def __init__(self, path, batch_size):
super().__init__(path, batch_size)
self.session_paths = []
self.stats = None
self.stat_list = []
for p in self.paths:
if "Session" in p:
self.session_paths.append(p)
elif "stats" in p:
with open(p, 'rb') as handle:
self.stat_list.append(pickle.load(handle))
if len(self.stat_list) == 0:
print("could not find stats!")
self.set_stats()
self.set_name(path)
self.classifier = model.get_classifier(int(
self.stats['num_classes'])).to(device)
self.session_paths = np.random.permutation(self.session_paths).tolist()
self.train_paths = self.session_paths[:TRAIN_SESSIONS]
self.val_paths = self.session_paths[TRAIN_SESSIONS:]
self.train_dataset = dataset_(self.train_paths)
self.val_dataset = dataset_(self.val_paths)
self.train_dataloader = DataLoader(self.train_dataset,
batch_size=batch_size,
shuffle=True)
self.val_dataloader = DataLoader(self.val_dataset,
batch_size=batch_size,
shuffle=True)
self.trainloader_iterator = iter(self.train_dataloader)
self.valloader_iterator = iter(self.val_dataloader)
def __init__(self, hparams):
super().__init__(hparams)
from model.utils import model_init
from model.utils import freeze
self.num = hparams["model_num"]
self.models = load_models(hparams)
assert len(self.models) == self.num
if not hparams["train_teachers"]:
freeze(self.models)
self.models.cuda()
self.large_gens = nn.ModuleList([self._make_large_generator(size * 4, size)
for size in [64, 128, 256, 512]])
self.small_gens = nn.ModuleList(
[nn.ModuleList([copy.copy(self._make_small_generator(size, size * 4)) for i in range(self.num)])
for size in [64, 128, 256, 512]])
if not hparams["train_gens"]:
freeze(self.large_gens)
freeze(self.small_gens)
else:
model_init(self.large_gens)
model_init(self.small_gens)
self.student = get_classifier(hparams["backbone"], hparams["dataset"])
if not hparams["train_student"]:
freeze(self.student)
else:
model_init(self.student)
from model.feature_similarity_measurement import cka_loss
self.feature_loss = cka_loss()
def __init__(self, hparams):
LightningModule.__init__(self, hparams)
self.scale = hparams['scale']
self.self_ensemble = hparams['self_ensemble']
self.model = get_classifier(hparams["backbone"], hparams["dataset"])
self.teacher = self.load_teacher()
self.dist_method = self.get_distillation_module()
def predict(date, time):
classifier = model.get_classifier()
file = open("../server/data/" + date, "r")
lines = file.readlines()
file.close()
line_size = model.FEATURE_SIZE + model.LABEL_PAD
if time == None:
last_lines = lines[line_size * -1:]
else:
for i in range(len(lines)):
if lines[i].startswith(time) and i >= line_size:
last_lines = lines[i-line_size:i]
break;
else:
return "Time is not valued"
q = []
for line in last_lines:
origin_item = data_filter.parse_line(line)
q.append(origin_item)
add_item = origin_item[:]
add_item[0] = q[-1:][0][0] + 1
q.append(add_item)
item = data_filter.cal_que(q)
if item is None:
return "Data is bad"
item.pop()
last_time = origin_item[0]
predict_x = {}
for i in range(len(item)):
predict_x["C" + str(i)] = []
predict_x["C" + str(i)].append(item[i])
predictions = classifier.predict(
input_fn=lambda:data_loader.eval_input_fn(predict_x,
labels=None,
batch_size=100))
for pred_dict in predictions:
template = ('"time":{}, "label":{}, "probability":{:.10f}')
class_id = pred_dict['class_ids'][0]
probability = pred_dict['probabilities'][class_id]
return "{" + template.format(last_time,
model.LABELS[class_id],
100 * probability) + "}"
def __init__(self, hparams):
super().__init__(hparams)
self.M_maps = []
self.teacher_model: ResNet_CIFAR = get_classifier(
hparams["backbone"], hparams["dataset"])
freeze(self.teacher_model.eval())
self.plane_model = nn.ModuleList()
self.teacher_start_layer = 0
self.last_channel = self.params['input_channel']
self.init_student()
def train_classifier():
train_generator = NoteIsoSequence(train_wav_files,
sample_duration=sample_duration,
sample_rate=sample_rate,
n_fft=n_fft,
batch_size=batch_size,
epsilon=epsilon,
song_indices=song_indices,
instr_indices=instr_indices,
note_indices=note_indices)
encoder, classifier = get_classifier()
classifier.summary()
test_generator = NoteIsoSequence(test_wav_files,
sample_duration=sample_duration,
sample_rate=sample_rate,
n_fft=n_fft,
batch_size=batch_size,
epsilon=epsilon,
song_indices=song_indices,
instr_indices=instr_indices,
note_indices=note_indices)
now = datetime.now()
log_dir = "logs/ae+-" + now.strftime("%Y-%m-%d-%H:%M:%S") + "/"
callbacks = [
TensorBoardWrapper(test_generator,
log_dir=log_dir,
nb_steps=5,
histogram_freq=1,
batch_size=batch_size,
write_graph=False,
write_grads=True,
write_images=False)
]
classifier.fit_generator(generator=train_generator,
validation_data=test_generator,
use_multiprocessing=use_multiprocessing,
workers=workers,
epochs=epochs,
steps_per_epoch=steps_per_epoch,
validation_steps=validation_steps,
callbacks=callbacks)
print("saving models...")
encoder.save("ae+/encoder-5.h")
classifier.save("ae+/classifier-5.h")
print("saved ae+.")
def __init__(self, hparams):
super().__init__(hparams)
from model.utils import model_init
from model.utils import freeze
self.models = nn.ModuleList(
[get_classifier(hparams["classifiers"][i], hparams["dataset"]) for i in range(self.num)])
for idx, model in enumerate(self.models):
checkpoint = torch.load(hparams["pretrain_paths"][idx], map_location='cpu')
model.load_state_dict({key[6:]: value for key, value in checkpoint["state_dict"].items()})
freeze(model)
model.cuda()
self.channels = [32, 64, 128, 256, 512]
assert 1 <= hparams["pos"] <= 5
self.gen0 = self._make_generator(self.channels[hparams["pos"]], hparams["gen_level"])
self.gen1 = self._make_generator(self.channels[hparams["pos"]], hparams["gen_level"])
model_init(self.gen0)
model_init(self.gen1)
def main(argv):
classifier = model.get_classifier()
items = [
[6.0,20.0426,5.0,1.6884,3.0,1.0,2.0,1.0999,0.0,1.0,6.0,20.0426,5.0,1.6884,3.0,1.0,2.0,1.0999,0.0,3.94,7.0,52.0,6.0,20.0426,5.0,3.9721,2.0,0.0999,0.0,2.94,7.0,53.0,6.0,20.0426,5.0,3.9721,2.0,0.0999,0.0,2.94,7.0,52.0,6.0,21.0426,5.0,3.9721,2.0,0.0999,1.0,2.94,7.0,52.0,6.0,21.0426,5.0,4.31,3.0,2.94,2.0,0.0999,7.0,51.0,6.0,20.0426,5.0,1.6884,4.0,5.88,2.0,0.0999,7.0,51.0,6.0,21.0426,5.0,23.6784,4.0,5.5616,2.0,0.0999,7.0,51.0,6.0,21.0426,5.0,1.6884,4.0,6.5616,2.0,0.0526,8.0,23.9563,7.0,52.0,6.0,20.0426,5.0,29.6884,4.0,6.3796,8.0,23.9563,7.0,51.0,6.0,20.0426,5.0,14.946,4.0,6.3638,8.0,23.9563,7.0,52.0,6.0,20.0426,5.0,14.946,4.0,4.706,8.0,34.9563,7.0,52.0,6.0,20.0426,5.0,2.6884,4.0,6.4943,7.0,52.0,6.0,20.0426,5.0,24.6684,4.0,6.338,3.0,0.8956,8.0,34.9563,7.0,52.0,6.0,20.0426,5.0,13.946,4.0,7.2696,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,1.6884,4.0,7.2543,8.0,34.9563,7.0,53.0,6.0,20.0426,5.0,1.6884,4.0,3.8636,8.0,34.9563,7.0,53.0,6.0,20.0426,5.0,16.5676,4.0,3.8553,8.0,34.9563,7.0,52.0,6.0,20.0426,5.0,4.31,4.0,3.8333,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,1.6884,4.0,3.8333,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,1.6884,4.0,3.9008,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,1.6884,4.0,4.711,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,1.6884,3.0,3.883,8.0,34.9563,7.0,52.0,6.0,20.0426,5.0,1.6884,3.0,3.8992,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,4.339,3.0,3.8724,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,33.319,3.0,3.8616,10.0,17.6222,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,5.4575,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,5.339,2.0,4.1805,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,1.8068,2.0,1.1828,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,6.0647,1.0,2.94,5]
,[6.0,20.0426,5.0,1.6884,3.0,1.0,2.0,1.0999,0.0,3.94,7.0,52.0,6.0,20.0426,5.0,3.9721,2.0,0.0999,0.0,2.94,7.0,53.0,6.0,20.0426,5.0,3.9721,2.0,0.0999,0.0,2.94,7.0,52.0,6.0,21.0426,5.0,3.9721,2.0,0.0999,1.0,2.94,7.0,52.0,6.0,21.0426,5.0,4.31,3.0,2.94,2.0,0.0999,7.0,51.0,6.0,20.0426,5.0,1.6884,4.0,5.88,2.0,0.0999,7.0,51.0,6.0,21.0426,5.0,23.6784,4.0,5.5616,2.0,0.0999,7.0,51.0,6.0,21.0426,5.0,1.6884,4.0,6.5616,2.0,0.0526,8.0,23.9563,7.0,52.0,6.0,20.0426,5.0,29.6884,4.0,6.3796,8.0,23.9563,7.0,51.0,6.0,20.0426,5.0,14.946,4.0,6.3638,8.0,23.9563,7.0,52.0,6.0,20.0426,5.0,14.946,4.0,4.706,8.0,34.9563,7.0,52.0,6.0,20.0426,5.0,2.6884,4.0,6.4943,7.0,52.0,6.0,20.0426,5.0,24.6684,4.0,6.338,3.0,0.8956,8.0,34.9563,7.0,52.0,6.0,20.0426,5.0,13.946,4.0,7.2696,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,1.6884,4.0,7.2543,8.0,34.9563,7.0,53.0,6.0,20.0426,5.0,1.6884,4.0,3.8636,8.0,34.9563,7.0,53.0,6.0,20.0426,5.0,16.5676,4.0,3.8553,8.0,34.9563,7.0,52.0,6.0,20.0426,5.0,4.31,4.0,3.8333,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,1.6884,4.0,3.8333,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,1.6884,4.0,3.9008,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,1.6884,4.0,4.711,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,1.6884,3.0,3.883,8.0,34.9563,7.0,52.0,6.0,20.0426,5.0,1.6884,3.0,3.8992,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,4.339,3.0,3.8724,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,33.319,3.0,3.8616,10.0,17.6222,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,5.4575,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,5.339,2.0,4.1805,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,1.8068,2.0,1.1828,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,6.0647,1.0,2.94,8.0,34.9563,7.0,51.0,6.0,20.0426,5.0,26.0447,2.0,1.1828,5]
,[9.0,23.9563,8.0,54.0723,7.0,20.0426,6.0,29.0447,4.0,2.94,9.0,23.9563,8.0,54.0723,7.0,20.0426,6.0,1.9584,4.0,2.94,9.0,23.9563,8.0,53.0723,7.0,21.0426,6.0,18.3223,4.0,2.94,9.0,23.9563,8.0,53.0723,7.0,21.0426,6.0,6.0647,4.0,2.94,8.0,53.0723,7.0,21.0426,6.0,6.0647,4.0,2.94,2.0,1.172,8.0,50.615,7.0,21.0426,6.0,6.0647,4.0,2.94,2.0,0.671,9.0,34.9563,8.0,51.0,7.0,20.0426,6.0,6.0647,4.0,2.94,8.0,51.82,7.0,20.0426,6.0,6.0647,4.0,2.94,2.0,0.6706,9.0,34.9563,8.0,51.82,7.0,20.0426,6.0,6.0647,2.0,2.94,9.0,34.9563,8.0,51.0,7.0,21.0681,6.0,6.0647,2.0,3.6103,11.0,17.0258,9.0,34.9563,8.0,51.0,7.0,43.0226,6.0,6.0647,8.0,51.0,7.0,20.0426,6.0,3.1918,5.0,0.0206,2.0,2.1176,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.0206,3.0,2.94,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.0206,3.0,2.2296,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.0206,2.0,2.94,9.0,34.9563,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.7599,9.0,34.9563,8.0,51.0,7.0,20.0426,6.0,14.4904,5.0,0.7599,8.0,51.0,7.0,20.0426,6.0,25.2128,5.0,0.7599,4.0,2.2396,8.0,51.0,7.0,20.0426,6.0,31.2128,5.0,0.7599,4.0,2.2396,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.7599,4.0,1.1623,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1623,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1623,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1612,8.0,50.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1612,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.2073,8.0,50.0,7.0,27.1503,6.0,24.6684,5.0,1.4599,4.0,1.2073,8.0,51.0,7.0,20.587,6.0,21.6684,5.0,2.1599,3.0,2.863,8.0,50.0,7.0,20.0426,6.0,22.2128,5.0,2.1599,3.0,2.814,8.0,50.0,7.0,21.0426,6.0,25.2128,5.0,2.1599,3.0,2.1136,8.0,51.0,7.0,20.0426,6.0,29.2328,5.0,2.1599,3.0,2.1136,5]
,[9.0,23.9563,8.0,54.0723,7.0,20.0426,6.0,1.9584,4.0,2.94,9.0,23.9563,8.0,53.0723,7.0,21.0426,6.0,18.3223,4.0,2.94,9.0,23.9563,8.0,53.0723,7.0,21.0426,6.0,6.0647,4.0,2.94,8.0,53.0723,7.0,21.0426,6.0,6.0647,4.0,2.94,2.0,1.172,8.0,50.615,7.0,21.0426,6.0,6.0647,4.0,2.94,2.0,0.671,9.0,34.9563,8.0,51.0,7.0,20.0426,6.0,6.0647,4.0,2.94,8.0,51.82,7.0,20.0426,6.0,6.0647,4.0,2.94,2.0,0.6706,9.0,34.9563,8.0,51.82,7.0,20.0426,6.0,6.0647,2.0,2.94,9.0,34.9563,8.0,51.0,7.0,21.0681,6.0,6.0647,2.0,3.6103,11.0,17.0258,9.0,34.9563,8.0,51.0,7.0,43.0226,6.0,6.0647,8.0,51.0,7.0,20.0426,6.0,3.1918,5.0,0.0206,2.0,2.1176,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.0206,3.0,2.94,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.0206,3.0,2.2296,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.0206,2.0,2.94,9.0,34.9563,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.7599,9.0,34.9563,8.0,51.0,7.0,20.0426,6.0,14.4904,5.0,0.7599,8.0,51.0,7.0,20.0426,6.0,25.2128,5.0,0.7599,4.0,2.2396,8.0,51.0,7.0,20.0426,6.0,31.2128,5.0,0.7599,4.0,2.2396,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.7599,4.0,1.1623,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1623,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1623,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1612,8.0,50.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1612,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.2073,8.0,50.0,7.0,27.1503,6.0,24.6684,5.0,1.4599,4.0,1.2073,8.0,51.0,7.0,20.587,6.0,21.6684,5.0,2.1599,3.0,2.863,8.0,50.0,7.0,20.0426,6.0,22.2128,5.0,2.1599,3.0,2.814,8.0,50.0,7.0,21.0426,6.0,25.2128,5.0,2.1599,3.0,2.1136,8.0,51.0,7.0,20.0426,6.0,29.2328,5.0,2.1599,3.0,2.1136,8.0,51.0,7.0,20.0426,6.0,14.4904,5.0,2.1599,3.0,1.4132,5]
,[9.0,23.9563,8.0,53.0723,7.0,21.0426,6.0,18.3223,4.0,2.94,9.0,23.9563,8.0,53.0723,7.0,21.0426,6.0,6.0647,4.0,2.94,8.0,53.0723,7.0,21.0426,6.0,6.0647,4.0,2.94,2.0,1.172,8.0,50.615,7.0,21.0426,6.0,6.0647,4.0,2.94,2.0,0.671,9.0,34.9563,8.0,51.0,7.0,20.0426,6.0,6.0647,4.0,2.94,8.0,51.82,7.0,20.0426,6.0,6.0647,4.0,2.94,2.0,0.6706,9.0,34.9563,8.0,51.82,7.0,20.0426,6.0,6.0647,2.0,2.94,9.0,34.9563,8.0,51.0,7.0,21.0681,6.0,6.0647,2.0,3.6103,11.0,17.0258,9.0,34.9563,8.0,51.0,7.0,43.0226,6.0,6.0647,8.0,51.0,7.0,20.0426,6.0,3.1918,5.0,0.0206,2.0,2.1176,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.0206,3.0,2.94,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.0206,3.0,2.2296,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.0206,2.0,2.94,9.0,34.9563,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.7599,9.0,34.9563,8.0,51.0,7.0,20.0426,6.0,14.4904,5.0,0.7599,8.0,51.0,7.0,20.0426,6.0,25.2128,5.0,0.7599,4.0,2.2396,8.0,51.0,7.0,20.0426,6.0,31.2128,5.0,0.7599,4.0,2.2396,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.7599,4.0,1.1623,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1623,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1623,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1612,8.0,50.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1612,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.2073,8.0,50.0,7.0,27.1503,6.0,24.6684,5.0,1.4599,4.0,1.2073,8.0,51.0,7.0,20.587,6.0,21.6684,5.0,2.1599,3.0,2.863,8.0,50.0,7.0,20.0426,6.0,22.2128,5.0,2.1599,3.0,2.814,8.0,50.0,7.0,21.0426,6.0,25.2128,5.0,2.1599,3.0,2.1136,8.0,51.0,7.0,20.0426,6.0,29.2328,5.0,2.1599,3.0,2.1136,8.0,51.0,7.0,20.0426,6.0,14.4904,5.0,2.1599,3.0,1.4132,9.0,23.9563,8.0,50.0,7.0,32.0426,6.0,27.2228,5.0,2.8599,5]
,[9.0,23.9563,8.0,53.0723,7.0,21.0426,6.0,6.0647,4.0,2.94,8.0,53.0723,7.0,21.0426,6.0,6.0647,4.0,2.94,2.0,1.172,8.0,50.615,7.0,21.0426,6.0,6.0647,4.0,2.94,2.0,0.671,9.0,34.9563,8.0,51.0,7.0,20.0426,6.0,6.0647,4.0,2.94,8.0,51.82,7.0,20.0426,6.0,6.0647,4.0,2.94,2.0,0.6706,9.0,34.9563,8.0,51.82,7.0,20.0426,6.0,6.0647,2.0,2.94,9.0,34.9563,8.0,51.0,7.0,21.0681,6.0,6.0647,2.0,3.6103,11.0,17.0258,9.0,34.9563,8.0,51.0,7.0,43.0226,6.0,6.0647,8.0,51.0,7.0,20.0426,6.0,3.1918,5.0,0.0206,2.0,2.1176,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.0206,3.0,2.94,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.0206,3.0,2.2296,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.0206,2.0,2.94,9.0,34.9563,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.7599,9.0,34.9563,8.0,51.0,7.0,20.0426,6.0,14.4904,5.0,0.7599,8.0,51.0,7.0,20.0426,6.0,25.2128,5.0,0.7599,4.0,2.2396,8.0,51.0,7.0,20.0426,6.0,31.2128,5.0,0.7599,4.0,2.2396,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.7599,4.0,1.1623,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1623,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1623,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1612,8.0,50.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1612,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.2073,8.0,50.0,7.0,27.1503,6.0,24.6684,5.0,1.4599,4.0,1.2073,8.0,51.0,7.0,20.587,6.0,21.6684,5.0,2.1599,3.0,2.863,8.0,50.0,7.0,20.0426,6.0,22.2128,5.0,2.1599,3.0,2.814,8.0,50.0,7.0,21.0426,6.0,25.2128,5.0,2.1599,3.0,2.1136,8.0,51.0,7.0,20.0426,6.0,29.2328,5.0,2.1599,3.0,2.1136,8.0,51.0,7.0,20.0426,6.0,14.4904,5.0,2.1599,3.0,1.4132,9.0,23.9563,8.0,50.0,7.0,32.0426,6.0,27.2228,5.0,2.8599,9.0,23.9563,8.0,50.0,7.0,32.0426,6.0,14.4904,5.0,4.376,5]
,[8.0,53.0723,7.0,21.0426,6.0,6.0647,4.0,2.94,2.0,1.172,8.0,50.615,7.0,21.0426,6.0,6.0647,4.0,2.94,2.0,0.671,9.0,34.9563,8.0,51.0,7.0,20.0426,6.0,6.0647,4.0,2.94,8.0,51.82,7.0,20.0426,6.0,6.0647,4.0,2.94,2.0,0.6706,9.0,34.9563,8.0,51.82,7.0,20.0426,6.0,6.0647,2.0,2.94,9.0,34.9563,8.0,51.0,7.0,21.0681,6.0,6.0647,2.0,3.6103,11.0,17.0258,9.0,34.9563,8.0,51.0,7.0,43.0226,6.0,6.0647,8.0,51.0,7.0,20.0426,6.0,3.1918,5.0,0.0206,2.0,2.1176,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.0206,3.0,2.94,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.0206,3.0,2.2296,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.0206,2.0,2.94,9.0,34.9563,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.7599,9.0,34.9563,8.0,51.0,7.0,20.0426,6.0,14.4904,5.0,0.7599,8.0,51.0,7.0,20.0426,6.0,25.2128,5.0,0.7599,4.0,2.2396,8.0,51.0,7.0,20.0426,6.0,31.2128,5.0,0.7599,4.0,2.2396,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.7599,4.0,1.1623,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1623,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1623,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1612,8.0,50.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1612,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.2073,8.0,50.0,7.0,27.1503,6.0,24.6684,5.0,1.4599,4.0,1.2073,8.0,51.0,7.0,20.587,6.0,21.6684,5.0,2.1599,3.0,2.863,8.0,50.0,7.0,20.0426,6.0,22.2128,5.0,2.1599,3.0,2.814,8.0,50.0,7.0,21.0426,6.0,25.2128,5.0,2.1599,3.0,2.1136,8.0,51.0,7.0,20.0426,6.0,29.2328,5.0,2.1599,3.0,2.1136,8.0,51.0,7.0,20.0426,6.0,14.4904,5.0,2.1599,3.0,1.4132,9.0,23.9563,8.0,50.0,7.0,32.0426,6.0,27.2228,5.0,2.8599,9.0,23.9563,8.0,50.0,7.0,32.0426,6.0,14.4904,5.0,4.376,8.0,50.0,7.0,32.0426,6.0,14.4904,5.0,5.9465,4.0,2.94,5]
,[8.0,50.615,7.0,21.0426,6.0,6.0647,4.0,2.94,2.0,0.671,9.0,34.9563,8.0,51.0,7.0,20.0426,6.0,6.0647,4.0,2.94,8.0,51.82,7.0,20.0426,6.0,6.0647,4.0,2.94,2.0,0.6706,9.0,34.9563,8.0,51.82,7.0,20.0426,6.0,6.0647,2.0,2.94,9.0,34.9563,8.0,51.0,7.0,21.0681,6.0,6.0647,2.0,3.6103,11.0,17.0258,9.0,34.9563,8.0,51.0,7.0,43.0226,6.0,6.0647,8.0,51.0,7.0,20.0426,6.0,3.1918,5.0,0.0206,2.0,2.1176,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.0206,3.0,2.94,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.0206,3.0,2.2296,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.0206,2.0,2.94,9.0,34.9563,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.7599,9.0,34.9563,8.0,51.0,7.0,20.0426,6.0,14.4904,5.0,0.7599,8.0,51.0,7.0,20.0426,6.0,25.2128,5.0,0.7599,4.0,2.2396,8.0,51.0,7.0,20.0426,6.0,31.2128,5.0,0.7599,4.0,2.2396,8.0,51.0,7.0,20.0426,6.0,2.2328,5.0,0.7599,4.0,1.1623,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1623,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1623,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1612,8.0,50.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.1612,8.0,51.0,7.0,20.587,6.0,1.6884,5.0,1.4599,4.0,1.2073,8.0,50.0,7.0,27.1503,6.0,24.6684,5.0,1.4599,4.0,1.2073,8.0,51.0,7.0,20.587,6.0,21.6684,5.0,2.1599,3.0,2.863,8.0,50.0,7.0,20.0426,6.0,22.2128,5.0,2.1599,3.0,2.814,8.0,50.0,7.0,21.0426,6.0,25.2128,5.0,2.1599,3.0,2.1136,8.0,51.0,7.0,20.0426,6.0,29.2328,5.0,2.1599,3.0,2.1136,8.0,51.0,7.0,20.0426,6.0,14.4904,5.0,2.1599,3.0,1.4132,9.0,23.9563,8.0,50.0,7.0,32.0426,6.0,27.2228,5.0,2.8599,9.0,23.9563,8.0,50.0,7.0,32.0426,6.0,14.4904,5.0,4.376,8.0,50.0,7.0,32.0426,6.0,14.4904,5.0,5.9465,4.0,2.94,8.0,51.0,7.0,31.0426,6.0,25.2128,5.0,4.376,4.0,3.8101,5]
]
predict_x = {}
expected = []
column_names = model.get_column_names();
for i in range(len(column_names)):
predict_x[column_names[i]] = []
for item in items:
predict_x[column_names[i]].append(item[i])
for item in items:
expected.append(item[len(item) - 1])
predictions = classifier.predict(
input_fn=lambda:data_loader.eval_input_fn(predict_x,
labels=None,
batch_size=100))
for pred_dict, expec in zip(predictions, expected):
template = ('predict: {},{:.10f}, expected: {}')
class_id = pred_dict['class_ids'][0]
probability = pred_dict['probabilities'][class_id]
print(template.format(model.LABELS[class_id],
100 * probability,
model.LABELS[expec]))
def __init__(self, path, batch_size):
super().__init__(path, batch_size)
self.batch_size = batch_size
self.train_paths = self.paths[0]
self.stats = {'num_classes': 4, 'num_channels': 22, 'length': 250}
self.val_paths = self.paths[1]
self.name = path[path.rfind('\\') + 4] + '\'' + 'Comp'
self.classifier = model.get_classifier(4).to(device)
self.classifier_id = 'Comp'
self.filter_id = 'Comp'
self.train_dataset = dataset_(self.train_paths)
self.val_dataset = dataset_(self.val_paths)
self.train_dataloader = DataLoader(self.train_dataset,
batch_size=batch_size,
shuffle=True)
self.val_dataloader = DataLoader(self.val_dataset,
batch_size=batch_size,
shuffle=True)
self.trainloader_iterator = iter(self.train_dataloader)
self.valloader_iterator = iter(self.val_dataloader)
def main(do_test_run=False,
resume_checkpoint=None,
run_method='alternate_train_and_validate',
**kwargs):
"""This is the main entry of the script.
It's main responsibility is to create an Engine object using the default
config that is modified by the kwargs given to this function.
By default it creates an Engine with the no preloaded weights for the model
and optimizer. When `resume_checkpoint` is set to the path to a checkpoint
file, this checkpoint file will be used.
By default, the `Engine#alternate_train_and_validate` method will be called.
This can be modified by the `run_method` parameter of this function.
"""
config.update(kwargs)
if os.path.exists(config['pwd']):
if input(f"{config['pwd']} already exists, are you sure? [y/N]: "
) != 'y':
return
print(f"config =\n{config}")
cnn_model = get_cnn(config)
classifier_model = get_classifier(config)
detective_model = get_detective(config)
if resume_checkpoint is not None:
info(f"loading state_dict for the engine from {resume_checkpoint} ...")
the_state_dict = torch.load(resume_checkpoint)
cnn_model.load_state_dict(the_state_dict['cnn_model'])
classifier_model.load_state_dict(the_state_dict['classifier_model'])
detective_model.load_state_dict(the_state_dict['detective_model'])
info(f"loaded state_dict for the engine from {resume_checkpoint}.")
engine = Engine(
cnn_model=cnn_model,
cnn_optimizer=(torch.optim.Adam, {
'lr': 3e-4
}),
classifier_model=classifier_model,
classifier_optimizer=(torch.optim.Adam, {
'lr': 3e-4
}),
classifier_loss_fn=nn.BCEWithLogitsLoss(),
detective_model=detective_model,
detective_optimizer=(torch.optim.Adam, {
'lr': 3e-4
}),
detective_loss_fn=nn.SmoothL1Loss(),
max_epochs=config['max_epochs'],
loader_pair=get_loader_pair(config),
plugins=[
Timestamp(),
TrainingMetrics(
classifier_metrics={
# 'loss': loss,
# 'acc': acc,
},
detective_metrics={
'loss': loss,
},
residual_factor=max(1 - config['train_batch_size'] / 10000, 0),
),
ValidationMetrics(
classifier_metrics={
# 'loss': loss,
# 'acc': acc,
# 'std': std,
# 'auc': auc,
},
detective_metrics={
'loss': loss,
'std': std,
},
),
ReduceLROnPlateau(),
Checkpoint(),
Messages(),
],
device=config['device'],
pwd=config['pwd'],
)
if do_test_run:
engine.__getattribute__(run_method)(test_run=True)
engine.reset()
engine.__getattribute__(run_method)()
def __init__(self, hparams
): # must name after hparams or there will be plenty of bugs
super().__init__(hparams)
self.model = get_classifier(hparams["backbone"], hparams["dataset"])
from model import get_classifier
from torch import nn
import torch
if __name__ == '__main__':
params = {
'arch': 'resnet20_layerwise',
}
model = get_classifier(params, "cifar10")
from model.basic_cifar_models.resnet_layerwise_cifar import BasicBlock_1, BasicBlock_2, ConvBNReLULayer
x_test = torch.randn((16, 3, 32, 32))
for layer in model.sequential_models:
if isinstance(layer, (BasicBlock_1, BasicBlock_2, ConvBNReLULayer)):
conv = layer.simplify_layer()[0]
assert isinstance(conv, nn.Conv2d)
with torch.no_grad():
ans = layer(x_test)
out = nn.ReLU()(conv(x_test))
diff_max = (ans - out).abs().max()
print("diff_max = ",
diff_max) # this should be smaller than 1e-5
x_test = layer(x_test)
# model.cuda().eval()
# x_test = torch.randn((128, 3, 32, 32)).float().cuda()
# with torch.no_grad():
# feats, outs = model(x_test, with_feature=True)
def main():
args = get_args()
if args.opts:
cfg.merge_from_list(args.opts)
cfg.freeze()
start_epoch = 0
checkpoint_dir = Path(args.checkpoint)
checkpoint_dir.mkdir(parents=True, exist_ok=True)
# display nb of workers
print(f"number of train workers {cfg.TRAIN.WORKERS}")
# fetch features
with open('train_features.pkl', 'rb') as f:
train_features = pickle.load(f)
with open('train_labels.pkl', 'rb') as f:
train_labels = pickle.load(f)
with open('valid_features.pkl', 'rb') as f:
valid_features = pickle.load(f)
with open('valid_labels.pkl', 'rb') as f:
valid_labels = pickle.load(f)
n_features = train_features.shape[1] * train_features.shape[
2] * train_features.shape[3]
n_classes = 101
# create model
if cfg.MODEL.ARCH_STYLE == 'classifier':
print("=> creating classifier")
model = get_classifier(n_features=n_features, n_classes=n_classes)
else:
print("=> creating regressor")
model = get_regressor(n_features=n_features, n_classes=n_classes)
# Create optimizer
if cfg.TRAIN.OPT == "sgd":
optimizer = torch.optim.SGD(model.parameters(),
lr=cfg.TRAIN.LR,
momentum=cfg.TRAIN.MOMENTUM,
weight_decay=cfg.TRAIN.WEIGHT_DECAY)
else:
optimizer = torch.optim.Adam(model.parameters(), lr=cfg.TRAIN.LR)
device = "cuda" if torch.cuda.is_available() else "cpu"
model = model.to(device)
# GPU config
if args.multi_gpu:
model = nn.DataParallel(model)
if device == "cuda":
cudnn.benchmark = True
# criterion
if cfg.MODEL.ARCH_STYLE == 'classifier':
if cfg.MODEL.SMOOTHING == True:
print("=> using label smoothing")
criterion = LabelSmoothingLoss(
std_smoothing=cfg.MODEL.STD_SMOOTHING,
n_classes=n_classes).to(device)
else:
criterion = nn.CrossEntropyLoss().to(device)
else:
if cfg.MODEL.ALEATORIC:
criterion = HeteroscedasticGaussianLoss().to(device)
else:
criterion = nn.L1Loss(reduction="sum").to(device)
# loaders
train_loader = get_feature_loader(train_features,
train_labels,
batch_size=cfg.TEST.BATCH_SIZE,
shuffle=True,
num_workers=cfg.TRAIN.WORKERS,
drop_last=True)
val_loader = get_feature_loader(valid_features,
valid_labels,
batch_size=cfg.TEST.BATCH_SIZE,
shuffle=False,
num_workers=cfg.TRAIN.WORKERS,
drop_last=False)
scheduler = StepLR(optimizer,
step_size=cfg.TRAIN.LR_DECAY_STEP,
gamma=cfg.TRAIN.LR_DECAY_RATE,
last_epoch=start_epoch - 1)
best_val_mae = 10000.0
train_writer = None
if args.tensorboard is not None:
opts_prefix = "_".join(args.opts)
train_writer = SummaryWriter(log_dir=args.tensorboard + "/" +
opts_prefix + "_train")
val_writer = SummaryWriter(log_dir=args.tensorboard + "/" +
opts_prefix + "_val")
print('=> Start training')
for epoch in range(start_epoch, cfg.TRAIN.EPOCHS):
# train
train_loss, train_acc = train(train_loader, model, criterion,
optimizer, epoch, device)
# validate
val_loss, val_acc, val_mae = validate(val_loader, model, criterion,
epoch, device)
if args.tensorboard is not None:
train_writer.add_scalar("loss", train_loss, epoch)
train_writer.add_scalar("acc", train_acc, epoch)
val_writer.add_scalar("loss", val_loss, epoch)
val_writer.add_scalar("acc", val_acc, epoch)
val_writer.add_scalar("mae", val_mae, epoch)
# checkpoint
if val_mae < best_val_mae:
print(
f"=> [epoch {epoch:03d}] best val mae was improved from {best_val_mae:.3f} to {val_mae:.3f}"
)
model_state_dict = model.module.state_dict(
) if args.multi_gpu else model.state_dict()
torch.save(
{
'epoch': epoch + 1,
'arch': cfg.MODEL.ARCH,
'state_dict': model_state_dict,
'optimizer_state_dict': optimizer.state_dict()
},
str(
checkpoint_dir.joinpath(
"epoch{:03d}_{:.5f}_{:.4f}.pth".format(
epoch, val_loss, val_mae))))
best_val_mae = val_mae
else:
print(
f"=> [epoch {epoch:03d}] best val mae was not improved from {best_val_mae:.3f} ({val_mae:.3f})"
)
# adjust learning rate
scheduler.step()
print("=> training finished")
print(f"additional opts: {args.opts}")
print(f"best val mae: {best_val_mae:.3f}")
def __init__(self, hparams):
super().__init__(hparams)
self.scale = hparams['scale']
self.self_ensemble = hparams['self_ensemble']
self.model = get_classifier(hparams["backbone"], hparams["dataset"])
def __init__(self,
task_batch_size,
data_batch_size,
shuffle=True,
train_size=0.80,
share=False,
train_task_types=None,
val_task_ids=None,
use_comp=False,
use_filters=False):
all_paths = glob.glob("data/*")
self.shuffle = shuffle
self.task_paths = []
for p in all_paths:
self.task_paths.append(p)
self.task_batch_size = task_batch_size
self.data_batch_size = data_batch_size
self.train_tasks = []
self.val_tasks = []
num_train_tasks = int(len(self.task_paths) * train_size)
if train_task_types is None and val_task_ids is None:
tasks = []
if shuffle:
self.task_paths = np.random.permutation(self.task_paths)
for p in self.task_paths:
if 'Comp' in p:
tasks.append(competetask(p, self.data_batch_size))
elif 'Large' in p:
tasks.append(largetask(p, self.data_batch_size))
self.train_tasks = tasks[:num_train_tasks]
self.val_tasks = tasks[num_train_tasks:]
else:
for p in self.task_paths:
cut_path = p[p.rfind('\\') + 1:]
task_id = cut_path[:2]
task_type = p[p.rfind('-') + 1:]
if int(task_id) in val_task_ids:
print("Val task ", task_id)
if 'Comp' in p:
if use_comp:
self.val_tasks.append(
competetask(p, self.data_batch_size))
else:
self.val_tasks.append(
largetask(p, self.data_batch_size))
elif task_type in train_task_types or train_task_types == "All":
print("Train task ", task_id)
if 'Comp' in p:
if use_comp:
self.train_tasks.append(
competetask(p, self.data_batch_size))
else:
self.train_tasks.append(
largetask(p, self.data_batch_size))
if shuffle:
self.train_tasks = np.random.permutation(
self.train_tasks).tolist()
self.val_tasks = np.random.permutation(self.val_tasks).tolist()
if share:
self.classifiers = {}
if use_filters:
self.filters = {}
for ta in self.train_tasks + self.val_tasks:
self.classifiers[ta.classifier_id] = model.get_classifier(
int(ta.stats['num_classes'])).to(device)
if use_filters:
self.filters[ta.filter_id] = model.Filter(
ta.stats['num_channels'], ta.stats['length'],
128).to(device)
for ta in self.train_tasks + self.val_tasks:
ta.classifier = self.classifiers[ta.classifier_id]
if use_filters:
ta.filter = self.filters[ta.filter_id]
print(self.classifiers)
if use_filters:
print(self.filters)
self.num_train_iters = int(
len(self.train_tasks) / self.task_batch_size)
self.num_val_iters = int(len(self.val_tasks) / self.task_batch_size)
