2022-04-16 12:20:44 +00:00
|
|
|
from __future__ import print_function
|
|
|
|
|
|
|
|
import os
|
|
|
|
import data
|
|
|
|
import utils
|
|
|
|
import torch
|
|
|
|
import pickle
|
|
|
|
import metrics
|
|
|
|
import argparse
|
|
|
|
import numpy as np
|
2023-06-07 06:51:07 +00:00
|
|
|
import amd_sample_draw
|
2022-04-16 12:20:44 +00:00
|
|
|
import config_bayesian as cfg
|
|
|
|
from datetime import datetime
|
|
|
|
from torch.nn import functional as F
|
|
|
|
from torch.optim import Adam, lr_scheduler
|
|
|
|
from models.BayesianModels.BayesianLeNet import BBBLeNet
|
|
|
|
from models.BayesianModels.BayesianAlexNet import BBBAlexNet
|
|
|
|
from models.BayesianModels.Bayesian3Conv3FC import BBB3Conv3FC
|
2023-06-01 08:20:51 +00:00
|
|
|
from stopping_crit import earlyStopping, energyBound, accuracyBound
|
2022-04-16 12:20:44 +00:00
|
|
|
|
|
|
|
# CUDA settings
|
2023-06-01 08:20:51 +00:00
|
|
|
device = torch.device("cuda:1" if torch.cuda.is_available() else "cpu")
|
2022-04-16 12:20:44 +00:00
|
|
|
|
|
|
|
def getModel(net_type, inputs, outputs, priors, layer_type, activation_type):
|
|
|
|
if (net_type == 'lenet'):
|
|
|
|
return BBBLeNet(outputs, inputs, priors, layer_type, activation_type,wide=cfg.wide)
|
|
|
|
elif (net_type == 'alexnet'):
|
|
|
|
return BBBAlexNet(outputs, inputs, priors, layer_type, activation_type)
|
|
|
|
elif (net_type == '3conv3fc'):
|
|
|
|
return BBB3Conv3FC(outputs, inputs, priors, layer_type, activation_type)
|
|
|
|
else:
|
|
|
|
raise ValueError('Network should be either [LeNet / AlexNet / 3Conv3FC')
|
|
|
|
|
|
|
|
|
|
|
|
def train_model(net, optimizer, criterion, trainloader, num_ens=1, beta_type=0.1, epoch=None, num_epochs=None):
|
|
|
|
net.train()
|
|
|
|
training_loss = 0.0
|
|
|
|
accs = []
|
|
|
|
kl_list = []
|
|
|
|
for i, (inputs, labels) in enumerate(trainloader, 1):
|
|
|
|
|
|
|
|
optimizer.zero_grad()
|
|
|
|
|
|
|
|
inputs, labels = inputs.to(device), labels.to(device)
|
|
|
|
outputs = torch.zeros(inputs.shape[0], net.num_classes, num_ens).to(device)
|
|
|
|
|
|
|
|
kl = 0.0
|
|
|
|
for j in range(num_ens):
|
|
|
|
net_out, _kl = net(inputs)
|
|
|
|
kl += _kl
|
|
|
|
outputs[:, :, j] = F.log_softmax(net_out, dim=1)
|
|
|
|
|
|
|
|
kl = kl / num_ens
|
|
|
|
kl_list.append(kl.item())
|
|
|
|
log_outputs = utils.logmeanexp(outputs, dim=2)
|
|
|
|
|
|
|
|
beta = metrics.get_beta(i-1, len(trainloader), beta_type, epoch, num_epochs)
|
|
|
|
loss = criterion(log_outputs, labels, kl, beta)
|
|
|
|
loss.backward()
|
|
|
|
optimizer.step()
|
|
|
|
|
|
|
|
accs.append(metrics.acc(log_outputs.data, labels))
|
|
|
|
training_loss += loss.cpu().data.numpy()
|
|
|
|
return training_loss/len(trainloader), np.mean(accs), np.mean(kl_list)
|
|
|
|
|
|
|
|
|
|
|
|
def validate_model(net, criterion, validloader, num_ens=1, beta_type=0.1, epoch=None, num_epochs=None):
|
|
|
|
"""Calculate ensemble accuracy and NLL Loss"""
|
|
|
|
net.train()
|
|
|
|
valid_loss = 0.0
|
|
|
|
accs = []
|
|
|
|
|
|
|
|
for i, (inputs, labels) in enumerate(validloader):
|
|
|
|
inputs, labels = inputs.to(device), labels.to(device)
|
|
|
|
outputs = torch.zeros(inputs.shape[0], net.num_classes, num_ens).to(device)
|
|
|
|
kl = 0.0
|
|
|
|
for j in range(num_ens):
|
|
|
|
net_out, _kl = net(inputs)
|
|
|
|
kl += _kl
|
|
|
|
outputs[:, :, j] = F.log_softmax(net_out, dim=1).data
|
|
|
|
|
|
|
|
log_outputs = utils.logmeanexp(outputs, dim=2)
|
|
|
|
|
|
|
|
beta = metrics.get_beta(i-1, len(validloader), beta_type, epoch, num_epochs)
|
|
|
|
valid_loss += criterion(log_outputs, labels, kl, beta).item()
|
|
|
|
accs.append(metrics.acc(log_outputs, labels))
|
|
|
|
|
|
|
|
return valid_loss/len(validloader), np.mean(accs)
|
|
|
|
|
|
|
|
|
|
|
|
def run(dataset, net_type):
|
|
|
|
|
|
|
|
# Hyper Parameter settings
|
|
|
|
layer_type = cfg.layer_type
|
|
|
|
activation_type = cfg.activation_type
|
|
|
|
priors = cfg.priors
|
|
|
|
|
|
|
|
train_ens = cfg.train_ens
|
|
|
|
valid_ens = cfg.valid_ens
|
|
|
|
n_epochs = cfg.n_epochs
|
|
|
|
lr_start = cfg.lr_start
|
|
|
|
num_workers = cfg.num_workers
|
|
|
|
valid_size = cfg.valid_size
|
|
|
|
batch_size = cfg.batch_size
|
|
|
|
beta_type = cfg.beta_type
|
|
|
|
|
|
|
|
trainset, testset, inputs, outputs = data.getDataset(dataset)
|
|
|
|
train_loader, valid_loader, test_loader = data.getDataloader(
|
|
|
|
trainset, testset, valid_size, batch_size, num_workers)
|
|
|
|
net = getModel(net_type, inputs, outputs, priors, layer_type, activation_type).to(device)
|
|
|
|
|
|
|
|
ckpt_dir = f'checkpoints/{dataset}/bayesian'
|
|
|
|
ckpt_name = f'checkpoints/{dataset}/bayesian/model_{net_type}_{layer_type}_{activation_type}_{cfg.wide}.pt'
|
|
|
|
|
|
|
|
if not os.path.exists(ckpt_dir):
|
|
|
|
os.makedirs(ckpt_dir, exist_ok=True)
|
|
|
|
|
2023-06-01 08:20:51 +00:00
|
|
|
with open("stp", "r") as file:
|
|
|
|
stp = int(file.read())
|
|
|
|
with open("sav", "r") as file:
|
|
|
|
sav = int(file.read())
|
|
|
|
|
2022-04-16 12:20:44 +00:00
|
|
|
criterion = metrics.ELBO(len(trainset)).to(device)
|
|
|
|
optimizer = Adam(net.parameters(), lr=lr_start)
|
|
|
|
lr_sched = lr_scheduler.ReduceLROnPlateau(optimizer, patience=6, verbose=True)
|
|
|
|
#valid_loss_max = np.Inf
|
2023-06-07 06:51:07 +00:00
|
|
|
#if stp == 2:
|
|
|
|
early_stop = []
|
2022-04-16 12:20:44 +00:00
|
|
|
train_data = []
|
|
|
|
for epoch in range(n_epochs): # loop over the dataset multiple times
|
|
|
|
|
|
|
|
train_loss, train_acc, train_kl = train_model(net, optimizer, criterion, train_loader, num_ens=train_ens, beta_type=beta_type, epoch=epoch, num_epochs=n_epochs)
|
|
|
|
valid_loss, valid_acc = validate_model(net, criterion, valid_loader, num_ens=valid_ens, beta_type=beta_type, epoch=epoch, num_epochs=n_epochs)
|
|
|
|
lr_sched.step(valid_loss)
|
|
|
|
|
|
|
|
|
|
|
|
train_data.append([epoch,train_loss,train_acc,valid_loss,valid_acc])
|
|
|
|
print('Epoch: {} \tTraining Loss: {:.4f} \tTraining Accuracy: {:.4f} \tValidation Loss: {:.4f} \tValidation Accuracy: {:.4f} \ttrain_kl_div: {:.4f}'.format(
|
|
|
|
epoch, train_loss, train_acc, valid_loss, valid_acc, train_kl))
|
|
|
|
|
2023-06-01 08:20:51 +00:00
|
|
|
if stp == 2:
|
2023-06-07 06:51:07 +00:00
|
|
|
#print('Using early stopping')
|
|
|
|
if earlyStopping(early_stop,train_acc,epoch,cfg.sens) == 1:
|
2023-06-01 08:20:51 +00:00
|
|
|
break
|
|
|
|
elif stp == 3:
|
2023-06-07 06:51:07 +00:00
|
|
|
#print('Using energy bound')
|
|
|
|
if energyBound(cfg.energy_thrs) == 1:
|
2023-06-01 08:20:51 +00:00
|
|
|
break
|
|
|
|
elif stp == 4:
|
2023-06-07 06:51:07 +00:00
|
|
|
#print('Using accuracy bound')
|
|
|
|
if accuracyBound(cfg.acc_thrs) == 1:
|
2023-06-01 08:20:51 +00:00
|
|
|
break
|
|
|
|
else:
|
|
|
|
print('Training for {} epochs'.format(cfg.n_epochs))
|
|
|
|
|
|
|
|
if sav == 1:
|
|
|
|
# save model when finished
|
2023-06-07 06:51:07 +00:00
|
|
|
if epoch == cfg.n_epochs-1:
|
2023-06-01 08:20:51 +00:00
|
|
|
torch.save(net.state_dict(), ckpt_name)
|
2022-04-16 12:20:44 +00:00
|
|
|
|
2023-06-07 06:51:07 +00:00
|
|
|
|
2022-04-16 12:20:44 +00:00
|
|
|
with open("bayes_exp_data_"+str(cfg.wide)+".pkl", 'wb') as f:
|
|
|
|
pickle.dump(train_data, f)
|
|
|
|
|
|
|
|
if __name__ == '__main__':
|
|
|
|
now = datetime.now()
|
|
|
|
current_time = now.strftime("%H:%M:%S")
|
|
|
|
print("Initial Time =", current_time)
|
|
|
|
parser = argparse.ArgumentParser(description = "PyTorch Bayesian Model Training")
|
|
|
|
parser.add_argument('--net_type', default='lenet', type=str, help='model')
|
|
|
|
parser.add_argument('--dataset', default='CIFAR10', type=str, help='dataset = [MNIST/CIFAR10/CIFAR100]')
|
|
|
|
args = parser.parse_args()
|
|
|
|
run(args.dataset, args.net_type)
|
|
|
|
now = datetime.now()
|
|
|
|
current_time = now.strftime("%H:%M:%S")
|
|
|
|
print("Final Time =", current_time)
|
|
|
|
|