def setUp(self):
self.C = mlogger.Container()
self.metric_a = mlogger.metric.Simple()
self.config = mlogger.Config(entry="running_a_test")
self.CC = mlogger.Container(b=mlogger.metric.Average())
self.CCC = mlogger.Container(c=mlogger.metric.Timer(),
d=mlogger.metric.Maximum())
def setup_xp(args, model, optimizer):
env_name = args.xp_name.split('/')[-1]
if args.visdom:
plotter = mlogger.VisdomPlotter({'env': env_name, 'server': args.server, 'port': args.port})
else:
plotter = None
xp = mlogger.Container()
xp.config = mlogger.Config(plotter=plotter, **vars(args))
xp.epoch = mlogger.metric.Simple()
xp.train = mlogger.Container()
xp.train.acc = mlogger.metric.Average(plotter=plotter, plot_title="Accuracy", plot_legend="training")
xp.train.loss = mlogger.metric.Average(plotter=plotter, plot_title="Objective", plot_legend="loss")
xp.train.obj = mlogger.metric.Simple(plotter=plotter, plot_title="Objective", plot_legend="objective")
xp.train.reg = mlogger.metric.Simple(plotter=plotter, plot_title="Objective", plot_legend="regularization")
xp.train.weight_norm = mlogger.metric.Simple(plotter=plotter, plot_title="Weight-Norm")
xp.train.step_size = mlogger.metric.Average(plotter=plotter, plot_title="Step-Size", plot_legend="clipped")
xp.train.step_size_u = mlogger.metric.Average(plotter=plotter, plot_title="Step-Size", plot_legend="unclipped")
xp.train.timer = mlogger.metric.Timer(plotter=plotter, plot_title="Time", plot_legend='training')
xp.val = mlogger.Container()
xp.val.acc = mlogger.metric.Average(plotter=plotter, plot_title="Accuracy", plot_legend="validation")
xp.val.timer = mlogger.metric.Timer(plotter=plotter, plot_title="Time", plot_legend='validation')
xp.max_val = mlogger.metric.Maximum(plotter=plotter, plot_title="Accuracy", plot_legend='best-validation')
xp.test = mlogger.Container()
xp.test.acc = mlogger.metric.Average(plotter=plotter, plot_title="Accuracy", plot_legend="test")
xp.test.timer = mlogger.metric.Timer(plotter=plotter, plot_title="Time", plot_legend='test')
if args.dataset == "imagenet":
xp.train.acc5 = mlogger.metric.Average(plotter=plotter, plot_title="Accuracy@5", plot_legend="training")
xp.val.acc5 = mlogger.metric.Average(plotter=plotter, plot_title="Accuracy@5", plot_legend="validation")
xp.test.acc5 = mlogger.metric.Average(plotter=plotter, plot_title="Accuracy@5", plot_legend="test")
if args.visdom:
plotter.set_win_opts("Step-Size", {'ytype': 'log'})
plotter.set_win_opts("Objective", {'ytype': 'log'})
if args.log:
# log at each epoch
xp.epoch.hook_on_update(lambda: xp.save_to('{}/results.json'.format(args.xp_name)))
xp.epoch.hook_on_update(lambda: save_state(model, optimizer, '{}/model.pkl'.format(args.xp_name)))
# log after final evaluation on test set
xp.test.acc.hook_on_update(lambda: xp.save_to('{}/results.json'.format(args.xp_name)))
xp.test.acc.hook_on_update(lambda: save_state(model, optimizer, '{}/model.pkl'.format(args.xp_name)))
# save results and model for best validation performance
xp.max_val.hook_on_new_max(lambda: save_state(model, optimizer, '{}/best_model.pkl'.format(args.xp_name)))
return xp
def get_xp(args, optimizer):
if args.visdom:
plotter = mlogger.VisdomPlotter({
'env': args.xp_name,
'server': 'http://localhost',
'port': args.port
})
else:
plotter = None
xp = mlogger.Container()
xp.config = mlogger.Config(plotter=plotter, **vars(args))
xp.epoch = mlogger.metric.Simple()
xp.train = mlogger.Container()
xp.train.error = mlogger.metric.Average(plotter=plotter,
plot_title="Error",
plot_legend="train")
xp.train.obj = mlogger.metric.Average(plotter=plotter,
plot_title="Objective",
plot_legend="objective")
xp.train.timer = mlogger.metric.Timer(plotter=plotter,
plot_title="Time",
plot_legend='training')
xp.val = mlogger.Container()
xp.val.error = mlogger.metric.Average(plotter=plotter,
plot_title="Error",
plot_legend="validation")
xp.val.timer = mlogger.metric.Timer(plotter=plotter,
plot_title="Time",
plot_legend='validation')
xp.test = mlogger.Container()
xp.test.error = mlogger.metric.Average(plotter=plotter,
plot_title="Error",
plot_legend="test")
xp.test.timer = mlogger.metric.Timer(plotter=plotter,
plot_title="Time",
plot_legend='test')
return xp
def test_load_state_dict(self):
self.C.a = self.metric_a
self.C.conf = self.config
self.C.CC = self.CC
self.C.CC.CCC = self.CCC
self.metric_a.update(10)
self.CC.b.update(12)
self.CC.b.update(15)
self.CCC.c.update()
self.CCC.d.update(12)
self.CCC.d.update(15)
state = self.C.state_dict()
new_C = mlogger.Container()
new_C.load_state_dict(state)
self.assertDictEqual(self.C.state_dict(), new_C.state_dict())
for old, new in zip(self.C.children(), new_C.children()):
assert old is not new
assert isinstance(new, type(old))
self.assertDictEqual(old.state_dict(), new.state_dict())
use_visdom = True
lr = 0.01
n_epochs = 10
#----------------------------------------------------------
# Prepare logging
#----------------------------------------------------------
# log the hyperparameters of the experiment
if use_visdom:
plotter = mlogger.VisdomPlotter({'env': 'my_experiment', 'server': 'http://localhost', 'port': 8097},
manual_update=True)
else:
plotter = None
xp = mlogger.Container()
xp.config = mlogger.Config(plotter=plotter)
xp.config.update(lr=lr, n_epochs=n_epochs)
xp.epoch = mlogger.metric.Simple()
xp.train = mlogger.Container()
xp.train.acc1 = mlogger.metric.Average(plotter=plotter, plot_title="Accuracy@1", plot_legend="training")
xp.train.acck = mlogger.metric.Average(plotter=plotter, plot_title="Accuracy@k", plot_legend="training")
xp.train.loss = mlogger.metric.Average(plotter=plotter, plot_title="Objective")
xp.train.timer = mlogger.metric.Timer(plotter=plotter, plot_title="Time", plot_legend="training")
xp.val = mlogger.Container()
xp.val.acc1 = mlogger.metric.Average(plotter=plotter, plot_title="Accuracy@1", plot_legend="validation")
xp.val.acck = mlogger.metric.Average(plotter=plotter, plot_title="Accuracy@k", plot_legend="validation")