def test_AverageMeter_str_smoke(self):
for show_local_avg in (True, False):
meter = optim.AverageMeter("test_average_meter",
show_local_avg=show_local_avg)
self.assertIsInstance(str(meter), str)
meter.update(0.0)
self.assertIsInstance(str(meter), str)
def test_AverageMeter_update_tensor(self):
actual_meter = optim.AverageMeter("actual_meter")
desired_meter = optim.AverageMeter("desired_meter")
torch.manual_seed(0)
vals = torch.rand(10)
actual_meter.update(vals)
desired_meter.update(vals.tolist())
for attr in (
"count",
"last_val",
"global_sum",
"global_min",
"global_max",
"global_avg",
"local_avg",
):
actual = getattr(actual_meter, attr)
desired = getattr(desired_meter, attr)
self.assertAlmostEqual(actual, desired)
def test_ProgressMeter_update(self):
meters = [optim.AverageMeter(str(idx)) for idx in range(3)]
progress_meter = optim.ProgressMeter(1, *meters)
vals = [float(val) for val in range(len(meters))]
progress_meter.update(
**{meter.name: val
for meter, val in zip(meters, vals)})
actual = progress_meter.count
desired = 1
self.assertEqual(actual, desired)
for meter, val in zip(meters, vals):
actual = meter.last_val
desired = val
self.assertFloatAlmostEqual(actual, desired)
def test_ProgressMeter_update():
meters = [optim.AverageMeter(str(idx)) for idx in range(3)]
progress_meter = optim.ProgressMeter(1, *meters)
vals = [float(val) for val in range(len(meters))]
progress_meter.update(
**{meter.name: val
for meter, val in zip(meters, vals)})
actual = progress_meter.count
desired = 1
assert actual == desired
for meter, val in zip(meters, vals):
actual = meter.last_val
desired = val
assert actual == pytest.approx(desired)