16, Iteration: 400 loss: 0.523 861 | Epoch: 16, Iteration: 420 loss: 0.505 862 | Epoch: 16, Iteration: 440 loss: 0.555 863 | Epoch: 16, Iteration: 460 loss: 0.571 864 | Epoch: 16, Iteration: 480 loss: 0.618 865 | Epoch: 16, Iteration: 500 loss: 0.575 866 | Epoch: 16, Iteration: 520 loss: 0.569 867 | Epoch: 16, Iteration: 540 loss: 0.574 868 | Epoch: 16, Iteration: 560 loss: 0.556 869 | Epoch: 16, Iteration: 580 loss: 0.562 870 | Epoch: 16, Iteration: 600 loss: 0.535 871 | Epoch: 16, Iteration: 620 loss: 0.587 872 | Epoch: 16, Iteration: 640 loss: 0.494 873 | Epoch: 16, Iteration: 660 loss: 0.588 874 | Epoch: 16, Iteration: 680 loss: 0.603 875 | Epoch: 16, Iteration: 700 loss: 0.568 876 | Epoch: 16, Iteration: 720 loss: 0.546 877 | Epoch: 16, Iteration: 740 loss: 0.507 878 | Epoch: 16, Iteration: 760 loss: 0.574 879 | Epoch: 16, Iteration: 780 loss: 0.562 880 | Loss: 1.108 881 | Accuracy of the network on the 10000 test images: 63 % 882 | Accuracy of plane : 75 % 883 | Accuracy of car : 76 % 884 | Accuracy of bird : 55 % 885 | Accuracy of cat : 49 % 886 | Accuracy of deer : 50 % 887 | Accuracy of dog : 54 % 888 | Accuracy of frog : 73 % 889 | Accuracy of horse : 60 % 890 | Accuracy of ship : 81 % 891 | Accuracy of truck : 65 % 892 | Epoch: 17, Iteration: 0 loss: 0.022 893 | Epoch: 17, Iteration: 20 loss: 0.531 894 | Epoch: 17, Iteration: 40 loss: 0.556 895 | Epoch: 17, Iteration: 60 loss: 0.545 896 | Epoch: 17, Iteration: 80 loss: 0.563 897 | Epoch: 17, Iteration: 100 loss: 0.515 898 | Epoch: 17, Iteration: 120 loss: 0.549 899 | Epoch: 17, Iteration: 140 loss: 0.548 900 | Epoch: 17, Iteration: 160 loss: 0.541 901 | Epoch: 17, Iteration: 180 loss: 0.516 902 | Epoch: 17, Iteration: 200 loss: 0.544 903 | Epoch: 17, Iteration: 220 loss: 0.567 904 | Epoch: 17, Iteration: 240 loss: 0.554 905 | Epoch: 17, Iteration: 260 loss: 0.576 906 | Epoch: 17, Iteration: 280 loss: 0.615 907 | Epoch: 17, Iteration: 300 loss: 0.555 908 | Epoch: 17, Iteration: 320 loss: 0.534 909 | Epoch: 17, Iteration: 340 loss: 0.530 910 | Epoch: 17, Iteration: 360 loss: 0.542 911 | Epoch: 17, Iteration: 380 loss: 0.569 912 | Epoch: 17, Iteration: 400 loss: 0.566 913 | Epoch: 17, Iteration: 420 loss: 0.552 914 | Epoch: 17, Iteration: 440 loss: 0.545 915 | Epoch: 17, Iteration: 460 loss: 0.573 916 | Epoch: 17, Iteration: 480 loss: 0.547 917 | Epoch: 17, Iteration: 500 loss: 0.585 918 | Epoch: 17, Iteration: 520 loss: 0.550 919 | Epoch: 17, Iteration: 540 loss: 0.569 920 | Epoch: 17, Iteration: 560 loss: 0.577 921 | Epoch: 17, Iteration: 580 loss: 0.568 922 | Epoch: 17, Iteration: 600 loss: 0.540 923 | Epoch: 17, Iteration: 620 loss: 0.535 924 | Epoch: 17, Iteration: 640 loss: 0.550 925 | Epoch: 17, Iteration: 660 loss: 0.521 926 | Epoch: 17, Iteration: 680 loss: 0.555 927 | Epoch: 17, Iteration: 700 loss: 0.544 928 | Epoch: 17, Iteration: 720 loss: 0.546 929 | Epoch: 17, Iteration: 740 loss: 0.576 930 | Epoch: 17, Iteration: 760 loss: 0.564 931 | Epoch: 17, Iteration: 780 loss: 0.558 932 | Loss: 1.108 933 | Accuracy of the network on the 10000 test images: 63 % 934 | Accuracy of plane : 75 % 935 | Accuracy of car : 76 % 936 | Accuracy of bird : 55 % 937 | Accuracy of cat : 49 % 938 | Accuracy of deer : 50 % 939 | Accuracy of dog : 54 % 940 | Accuracy of frog : 73 % 941 | Accuracy of horse : 60 % 942 | Accuracy of ship : 81 % 943 | Accuracy of truck : 65 % 944 | Epoch 17: reducing learning rate of group 0 to 1.0000e-07. 945 | Epoch: 18, Iteration: 0 loss: 0.028 946 | Epoch: 18, Iteration: 20 loss: 0.542 947 | Epoch: 18, Iteration: 40 loss: 0.533 948 | Epoch: 18, Iteration: 60 loss: 0.569 949 | Epoch: 18, Iteration: 80 loss: 0.528 950 | Epoch: 18, Iteration: 100 loss: 0.554 951 | Epoch: 18, Iteration: 120 loss: 0.496 952 | Epoch: 18, Iteration: 140 loss: 0.554 953 | Epoch: 18, Iteration: 160 loss: 0.607 954 | Epoch: 18, Iteration: 180 loss: 0.560 955 | Epoch: 18, Iteration: 200 loss: 0.557 956 | Epoch: 18, Iteration: 220 loss: 0.528 957 | Epoch: 18, Iteration: 240 loss: 0.516 958 | Epoch: 18, Iteration: 260 loss: 0.543 959 | Epoch: 18, Iteration: 280 loss: 0.544 960 | Epoch: 18, Iteration: 300 loss: 0.571 961 | Epoch: 18, Iteration: 320 loss: 0.554 962 | Epoch: 18, Iteration: 340 loss: 0.560 963 | Epoch: 18, Iteration: 360 loss: 0.525 964 | Epoch: 18, Iteration: 380 loss: 0.548 965 | Epoch: 18, Iteration: 400 loss: 0.572 966 | Epoch: 18, Iteration: 420 loss: 0.532 967 | Epoch: 18, Iteration: 440 loss: 0.555 968 | Epoch: 18, Iteration: 460 loss: 0.576 969 | Epoch: 18, Iteration: 480 loss: 0.530 970 | Epoch: 18, Iteration: 500 loss: 0.604 971 | Epoch: 18, Iteration: 520 loss: 0.572 972 | Epoch: 18, Iteration: 540 loss: 0.545 973 | Epoch: 18, Iteration: 560 loss: 0.511 974 | Epoch: 18, Iteration: 580 loss: 0.564 975 | Epoch: 18, Iteration: 600 loss: 0.561 976 | Epoch: 18, Iteration: 620 loss: 0.544 977 | Epoch: 18, Iteration: 640 loss: 0.527 978 | Epoch: 18, Iteration: 660 loss: 0.592 979 | Epoch: 18, Iteration: 680 loss: 0.573 980 | Epoch: 18, Iteration: 700 loss: 0.525 981 | Epoch: 18, Iteration: 720 loss: 0.583 982 | Epoch: 18, Iteration: 740 loss: 0.555 983 | Epoch: 18, Iteration: 760 loss: 0.584 984 | Epoch: 18, Iteration: 780 loss: 0.573 985 | Loss: 1.108 986 | Accuracy of the network on the 10000 test images: 63 % 987 | Accuracy of plane : 75 % 988 | Accuracy of car : 76 % 989 | Accuracy of bird : 55 % 990 | Accuracy of cat : 49 % 991 | Accuracy of deer : 50 % 992 | Accuracy of dog : 54 % 993 | Accuracy of frog : 73 % 994 | Accuracy of horse : 60 % 995 | Accuracy of ship : 81 % 996 | Accuracy of truck : 65 % 997 | Epoch: 19, Iteration: 0 loss: 0.021 998 | Epoch: 19, Iteration: 20 loss: 0.552 999 | Epoch: 19, Iteration: 40 loss: 0.545 1000 | Epoch: 19, Iteration: 60 loss: 0.540 1001 | Epoch: 19, Iteration: 80 loss: 0.578 1002 | Epoch: 19, Iteration: 100 loss: 0.547 1003 | Epoch: 19, Iteration: 120 loss: 0.569 1004 | Epoch: 19, Iteration: 140 loss: 0.569 1005 | Epoch: 19, Iteration: 160 loss: 0.554 1006 | Epoch: 19, Iteration: 180 loss: 0.505 1007 | Epoch: 19, Iteration: 200 loss: 0.527 1008 | Epoch: 19, Iteration: 220 loss: 0.527 1009 | Epoch: 19, Iteration: 240 loss: 0.608 1010 | Epoch: 19, Iteration: 260 loss: 0.495 1011 | Epoch: 19, Iteration: 280 loss: 0.551 1012 | Epoch: 19, Iteration: 300 loss: 0.513 1013 | Epoch: 19, Iteration: 320 loss: 0.517 1014 | Epoch: 19, Iteration: 340 loss: 0.529 1015 | Epoch: 19, Iteration: 360 loss: 0.566 1016 | Epoch: 19, Iteration: 380 loss: 0.585 1017 | Epoch: 19, Iteration: 400 loss: 0.543 1018 | Epoch: 19, Iteration: 420 loss: 0.563 1019 | Epoch: 19, Iteration: 440 loss: 0.596 1020 | Epoch: 19, Iteration: 460 loss: 0.565 1021 | Epoch: 19, Iteration: 480 loss: 0.595 1022 | Epoch: 19, Iteration: 500 loss: 0.540 1023 | Epoch: 19, Iteration: 520 loss: 0.577 1024 | Epoch: 19, Iteration: 540 loss: 0.568 1025 | Epoch: 19, Iteration: 560 loss: 0.538 1026 | Epoch: 19, Iteration: 580 loss: 0.532 1027 | Epoch: 19, Iteration: 600 loss: 0.572 1028 | Epoch: 19, Iteration: 620 loss: 0.592 1029 | Epoch: 19, Iteration: 640 loss: 0.586 1030 | Epoch: 19, Iteration: 660 loss: 0.528 1031 | Epoch: 19, Iteration: 680 loss: 0.568 1032 | Epoch: 19, Iteration: 700 loss: 0.531 1033 | Epoch: 19, Iteration: 720 loss: 0.555 1034 | Epoch: 19, Iteration: 740 loss: 0.580 1035 | Epoch: 19, Iteration: 760 loss: 0.548 1036 | Epoch: 19, Iteration: 780 loss: 0.521 1037 | Loss: 1.108 1038 | Accuracy of the network on the 10000 test images: 63 % 1039 | Accuracy of plane : 75 % 1040 | Accuracy of car : 76 % 1041 | Accuracy of bird : 55 % 1042 | Accuracy of cat : 49 % 1043 | Accuracy of deer : 50 % 1044 | Accuracy of dog : 54 % 1045 | Accuracy of frog : 73 % 1046 | Accuracy of horse : 60 % 1047 | Accuracy of ship : 81 % 1048 | Accuracy of truck : 65 % 1049 | Epoch 19: reducing learning rate of group 0 to 1.0000e-08. 1050 | Epoch: 20, Iteration: 0 loss: 0.028 1051 | Epoch: 20, Iteration: 20 loss: 0.546 1052 | Epoch: 20, Iteration: 40 loss: 0.523 1053 | Epoch: 20, Iteration: 60 loss: 0.582 1054 | Epoch: 20, Iteration: 80 loss: 0.577 1055 | Epoch: 20, Iteration: 100 loss: 0.540 1056 | Epoch: 20, Iteration: 120 loss: 0.526 1057 | Epoch: 20, Iteration: 140 loss: 0.571 1058 | Epoch: 20, Iteration: 160 loss: 0.591 1059 | Epoch: 20, Iteration: 180 loss: 0.508 1060 | Epoch: 20, Iteration: 200 loss: 0.559 1061 | Epoch: 20, Iteration: 220 loss: 0.556 1062 | Epoch: 20, Iteration: 240 loss: 0.572 1063 | Epoch: 20, Iteration: 260 loss: 0.552 1064 | Epoch: 20, Iteration: 280 loss: 0.519 1065 | Epoch: 20, Iteration: 300 loss: 0.569 1066 | Epoch: 20, Iteration: 320 loss: 0.558 1067 | Epoch: 20, Iteration: 340 loss: 0.558 1068 | Epoch: 20, Iteration: 360 loss: 0.559 1069 | Epoch: 20, Iteration: 380 loss: 0.491 1070 | Epoch: 20, Iteration: 400 loss: 0.536 1071 | Epoch: 20, Iteration: 420 loss: 0.528 1072 | Epoch: 20, Iteration: 440 loss: 0.539 1073 | Epoch: 20, Iteration: 460 loss: 0.597 1074 | Epoch: 20, Iteration: 480 loss: 0.580 1075 | Epoch: 20, Iteration: 500 loss: 0.525 1076 | Epoch: 20, Iteration: 520 loss: 0.545 1077 | Epoch: 20, Iteration: 540 loss: 0.543 1078 | Epoch: 20, Iteration: 560 loss: 0.587 1079 | Epoch: 20, Iteration: 580 loss: 0.573 1080 | Epoch: 20, Iteration: 600 loss: 0.522 1081 | Epoch: 20, Iteration: 620 loss: 0.568 1082 | Epoch: 20, Iteration: 640 loss: 0.558 1083 | Epoch: 20, Iteration: 660 loss: 0.559 1084 | Epoch: 20, Iteration: 680 loss: 0.568 1085 | Epoch: 20, Iteration: 700 loss: 0.558 1086 | Epoch: 20, Iteration: 720 loss: 0.520 1087 | Epoch: 20, Iteration: 740 loss: 0.590 1088 | Epoch: 20, Iteration: 760 loss: 0.539 1089 | Epoch: 20, Iteration: 780 loss: 0.572 1090 | Loss: 1.108 1091 | Accuracy of the network on the 10000 test images: 63 % 1092 | Accuracy of plane : 75 % 1093 | Accuracy of car : 76 % 1094 | Accuracy of bird : 55 % 1095 | Accuracy of cat : 49 % 1096 | Accuracy of deer : 50 % 1097 | Accuracy of dog : 54 % 1098 | Accuracy of frog : 73 % 1099 | Accuracy of horse : 60 % 1100 | Accuracy of ship : 81 % 1101 | Accuracy of truck : 65 % 1102 | Epoch: 21, Iteration: 0 loss: 0.032 1103 | Epoch: 21, Iteration: 20 loss: 0.555 1104 | Epoch: 21, Iteration: 40 loss: 0.575 1105 | Epoch: 21, Iteration: 60 loss: 0.566 1106 | Epoch: 21, Iteration: 80 loss: 0.577 1107 | Epoch: 21, Iteration: 100 loss: 0.539 1108 | Epoch: 21, Iteration: 120 loss: 0.512 1109 | Epoch: 21, Iteration: 140 loss: 0.543 1110 | Epoch: 21, Iteration: 160 loss: 0.574 1111 | Epoch: 21, Iteration: 180 loss: 0.562 1112 | Epoch: 21, Iteration: 200 loss: 0.533 1113 | Epoch: 21, Iteration: 220 loss: 0.525 1114 | Epoch: 21, Iteration: 240 loss: 0.565 1115 | Epoch: 21, Iteration: 260 loss: 0.518 1116 | Epoch: 21, Iteration: 280 loss: 0.575 1117 | Epoch: 21, Iteration: 300 loss: 0.556 1118 | Epoch: 21, Iteration: 320 loss: 0.544 1119 | Epoch: 21, Iteration: 340 loss: 0.561 1120 | Epoch: 21, Iteration: 360 loss: 0.567 1121 | Epoch: 21, Iteration: 380 loss: 0.540 1122 | Epoch: 21, Iteration: 400 loss: 0.554 1123 | Epoch: 21, Iteration: 420 loss: 0.520 1124 | Epoch: 21, Iteration: 440 loss: 0.561 1125 | Epoch: 21, Iteration: 460 loss: 0.552 1126 | Epoch: 21, Iteration: 480 loss: 0.535 1127 | Epoch: 21, Iteration: 500 loss: 0.605 1128 | Epoch: 21, Iteration: 520 loss: 0.559 1129 | Epoch: 21, Iteration: 540 loss: 0.594 1130 | Epoch: 21, Iteration: 560 loss: 0.533 1131 | Epoch: 21, Iteration: 580 loss: 0.542 1132 | Epoch: 21, Iteration: 600 loss: 0.561 1133 | Epoch: 21, Iteration: 620 loss: 0.545 1134 | Epoch: 21, Iteration: 640 loss: 0.545 1135 | Epoch: 21, Iteration: 660 loss: 0.534 1136 | Epoch: 21, Iteration: 680 loss: 0.593 1137 | Epoch: 21, Iteration: 700 loss: 0.574 1138 | Epoch: 21, Iteration: 720 loss: 0.534 1139 | Epoch: 21, Iteration: 740 loss: 0.562 1140 | Epoch: 21, Iteration: 760 loss: 0.518 1141 | Epoch: 21, Iteration: 780 loss: 0.552 1142 | Loss: 1.108 1143 | Accuracy of the network on the 10000 test images: 63 % 1144 | Accuracy of plane : 75 % 1145 | Accuracy of car : 76 % 1146 | Accuracy of bird : 55 % 1147 | Accuracy of cat : 49 % 1148 | Accuracy of deer : 50 % 1149 | Accuracy of dog : 54 % 1150 | Accuracy of frog : 73 % 1151 | Accuracy of horse : 60 % 1152 | Accuracy of ship : 81 % 1153 | Accuracy of truck : 65 % 1154 | Epoch: 22, Iteration: 0 loss: 0.035 1155 | Epoch: 22, Iteration: 20 loss: 0.547 1156 | Epoch: 22, Iteration: 40 loss: 0.573 1157 | Epoch: 22, Iteration: 60 loss: 0.570 1158 | Epoch: 22, Iteration: 80 loss: 0.529 1159 | Epoch: 22, Iteration: 100 loss: 0.591 1160 | Epoch: 22, Iteration: 120 loss: 0.530 1161 | Epoch: 22, Iteration: 140 loss: 0.583 1162 | Epoch: 22, Iteration: 160 loss: 0.528 1163 | Epoch: 22, Iteration: 180 loss: 0.579 1164 | Epoch: 22, Iteration: 200 loss: 0.493 1165 | Epoch: 22, Iteration: 220 loss: 0.554 1166 | Epoch: 22, Iteration: 240 loss: 0.547 1167 | Epoch: 22, Iteration: 260 loss: 0.546 1168 | Epoch: 22, Iteration: 280 loss: 0.584 1169 | Epoch: 22, Iteration: 300 loss: 0.493 1170 | Epoch: 22, Iteration: 320 loss: 0.519 1171 | Epoch: 22, Iteration: 340 loss: 0.619 1172 | Epoch: 22, Iteration: 360 loss: 0.570 1173 | Epoch: 22, Iteration: 380 loss: 0.533 1174 | Epoch: 22, Iteration: 400 loss: 0.581 1175 | Epoch: 22, Iteration: 420 loss: 0.573 1176 | Epoch: 22, Iteration: 440 loss: 0.566 1177 | Epoch: 22, Iteration: 460 loss: 0.546 1178 | Epoch: 22, Iteration: 480 loss: 0.540 1179 | Epoch: 22, Iteration: 500 loss: 0.534 1180 | Epoch: 22, Iteration: 520 loss: 0.535 1181 | Epoch: 22, Iteration: 540 loss: 0.549 1182 | Epoch: 22, Iteration: 560 loss: 0.563 1183 | Epoch: 22, Iteration: 580 loss: 0.597 1184 | Epoch: 22, Iteration: 600 loss: 0.531 1185 | Epoch: 22, Iteration: 620 loss: 0.519 1186 | Epoch: 22, Iteration: 640 loss: 0.542 1187 | Epoch: 22, Iteration: 660 loss: 0.563 1188 | Epoch: 22, Iteration: 680 loss: 0.527 1189 | Epoch: 22, Iteration: 700 loss: 0.531 1190 | Epoch: 22, Iteration: 720 loss: 0.602 1191 | Epoch: 22, Iteration: 740 loss: 0.568 1192 | Epoch: 22, Iteration: 760 loss: 0.545 1193 | Epoch: 22, Iteration: 780 loss: 0.559 1194 | Loss: 1.108 1195 | Accuracy of the network on the 10000 test images: 63 % 1196 | Accuracy of plane : 75 % 1197 | Accuracy of car : 76 % 1198 | Accuracy of bird : 55 % 1199 | Accuracy of cat : 49 % 1200 | Accuracy of deer : 50 % 1201 | Accuracy of dog : 54 % 1202 | Accuracy of frog : 73 % 1203 | Accuracy of horse : 60 % 1204 | Accuracy of ship : 81 % 1205 | Accuracy of truck : 65 % 1206 | Epoch: 23, Iteration: 0 loss: 0.026 1207 | Epoch: 23, Iteration: 20 loss: 0.548 1208 | Epoch: 23, Iteration: 40 loss: 0.571 1209 | Epoch: 23, Iteration: 60 loss: 0.526 1210 | Epoch: 23, Iteration: 80 loss: 0.586 1211 | Epoch: 23, Iteration: 100 loss: 0.541 1212 | Epoch: 23, Iteration: 120 loss: 0.588 1213 | Epoch: 23, Iteration: 140 loss: 0.521 1214 | Epoch: 23, Iteration: 160 loss: 0.569 1215 | Epoch: 23, Iteration: 180 loss: 0.565 1216 | Epoch: 23, Iteration: 200 loss: 0.552 1217 | Epoch: 23, Iteration: 220 loss: 0.546 1218 | Epoch: 23, Iteration: 240 loss: 0.536 1219 | Epoch: 23, Iteration: 260 loss: 0.533 1220 | Epoch: 23, Iteration: 280 loss: 0.549 1221 | Epoch: 23, Iteration: 300 loss: 0.625 1222 | Epoch: 23, Iteration: 320 loss: 0.565 1223 | Epoch: 23, Iteration: 340 loss: 0.564 1224 | Epoch: 23, Iteration: 360 loss: 0.580 1225 | Epoch: 23, Iteration: 380 loss: 0.527 1226 | Epoch: 23, Iteration: 400 loss: 0.578 1227 | Epoch: 23, Iteration: 420 loss: 0.571 1228 | Epoch: 23, Iteration: 440 loss: 0.518 1229 | Epoch: 23, Iteration: 460 loss: 0.537 1230 | Epoch: 23, Iteration: 480 loss: 0.564 1231 | Epoch: 23, Iteration: 500 loss: 0.543 1232 | Epoch: 23, Iteration: 520 loss: 0.580 1233 | Epoch: 23, Iteration: 540 loss: 0.547 1234 | Epoch: 23, Iteration: 560 loss: 0.581 1235 | Epoch: 23, Iteration: 580 loss: 0.545 1236 | Epoch: 23, Iteration: 600 loss: 0.526 1237 | Epoch: 23, Iteration: 620 loss: 0.528 1238 | Epoch: 23, Iteration: 640 loss: 0.541 1239 | Epoch: 23, Iteration: 660 loss: 0.532 1240 | Epoch: 23, Iteration: 680 loss: 0.528 1241 | Epoch: 23, Iteration: 700 loss: 0.543 1242 | Epoch: 23, Iteration: 720 loss: 0.535 1243 | Epoch: 23, Iteration: 740 loss: 0.569 1244 | Epoch: 23, Iteration: 760 loss: 0.545 1245 | Epoch: 23, Iteration: 780 loss: 0.566 1246 | Loss: 1.108 1247 | Accuracy of the network on the 10000 test images: 63 % 1248 | Accuracy of plane : 75 % 1249 | Accuracy of car : 76 % 1250 | Accuracy of bird : 55 % 1251 | Accuracy of cat : 49 % 1252 | Accuracy of deer : 50 % 1253 | Accuracy of dog : 54 % 1254 | Accuracy of frog : 73 % 1255 | Accuracy of horse : 60 % 1256 | Accuracy of ship : 81 % 1257 | Accuracy of truck : 65 % 1258 | Epoch: 24, Iteration: 0 loss: 0.029 1259 | Epoch: 24, Iteration: 20 loss: 0.552 1260 | Epoch: 24, Iteration: 40 loss: 0.599 1261 | Epoch: 24, Iteration: 60 loss: 0.565 1262 | Epoch: 24, Iteration: 80 loss: 0.529 1263 | Epoch: 24, Iteration: 100 loss: 0.563 1264 | Epoch: 24, Iteration: 120 loss: 0.550 1265 | Epoch: 24, Iteration: 140 loss: 0.571 1266 | Epoch: 24, Iteration: 160 loss: 0.575 1267 | Epoch: 24, Iteration: 180 loss: 0.557 1268 | Epoch: 24, Iteration: 200 loss: 0.502 1269 | Epoch: 24, Iteration: 220 loss: 0.552 1270 | Epoch: 24, Iteration: 240 loss: 0.541 1271 | Epoch: 24, Iteration: 260 loss: 0.569 1272 | Epoch: 24, Iteration: 280 loss: 0.566 1273 | Epoch: 24, Iteration: 300 loss: 0.555 1274 | Epoch: 24, Iteration: 320 loss: 0.535 1275 | Epoch: 24, Iteration: 340 loss: 0.588 1276 | Epoch: 24, Iteration: 360 loss: 0.586 1277 | Epoch: 24, Iteration: 380 loss: 0.563 1278 | Epoch: 24, Iteration: 400 loss: 0.542 1279 | Epoch: 24, Iteration: 420 loss: 0.573 1280 | Epoch: 24, Iteration: 440 loss: 0.545 1281 | Epoch: 24, Iteration: 460 loss: 0.539 1282 | Epoch: 24, Iteration: 480 loss: 0.555 1283 | Epoch: 24, Iteration: 500 loss: 0.506 1284 | Epoch: 24, Iteration: 520 loss: 0.535 1285 | Epoch: 24, Iteration: 540 loss: 0.570 1286 | Epoch: 24, Iteration: 560 loss: 0.579 1287 | Epoch: 24, Iteration: 580 loss: 0.524 1288 | Epoch: 24, Iteration: 600 loss: 0.574 1289 | Epoch: 24, Iteration: 620 loss: 0.534 1290 | Epoch: 24, Iteration: 640 loss: 0.541 1291 | Epoch: 24, Iteration: 660 loss: 0.522 1292 | Epoch: 24, Iteration: 680 loss: 0.578 1293 | Epoch: 24, Iteration: 700 loss: 0.540 1294 | Epoch: 24, Iteration: 720 loss: 0.546 1295 | Epoch: 24, Iteration: 740 loss: 0.548 1296 | Epoch: 24, Iteration: 760 loss: 0.556 1297 | Epoch: 24, Iteration: 780 loss: 0.534 1298 | Loss: 1.108 1299 | Accuracy of the network on the 10000 test images: 63 % 1300 | Accuracy of plane : 75 % 1301 | Accuracy of car : 76 % 1302 | Accuracy of bird : 55 % 1303 | Accuracy of cat : 49 % 1304 | Accuracy of deer : 50 % 1305 | Accuracy of dog : 54 % 1306 | Accuracy of frog : 73 % 1307 | Accuracy of horse : 60 % 1308 | Accuracy of ship : 81 % 1309 | Accuracy of truck : 65 % 1310 | Finished Training 1311 | --------------------------------------------------------------------------------