def __init__(self, config, install_dir):
AgentLogicBase.__init__(self, config)
self.dr = WinDataRetriver()
self.commandHandler = CommandHandlerWin()
hooks_dir = os.path.join(install_dir, 'hooks')
self.hooks = Hooks(logging.getLogger('Hooks'), hooks_dir)
set_bcd_useplatformclock()
def __init__(self, config, install_dir):
AgentLogicBase.__init__(self, config)
self.dr = WinDataRetriver()
self.commandHandler = CommandHandlerWin(self.dr)
hooks_dir = os.path.join(install_dir, 'hooks')
self.hooks = Hooks(logging.getLogger('Hooks'), hooks_dir)
if config.getboolean('general', 'apply_timer_configuration'):
apply_clock_tuning()
def __init__(self, config):
AgentLogicBase.__init__(self, config)
self.dr = LinuxDataRetriver()
self.dr.app_list = config.get("general", "applications_list")
self.dr.ignored_fs = set(config.get("general", "ignored_fs").split())
self.dr.ignore_zero_size_fs = config.get("general",
"ignore_zero_size_fs")
self.commandHandler = CommandHandlerLinux(self)
self.cred_server = CredServer()
self.hooks = Hooks(logging.getLogger('Hooks'),
_GUEST_HOOKS_CONFIG_PATH)
def main():
train_dataset = ImagenetteDataset(training=True)
val_dataset = ImagenetteDataset(training=False)
bs = 64
train_dataloader = DataLoader(train_dataset, shuffle=True, batch_size=bs)
val_dataloader = DataLoader(val_dataset, shuffle=False, batch_size=bs)
net = ConvNet(in_ch=3).cuda()
hooks = Hooks(net)
optim = torch.optim.SGD(net.parameters(), lr=0.5)
lossfxn = nn.CrossEntropyLoss()
for epoch in range(5):
pbar = tqdm(train_dataloader)
for batch in pbar:
optim.zero_grad()
imgs = batch['image'].cuda()
labels = batch['label'].cuda()
preds = net(imgs)
loss = lossfxn(preds, labels)
pbar.set_postfix({'Loss': float(loss)})
loss.backward()
optim.step()
hooks.show_me()
pbar = tqdm(val_dataloader)
total = 0
correct = 0
for batch in pbar:
imgs = batch['image'].cuda()
labels = batch['label'].cuda()
with torch.no_grad():
preds = net(imgs)
assert_shapes(labels, ['bs'], preds, ['bs', 10])
total += labels.numel()
correct += (labels == torch.argmax(preds, dim=1)).sum().item()
print(f'Top 1 accuracy is {correct/total}')
