def __init__(self, cfg):
super(DemoAgent, self).__init__(cfg)
use_cuda = self._use_cuda
in_size, out_size = None, None
# -- Get variables from cfg
train_cfg = cfg.train
self.max_grad_norm = train_cfg.max_grad_norm
# -- Initialize model
model_class = get_models(cfg.model)
self.model = model_class[0](cfg.model, in_size, out_size)
self._models.append(
self.model) # -- Add models & optimizers to base for saving
if use_cuda:
self.cuda()
# -- Initialize optimizers
self.optimizer = self.get_optim(cfg.train.algorithm,
cfg.train.algorithm_args, self.model)
self._optimizers.append(
self.optimizer) # -- Add models & optimizers to base for saving
# -- Initialize criterion
self.criterion = getattr(torch.nn, cfg.train.criterion)()
# -- Change settings
self.set_eval_metric_comparison(True)
super(DemoAgent, self).__end_init__()
def __init__(self, cfg):
super(ConditionalImitationAgent, self).__init__(cfg)
use_cuda = self._use_cuda # ++ Parent class already saves some configuration variables
# ++ All parent variables should start with _.
# -- Get necessary variables from cfg
self.cfg = cfg
# -- Initialize model
model_class = get_models(cfg.model)
self.model = model_class[0](cfg, cfg.data_info.image_shape, None)
# ++ All models receive as parameters (configuration namespace, input data size,
# ++ output data size)
self._models.append(
self.model) # -- Add models & optimizers to base for saving
# ++ After adding model you can set the agent to cuda mode
# ++ Parent class already makes some adjustments. E.g. turns model to cuda mode
if use_cuda:
self.cuda()
# -- Initialize optimizers
self.optimizer = self.get_optim(cfg.train.algorithm,
cfg.train.algorithm_args, self.model)
self.scheduler = StepLR(self.optimizer, cfg.train.step_size,
cfg.train.decay)
self._optimizers.append(
self.optimizer) # -- Add models & optimizers to base for saving
# -- Change settings from parent class
# ++ Parent class automatically initializes 4 metrics: loss/acc for train/test
# ++ E.g switch metric slope
self.set_eval_metric_comparison(True)
# ++ E.g. to add variable name to be saved at checkpoints
self._save_data.append("scheduler")
self._tensorboard_model = False
self.loss_values_train = []
self.loss_values_test = []
##### Make directories and shit for demo########
self.img_dir = os.getcwd() + "/" +image_dir
self.act_dir = os.getcwd() + "/" + activations_dir
self.steer_dir = os.getcwd() + "/" + steer_distr_dir
if not os.path.isdir(self.img_dir):
os.mkdir(self.img_dir)
if not os.path.isdir(self.act_dir):
os.mkdir(self.act_dir)
if not os.path.isdir(self.steer_dir):
os.mkdir(self.steer_dir)
self.nr_img = 0
################################################
super(ConditionalImitationAgent, self).__end_init__()
def __init__(self, args):
super(SARPN, self).__init__()
print("backbone:", args.backbone)
self.feature_extraction = get_models(args)
if args.backbone in ["ResNet18", "ResNet34"]:
adff_num_features = 640
rpd_num_features = 512
block_channel = [64, 64, 128, 256, 512]
top_num_features = block_channel[-1]
if args.backbone in ["ResNet50", "ResNet101", "ResNet152"]:
adff_num_features = 1280
rpd_num_features = 2048
block_channel = [64, 256, 512, 1024, 2048]
top_num_features = block_channel[-1]
if args.backbone in ["DenseNet121"]:
adff_num_features = 640
rpd_num_features = 1024
block_channel = [64, 128, 256, 512, 1024]
top_num_features = block_channel[-1]
if args.backbone in ["DenseNet161"]:
adff_num_features = 1280
rpd_num_features = 2048
block_channel = [96, 192, 384, 1056, 2208]
top_num_features = block_channel[-1]
if args.backbone in ["DenseNet169"]:
adff_num_features = 1280
rpd_num_features = 2048
block_channel = [64, 128, 256, 640, 1664]
top_num_features = block_channel[-1]
if args.backbone in ["DenseNet201"]:
adff_num_features = 1280
rpd_num_features = 2048
block_channel = [64, 128, 256, 896, 1920]
top_num_features = block_channel[-1]
if args.backbone in ["SENet154"]:
adff_num_features = 1280
rpd_num_features = 2048
block_channel = [128, 256, 512, 1024, 2048]
top_num_features = block_channel[-1]
if args.backbone in [
"SE_ResNet50", "SE_ResNet101", "SE_ResNet152",
"SE_ResNext50_32x4d", "SE_ResNext101_32x4d"
]:
adff_num_features = 1280
rpd_num_features = 2048
block_channel = [64, 256, 512, 1024, 2048]
top_num_features = block_channel[-1]
self.residual_pyramid_decoder = modules.RPD(rpd_num_features,
top_num_features)
self.adaptive_dense_feature_fusion = modules.ADFF(
block_channel, adff_num_features, rpd_num_features)
def main():
logging.basicConfig(level=logging.INFO)
parser = argparse.ArgumentParser()
# data
parser.add_argument('--n-classes', type=int, default=10)
parser.add_argument('--feat-std-min', type=float, default=0.1)
# model
parser.add_argument('--model-names', type=str, nargs='+')
parser.add_argument('--n-comp', type=int, default=10)
parser.add_argument('--n-trunc', type=int, default=30)
parser.add_argument('--dp-alpha', type=float, default=1)
parser.add_argument('--pcomp-dirichlet-dist-alpha', type=float, default=1)
parser.add_argument('--pkw-beta-dist-alpha', type=float, default=1)
parser.add_argument('--pkw-beta-dist-beta', type=float, default=1)
parser.add_argument('--pkw-dirichlet-dist-alpha', type=float, default=1)
# trace
parser.add_argument('--samples', type=int, default=500)
parser.add_argument('--njobs', type=int, default=1)
# other
parser.add_argument('--exp-name', type=str, default='A')
args = parser.parse_args()
logger.info("args=%s", args)
save_config(vars(args), args.exp_name)
dataset = data.Mnist(args.n_classes, args.feat_std_min)
for model_name, model in models.get_models(dataset.X_count, dataset.X_bin,
vars(args)).items():
if args.model_names is None or model_name in args.model_names:
exp_name = "{}_{}".format(args.exp_name, model_name)
pred_clusters = get_pred_clusters(model, args.samples, args.njobs)
dataset.evaluate_clusters(pred_clusters, exp_name)
def loadModelAndCheckpoint(self):
self.model=get_models(self.modelname,num_classes=self.numclasses)
assert os.path.isfile(self.checkpoint)
checkpoint = torch.load(self.checkpoint, map_location=torch.device("cpu"))
extras=checkpoint["extras"]
epoch=checkpoint["epoch"]
self.model.load_state_dict(checkpoint['state_dict'])
self.model.eval()
def configurate_extension(self):
from models import get_models
for m in get_models():
self.extension.register(m)
self.extension.setup()
def main():
# Set up Chain Objects
site = chainclient.get(SITE_URL)
metric_hash, device_hash, sensor_hash = get_models(site)
# Pass through websocket events from chainAPI
def get_ws_values_loop():
if True:
stream_url = 'ws://localhost:8000/'
else:
stream_url = site.links['ch:websocketStream'].href
logger.info('Connecting to %s' % stream_url)
ws = create_connection(stream_url)
logger.info('Connected!')
while True:
resource_data = ws.recv()
logger.debug(resource_data)
in_data = HALDoc(json.loads(resource_data))
try:
sensor = sensor_hash[in_data.links['ch:sensor'].href]
except KeyError:
logger.warning('Hash miss: %s' %
in_data.links['ch:sensor'].href)
continue
logger.debug('Received value of %f from sensor %s' %
(in_data.value, sensor))
sensor.value = in_data.value
liblo.send(outgoing_addr, '/device/data', sensor.device.index,
sensor.metric, in_data.value)
t = Thread(target=get_ws_values_loop)
t.daemon = True
t.start()
# OSC Server to pass through Unity player information
# /player/location x y z
# /player/angle yaw pitch roll
# /time seconds
def osc_pass_through_loop():
server = liblo.Server(OSC_UNITY_PORT)
def pass_through(path, args):
logger.info("Received data from Unity: %s : %s" % (path, args))
liblo.send(outgoing_addr, path, *args)
server.add_method(None, None, pass_through)
while True:
server.recv(100)
t2 = Thread(target=osc_pass_through_loop)
t2.daemon = True
t2.start()
# OSC Server to communicated with Music client
try:
server = liblo.Server(OSC_IN_PORT)
except liblo.ServerError, err:
print str(err)
def test_func(test_path, save_path):
model_names = [
'ExtraTreesClassifier',
'AdaBoostClassifier',
'GradientBoostingClassifier',
'RandomForestClassifier',
'BaggingClassifier',
]
encode_type = 'utf-8'
test = pd.read_csv(test_path, keep_default_na=False, encoding=encode_type)
encoders = joblib.load('./dataSet/encoders.pkl')
# 数据预处理
test, _ = ProcessData(test, encoders)
# 扔掉不用的标签字段（例如：值类型为字符串的字段）
drop_cols = args['drop_cols']
x_columns = [x for x in test.columns if x not in drop_cols]
X = test[x_columns]
# 提取一些信息用于后处理
data = pd.DataFrame(test[['srcAddress', 'destAddress', 'eventId']])
# scale
X = pd.DataFrame(encoders[len(encoders) - 1].transform(X),
columns=x_columns)
# 在多个模型上分别预测
models = get_models(model_names, True)
for model_name in model_names:
y_pred = models[model_name].predict(X)
data[model_name] = y_pred
# 统计投票结果
y_preds = np.sum(data[models].to_numpy(), axis=1)
# y_preds = (y_preds > (len(models)//2)) + 0
data['y_preds'] = y_preds
# 后处理；认为同一组['srcAddress','destAddress']的标签结果应该相同，故而通过均值聚合预测结果
data['index'] = list(range(len(data)))
data = data.merge(data[['srcAddress', 'destAddress',
'y_preds']].groupby(['srcAddress', 'destAddress'],
as_index=False).mean(),
on=['srcAddress', 'destAddress'],
suffixes=['', '_mean_by_addr']).sort_values('index')
# 加合同组均值预测结果与单样本预测结果
result = data[['y_preds' + 'y_preds_mean_by_addr']]
y_preds = np.sum(result.to_numpy(), axis=1)
y_preds = (y_preds > (len(models))) + 0
data['label'] = y_preds
submission = data[['eventId', 'label']]
submission.to_csv(save_path + 'S3_finalB.csv',
index=False,
encoding='utf-8')
def paramsFlopsCounter(models,num_classes=10,input_shape=(3,32,32)):
logger=get_logger("./")
for modelname in models:
model=get_models(modelname,num_classes=10)
model = model.eval()
pa1=getParams(model)
fl1=getFlops(model,input_shape)
fl2,pa2=get_model_complexity_info(model,input_shape,True)
#logger.info("{} v1: {}--{} ".format(model,pa1,fl1))
logger.info("{} v1: {}--{} v2: {}--{}".format(modelname,pa1,fl1,pa2,fl2))
def load_model(name, input_node):
""" Creates and returns an instance of the model given its class name.
The created model has a single placeholder node for feeding images.
"""
# Find the model class from its name
all_models = models.get_models()
net_class = [model for model in all_models if model.__name__ == name][0]
# Construct and return the model
return net_class({'data': input_node})
def evaluate(self, data_loader):
with tf.device('/cpu:0'):
input_images, input_labels, input_widths = data_loader.read_with_bucket_queue(
batch_size=cfg.TEST.BATCH_SIZE,
num_threads=cfg.TEST.THREADS,
num_epochs=1,
shuffle=False)
with tf.device('/gpu:0'):
logits = get_models(cfg.MODEL.BACKBONE)(cfg.MODEL.NUM_CLASSES).build(input_images, False)
seqlen = tf.cast(tf.floor_div(input_widths, 2), tf.int32, name='sequence_length')
softmax = tf.nn.softmax(logits, dim=-1, name='softmax')
decoded, log_prob = tf.nn.ctc_greedy_decoder(softmax, seqlen)
distance = tf.reduce_mean(tf.edit_distance(tf.cast(decoded[0], tf.int32), input_labels))
saver = tf.train.Saver(tf.global_variables())
gpu_options = tf.GPUOptions(allow_growth=True)
with tf.Session(config=tf.ConfigProto(gpu_options=gpu_options, allow_soft_placement=True)) as sess:
saver.restore(sess, tf.train.latest_checkpoint(self.output_dir))
sess.run(tf.local_variables_initializer())
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess=sess, coord=coord)
try:
cnt = 0
dm = 1e-5
while not coord.should_stop():
dt = sess.run([distance, ])[0]
cnt += 1
dm = (dm + dt) / cnt
if cfg.TEST.VIS:
dd, il, ii = sess.run([decoded, input_labels, input_images])
gts = self.decoder.sparse_to_strlist(il.indices, il.values, cfg.TEST.BATCH_SIZE)
pts = self.decoder.sparse_to_strlist(dd[0].indices, dd[0].values, cfg.TEST.BATCH_SIZE)
tb = PrettyTable()
tb.field_names = ['Index', 'GroundTruth', 'Predict', '{:.3f}/{:.3f}'.format(dt, dm)]
for i in range(len(gts)):
tb.add_row([i, gts[i], pts[i], ''])
print(tb)
else:
print('EditDistance: {:.3f}/{:.3f}'.format(dt, dm))
except tf.errors.OutOfRangeError:
print('Epochs Complete!')
finally:
coord.request_stop()
coord.join(threads)
def select_model(models_probe, X_train, y_train, cv=10):
"""
Compare performance of different models in a dataset
Print metrics for each model, and a boxplot with each model performance
Parameteres:
models_probe: List of models to test
X_train: Training features
y_train: Training labels
cv: K value for Cross Validation
Return:
None
"""
# Get the models implementations
models = get_models(models_probe)
# Variables to store results
results = []
names = []
# Every algorithm is tested and results are
# collected and printed
for name, model in models:
cv_results = cross_validation(model,
X_train,
y_train,
folds=cv,
message=False)
results.append(cv_results)
names.append(name)
msg = "Model Selection - %s: %f (%f)" % (name, cv_results.mean(),
cv_results.std())
print(msg)
# boxplot algorithm comparison
fig = plt.figure()
fig.suptitle('Algorithm Comparison')
ax = fig.add_subplot(111)
plt.boxplot(results)
ax.set_xticklabels(names)
plt.show()
# Check the best performing model
results_np = np.array(results)
results_mean = np.mean(results_np, axis=1)
best_model = names[np.argmax(results_mean)]
return best_model
def main():
args = init_args()
if args.list_models:
print('\n'.join(models.get_model_names()))
exit()
m = args.model.split(',')
dict_m = models.get_models(m)
x, y = preprocess.load_data(args.train_datesets)
for model_name in dict_m:
model = dict_m[model_name]
print('Training model %s' % model_name)
model.fit(x, y)
models.save_model(model, model_name, args.model_dir)
print('Train finished, save to %s' % args.model_dir)
def run(params, log_dir=None):
if log_dir is None:
log_dir = params["main_params"]["log_dir"]
else:
params["main_params"]["log_dir"] = log_dir
os.makedirs(log_dir, exist_ok=True)
save_yaml(params, out_file=os.path.join(log_dir, "init_params.yml"))
initialize(params)
print("loading data")
datasets = get_loaders(params)
models = get_models(params, datasets)
runners = get_runners(params, datasets, models)
for runner in runners:
runner()
def freeze_net(ckpt_path, save_path):
from tensorflow.python.tools import freeze_graph
logger.info('Freeze Model Will Saved at ', save_path)
with tf.Graph().as_default():
x1 = tf.placeholder(tf.float32,
shape=[None, 32, None, 3],
name='input_images')
x2 = tf.placeholder(tf.int32, shape=[None], name='input_widths')
logits = get_models(cfg.MODEL.BACKBONE)(cfg.MODEL.NUM_CLASSES).build(
x1, False)
seqlen = tf.cast(tf.floor_div(x2, 2), tf.int32, name='sequence_length')
softmax = tf.nn.softmax(logits, dim=-1, name='softmax')
decoded, log_prob = tf.nn.ctc_greedy_decoder(softmax, seqlen)
prob_value = -1.0 * tf.reshape(log_prob, shape=[-1]) / tf.cast(
seqlen, tf.float32)
tf.identity(tf.cast(decoded[0].indices, dtype=tf.int32),
name='indices')
tf.identity(tf.cast(decoded[0].values, dtype=tf.int32), name='values')
tf.identity(tf.shape(softmax)[0], name='length')
tf.identity(prob_value, name='prob')
saver = tf.train.Saver()
with tf.Session(graph=tf.Graph()) as sess:
saver.restore(sess, ckpt_path)
fdir, name = os.path.split(save_path)
tf.train.write_graph(sess.graph_def, fdir, name, as_text=True)
freeze_graph.freeze_graph(
input_graph=save_path,
input_saver='',
input_binary=False,
input_checkpoint=ckpt_path,
output_node_names='indices,values,prob,length',
restore_op_name='',
filename_tensor_name='',
output_graph=save_path,
clear_devices=True,
initializer_nodes='',
)
logger.info('Freeze Model done.')
def main(args):
# 0.762
# args.model='unet'
# model1=get_models(args)
# model1.load_state_dict(torch.load(r'E:\segmentation\Image_Segmentation\logs\cvc_logs\unet_ep1600\cvc\20200312-143050\model_best.pth.tar',map_location='cpu')['state_dict'])
# # 0.766/0.773
# args.model='unet++'
# model2=get_models(args)
# model2.load_state_dict(torch.load(r'E:\segmentation\Image_Segmentation\logs\cvc_logs\unet++_nodeep_ep800\cvc\no_deep\model_best.pth.tar',map_location='cpu')['state_dict'])
#
# # mutilres 0.695
# args.model='multires_unet'
# model3=get_models(args)
# model3.load_state_dict(torch.load(r'E:\segmentation\Image_Segmentation\logs\cvc_logs\multires_unet_800\cvc\20200310-172036\checkpoint.pth.tar',map_location='cpu')['state_dict'])
#
#
# attention_unet 0.778
args.model = 'attention_unet_v1'
model4 = get_models(args)
model4.load_state_dict(
torch.load(
r'E:\segmentation\Image_Segmentation\logs\cvc_logs\attention_unet_v1_ep1600\cvc\20200312-143413\model_best.pth.tar',
map_location='cpu')['state_dict'])
genotype = eval('genotypes.%s' % 'layer7_double_deep')
#BuildNasUnetPrune
model5 = BuildNasUnetPrune(
genotype=genotype,
input_c=3,
c=16,
num_classes=1,
meta_node_num=4,
layers=9,
dp=0,
use_sharing=True,
double_down_channel=True,
aux=True,
)
model5.load_state_dict(
torch.load(
r'E:\segmentation\Image_Segmentation\nas_search_unet\logs\cvc\layer7_double_deep_ep1600_20200320-200539\model_best.pth.tar',
map_location='cpu')['state_dict'])
models_list = [model4, model5]
inference_isic(models_list, args.image, args.mask)
def train(self, models_kwargs: Dict) -> List[List[ForecastModel]]:
"""
:param models_kwargs: A dictionary of models' init arguments
:return: List of models(a list of models) for each cluster.
"""
models = []
for cid in range(len(self._cluster_data)):
cluster_models = get_models(models_kwargs)
train_seqs = self._cluster_seqs(cid,
test_mode=False,
with_label=True)
for model_name, model in cluster_models.items():
# Fit the model
model.fit(train_seqs)
self.eval(cid, model)
models.append(cluster_models)
return models
def __init__(self, cfg):
super(DemoAgentWiki, self).__init__(cfg)
use_cuda = self._use_cuda # ++ Parent class already saves some configuration variables
# ++ All parent variables should start with _.
# -- Get necessary variables from cfg
self.train_cfg = cfg.train
# -- Initialize model
model_class = get_models(cfg.model)
self.model = model_class[0](cfg.model,
torch.zeros(3, 224, 224), torch.zeros(10))
# ++ All models receive as parameters (configuration namespace, input data size,
# ++ output data size)
self._models.append(
self.model
) # -- Add models & optimizers to base for saving
# ++ After adding model you can set the agent to cuda mode
# ++ Parent class already makes some adjustments. E.g. turns model to cuda mode
if use_cuda:
self.cuda()
# -- Initialize optimizers
self.optimizer = self.get_optim(cfg.train.algorithm,
cfg.train.algorithm_args, self.model)
self._optimizers.append(
self.optimizer) # -- Add models & optimizers to base for saving
# -- Initialize criterion
self.criterion = getattr(torch.nn, cfg.train.criterion)()
# -- Change settings from parent class
# ++ Parent class automatically initializes 4 metrics: loss/acc for train/test
# ++ E.g switch metric slope
self.set_eval_metric_comparison(True)
# ++ E.g. to add variable name to be saved at checkpoints
self._save_data.append("train_cfg")
super(DemoAgentWiki, self).__end_init__()
def tflite(ckpt_path, save_path):
logger.info('tflite Model Will Saved at ', save_path)
with tf.Graph().as_default():
x1 = tf.placeholder(tf.uint8,
shape=[None, 32, 1024, 3],
name='input_images')
logits = get_models(cfg.MODEL.BACKBONE)(cfg.MODEL.NUM_CLASSES).build(
x1, False)
y1 = tf.nn.softmax(logits, dim=-1, name='softmax')
saver = tf.train.Saver()
with tf.Session(graph=tf.Graph()) as sess:
saver.restore(sess, ckpt_path)
converter = tf.contrib.lite.TFLiteConverter.from_session(
sess, [x1], [y1])
tflite_model = converter.convert()
with open(save_path, "wb") as wt:
wt.write(tflite_model)
logger.info('Convert to TFLite Model done.')
def main():
logging.basicConfig(
level=logging.INFO,
format="%(asctime)s : " +
"%(module)s (%(lineno)s) - %(levelname)s - %(message)s")
conf = ConfigParser(os.environ)
conf.read(sys.argv[1])
logging.warning('loading datasets...')
datasets = get_data(conf)
logging.warning('loaded these: {0}'.format(datasets.keys()))
logging.warning('loading models...')
models = get_models(conf)
logging.warning('evaluating...')
for data_type, data in datasets.iteritems():
logging.warning('data: {0}'.format(data_type))
r = Regression(conf)
r.featurize_data(data, models)
r.evaluate()
def _load_weights(self, weights=None):
with self.graph.as_default():
with tf.device(self.device):
input_images = tf.placeholder(tf.uint8, shape=[None, 32, None, 3], name='input_images')
input_widths = tf.placeholder(tf.uint8, shape=[None], name='input_widths')
with tf.device('/gpu:0'):
logits = get_models(cfg.MODEL.BACKBONE)(cfg.MODEL.NUM_CLASSES).build(input_images, False)
seqlen = tf.cast(tf.floor_div(input_widths, 2), tf.int32, name='sequence_length')
softmax = tf.nn.softmax(logits, dim=-1, name='softmax')
decoded, log_prob = tf.nn.ctc_greedy_decoder(softmax, seqlen)
prob = -tf.divide(tf.cast(log_prob, tf.int32), seqlen[0])
saver = tf.train.Saver(tf.global_variables())
if weights is None:
saver.restore(self.sess, tf.train.latest_checkpoint(self.output_dir))
else:
saver.restore(self.sess, weights)
return {'input_images': input_images, 'input_widths': input_widths, 'decoded': decoded, 'prob': prob}
def load_model(name):
'''Creates and returns an instance of the model given its class name.
The created model has a single placeholder node for feeding images.
'''
# Find the model class from its name
all_models = models.get_models()
lut = {model.__name__: model for model in all_models}
if name not in lut:
print('Invalid model index. Options are:')
# Display a list of valid model names
for model in all_models:
print('\t* {}'.format(model.__name__))
return None
NetClass = lut[name]
# Create a placeholder for the input image
spec = models.get_data_spec(model_class=NetClass)
data_node = tf.placeholder(tf.float32,
shape=(None, spec.crop_size, spec.crop_size, spec.channels))
# Construct and return the model
return NetClass({'data': data_node})
def load_model(name):
'''Creates and returns an instance of the model given its class name.
The created model has a single placeholder node for feeding images.
'''
# Find the model class from its name
all_models = models.get_models()
lut = {model.__name__: model for model in all_models}
if name not in lut:
print('Invalid model index. Options are:')
# Display a list of valid model names
for model in all_models:
print('\t* {}'.format(model.__name__))
return None
NetClass = lut[name]
# Create a placeholder for the input image
spec = models.get_data_spec(model_class=NetClass)
data_node = tf.placeholder(tf.float32,
shape=(None, spec.crop_size, spec.crop_size, spec.channels))
# Construct and return the model
return NetClass({'data': data_node})
def main():
logging.basicConfig(level=logging.DEBUG)
parser = argparse.ArgumentParser()
# data
parser.add_argument('--max-df', type=float, default=0.3)
parser.add_argument('--min-df', type=int, default=20)
parser.add_argument('--doc-len-min', type=int, default=20)
parser.add_argument('--doc-len-max', type=int, default=200)
parser.add_argument('--classes', type=int, nargs='+', default=[0, 1, 2, 3])
# model
parser.add_argument('--model-names', type=str, nargs='+')
parser.add_argument('--n-comp', type=int, default=10)
parser.add_argument('--n-trunc', type=int, default=30)
parser.add_argument('--dp-alpha', type=float, default=1)
parser.add_argument('--pcomp-dirichlet-dist-alpha', type=float, default=1)
parser.add_argument('--pkw-beta-dist-alpha', type=float, default=1)
parser.add_argument('--pkw-beta-dist-beta', type=float, default=1)
parser.add_argument('--pkw-dirichlet-dist-alpha', type=float, default=1)
# trace
parser.add_argument('--samples', type=int, default=500)
parser.add_argument('--njobs', type=int, default=1)
# other
parser.add_argument('--exp-name', type=str, default='A')
args = parser.parse_args()
logger.info("args=%s", args)
save_config(vars(args), args.exp_name)
dataset = data.TwentyNewsGroups(args.max_df, args.min_df, args.doc_len_min,
args.doc_len_max, args.classes)
for model_name, model in models.get_models(dataset.X_count, dataset.X_bin,
vars(args)).items():
if args.model_names is None or model_name in args.model_names:
exp_name = "{}_{}".format(args.exp_name, model_name)
pred_clusters = get_pred_clusters(model, args.samples, args.njobs)
title = r"Clusters composition, $\alpha={}$".format(args.dp_alpha)
dataset.evaluate_clusters(pred_clusters, exp_name, title)
def main(args):
model1 = get_models(args)
ckpt1 = torch.load(args.model_weight1, map_location='cpu')
model1.load_state_dict(ckpt1['state_dict'])
# inference_isic(model,args.image,args.mask)
ckpt2 = torch.load(args.model_weight2, map_location='cpu')
genotype = eval('genotypes.%s' %
'stage1_layer9_110epoch_double_deep_final')
#BuildNasUnetPrune
model2 = BuildNasUnetPrune(
genotype=genotype,
input_c=3,
c=16,
num_classes=1,
meta_node_num=4,
layers=9,
dp=0,
use_sharing=True,
double_down_channel=True,
aux=True,
)
model2.load_state_dict(ckpt2['state_dict'])
inference_isic(model1, model2, args.image, args.mask)
def main(args):
############ init config ################
model_name = args.model
assert model_name in models_dict.keys(),"The Usage model is not exist !"
print('Usage model :{}'.format(model_name))
#################### init logger ###################################
log_dir = './logs/'+ args.model+'_'+args.note + '/{}'.format(time.strftime('%Y%m%d-%H%M%S'))
logger = get_logger(log_dir)
print('RUNDIR: {}'.format(log_dir))
logger.info('{}-Train'.format(args.model))
# setting
setting={k: v for k, v in args._get_kwargs()}
logger.info(setting)
args.save_path = log_dir
args.save_tbx_log = args.save_path + '/tbx_log'
writer = SummaryWriter(args.save_tbx_log)
##################### init device #################################
if args.manualSeed is None:
args.manualSeed = random.randint(1, 10000)
np.random.seed(args.manualSeed)
torch.manual_seed(args.manualSeed)
args.use_cuda= args.gpus>0 and torch.cuda.is_available()
args.device = torch.device('cuda' if args.use_cuda else 'cpu')
if args.use_cuda:
torch.cuda.manual_seed(args.manualSeed)
cudnn.benchmark = True
####################### init dataset ###########################################
train_loader=get_dataloder(args,split_flag="train")
val_loader=get_dataloder(args,split_flag="valid")
######################## init model ############################################
# model
logger.info("Model Dict has keys: \n {}".format(models_dict.keys()))
model=get_models(args)
if torch.cuda.device_count() > 1 and args.use_cuda:
logger.info('use: %d gpus', torch.cuda.device_count())
model = nn.DataParallel(model)
logger.info('param size = %fMB', calc_parameters_count(model))
# init loss
if args.loss=='bce':
criterion=nn.BCELoss()
elif args.loss=='bcelog':
criterion=nn.BCEWithLogitsLoss()
elif args.loss=="dice":
criterion=DiceLoss()
elif args.loss=="softdice":
criterion=SoftDiceLoss()
elif args.loss=='bcedice':
criterion=BCEDiceLoss()
else:
criterion=nn.CrossEntropyLoss()
if args.use_cuda:
logger.info("load model and criterion to gpu !")
model=model.to(args.device)
criterion=criterion.to(args.device)
# init optimizer
if args.model_optimizer=="sgd":
#torch.optim.SGD(parametetrs,lr=args.lr,weight_decay=args.weight_decay,momentum=args.momentum)
optimizer=torch.optim.SGD(model.parameters(),lr=args.lr,weight_decay=args.weight_decay,momentum=args.momentum)
else:
optimizer=torch.optim.Adam(model.parameters(),args.lr,[args.beta1, args.beta2],
weight_decay=args.weight_decay)
# init schedulers Steplr
scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer,args.epoch)
# scheduler=torch.optim.lr_scheduler.StepLR(optimizer=optimizer,step_size=30,gamma=0.1,last_epoch=-1)
############################### check resume #########################
start_epoch=0
if args.resume is not None:
if os.path.isfile(args.resume):
logger.info("Loading model and optimizer from checkpoint '{}'".format(args.resume))
checkpoint = torch.load(args.resume, map_location=args.device)
start_epoch = checkpoint['epoch']
optimizer.load_state_dict(checkpoint['optimizer'])
model.load_state_dict(checkpoint['state_dict'])
scheduler.load_state_dict(checkpoint['scheduler'])
else:
raise FileNotFoundError("No checkpoint found at '{}'".format(args.resume))
#################################### train and val ########################
max_value=0
for epoch in range(start_epoch,args.epoch):
# lr=adjust_learning_rate(args,optimizer,epoch)
scheduler.step()
logger.info('Epoch: %d lr %e', epoch, scheduler.get_lr()[0])
# train
mr, ms, mp, mf, mjc, md, macc, mean_loss=train(args, model, criterion, train_loader,
optimizer, epoch, logger)
# write
writer.add_scalar('Train/Loss', mean_loss, epoch)
writer.add_scalar('Train/mAcc', macc, epoch)
writer.add_scalar('Train/Recall', mr, epoch)
writer.add_scalar('Train/Specifi', ms, epoch)
writer.add_scalar('Train/Precision', mp, epoch)
writer.add_scalar('Train/F1', mf, epoch)
writer.add_scalar('Train/Jc', mjc, epoch)
writer.add_scalar('Train/Dice', md, epoch)
# val
vmr, vms, vmp, vmf, vmjc, vmd, vmacc, vmean_loss=val(args, model, criterion, val_loader, epoch, logger)
writer.add_scalar('Val/Loss', vmean_loss, epoch)
writer.add_scalar('Val/mAcc', vmacc, epoch)
writer.add_scalar('Val/Recall', vmr, epoch)
writer.add_scalar('Val/Specifi', vms, epoch)
writer.add_scalar('Val/Precision', vmp, epoch)
writer.add_scalar('Val/F1', vmf, epoch)
writer.add_scalar('Val/Jc', vmjc, epoch)
writer.add_scalar('Val/Dice', vmd, epoch)
is_best=True if (vmjc>=max_value) else False
max_value=max(max_value,vmjc)
state={
'epoch': epoch,
'optimizer': optimizer.state_dict(),
'state_dict': model.state_dict(),
'scheduler': model.state_dict(),
}
logger.info("epoch:{} best:{} max_value:{}".format(epoch,is_best,max_value))
if not is_best:
torch.save(state,os.path.join(args.save_path,"checkpoint.pth.tar"))
else:
torch.save(state,os.path.join(args.save_path,"checkpoint.pth.tar"))
torch.save(state,os.path.join(args.save_path,"model_best.pth.tar"))
writer.close()
def index():
page_desc = "Home"
headers = map(Item, ["Model name"])
data = map(lambda x: Item(x, url="/list/" + x, label=x),
models.get_models())
return render_template('index.html', **locals())
optlist, args = getopt.getopt(sys.argv[1:], 'vh', ['verbose', 'help'])
for (k, v) in optlist:
if k in ('-v', '--verbose'):
verbose = True
elif k in ('-h', '--help'):
usage()
sys.exit(0)
# create an instance of the Data class
dtrain = Data('datasets/train-exploration.csv')
dvalid = Data('datasets/validation-exploration.csv')
# create placeholders for x and y
x = tf.placeholder(tf.float32, shape=[None, 28, 28, 1])
y = tf.placeholder(tf.float32, shape=[None, 10])
# build the neural network
models = get_models(x, y)
with tf.Session() as sess:
# get a validation batch
vbatch = (dvalid.features, dvalid.labels) # get_batch(4207)
# iterate over all models
for m in models:
run_ensemble(sess, m, x, y, vbatch, dtrain)
def start():
system('clear')
print ''' ... MMM
. .,MMMMMMMMMD. . MMM.
.DMMMMMMMMMMMMMM . MMM
.MMMMMMM..MMMM? MMMM . MMM+. . .
.IMMMO?: ?$ ,. MMM . NMMM .8MM.
. ...MMM. MMM .. MMM. MMM
.OMM~ MMM8 OMM~. MMM
.7. .MMN .MMMM. MMM..... MMM .
.ZMM+ MMM7MMMM: ~MM~ MMMM . MMM .
OMMZ MMMMMMM MMM. MMN. . .IMMMMMMMMMMMMM
.?MMN .MMMMZ MMM ...DMM. +MMMMMMMMMMMMMMMM
.MMM MMM .. MMM MM8 OMMZ, MMM. .
..MMM. MMM MMM .IMM MMM.
..MMMN. 8MM .7MM, MMM. .DMM
MMM. .OMM MMMMMM . MM.
.ZMM 7MM.. ...
.. . ~ . . . .
..
.
... . ,,.. .MMM
ZMMMMMMMMMMMMMM, .. .. . .NMM7 .$MM . .
+MMMMMMMMM?. .. ... ?NZ DMMMMMMM MM$.. . ZD~
MMMMM? .. MM=. .MM . MMMMM8.. NMMMM:.+MM~ ... ... ...MMM. . . . DMMMM? MMM ..MMMMM,
MMM MM$ MMMMM ,MMMM .MM8. . MMMMMMM+ MMM . 7MMM. MMMMMMMMM. MMM MMMMZ
MMM MMM MMMM.. MM.. MMM. . MMMM..NMM. MMM $MMMMM= MMM.....$MM DMM= OMMMM
MMM MMM MMM$ .MMM MMM... . . MMMIOMMMMM, .MMM+$MMMMMMM :MMM.MMM.
.MMM IMM~MMZ . .$O: MMM .NMM ... . . $MMMMMM7. MMMMMMMMM~ . MMMMMM
MMM MMMMM . . MMMMMMMMMM MMM .MMMMMMI DMM..... . . . MMMMM7.
MMM. MMMM= . MMMM. MMMMM. .MMM . MD MMMMMMMMN . MMM. . . MM ..OMMMM.
MMM. .MMMM. . MMM= ...=MMMM...$MMM....$MMMMM .MMM. ,MMMMM. .,MMMM?7MMMM+ . MMMM
NMM. $MMM. ..MMM. ..MMMMMMI...OMMMMMMMMMM. NMM MMMMM .. MMMMMMMM MMMM
?MM MMM MMMMMMMMMM .:NMMI ~MM MMM . .MM=.
... .. DMMMMM: . MM+ ... ..
... . Z
'''
print 'Welcome to the iPhone 6+ CLI tracking tool v1.1.2, from Risto Keravuori (www.risto.io).\n' \
'Please enjoy, and good luck in your search!'
mode = get_mode()
if mode == 4:
with open(join(FILE_PATH, 'previous.pickle'), 'rb') as f:
data = load(f)
zip_code = data['zip_code']
target_stores = data['target_stores']
alert_models = data['alert_models']
beep_models = data['beep_models']
print '\nRestarting monitoring with the following settings:'
print '-- zip code: %s' % zip_code
print '-- target_stores: ',
for store in target_stores[:-1]:
print '%s, ' % store,
print target_stores[-1]
print '-- alert models: ',
for model in alert_models[:-1]:
print '%s, ' % nice_model_name(model),
print nice_model_name(alert_models[-1]) if alert_models else '--'
print '-- beep models: ',
for model in beep_models[:-1]:
print '%s, ' % nice_model_name(model),
print nice_model_name(beep_models[-1]) if beep_models else '--'
else:
zip_code = get_zip_code()
target_stores = get_target_stores(zip_code)
alert_models, beep_models = get_models(mode)
with open(join(FILE_PATH, 'previous.pickle'), 'wb') as f:
dump(
{
'zip_code': zip_code,
'target_stores': target_stores,
'alert_models': alert_models,
'beep_models': beep_models
}, f)
start_monitoring(zip_code, target_stores, alert_models, beep_models)
def main(args):
#################### init logger ###################################
args.model='unet'
model_weight_path='../logs/isic2018/unet_ep300/20200402-135108/model_best.pth.tar'
model=get_models(args)
model.load_state_dict(torch.load(model_weight_path, map_location='cpu')['state_dict'])
log_dir = './models/' + args.model+'_prune_'+args.note
logger = get_logger(log_dir)
print('RUNDIR: {}'.format(log_dir))
logger.info('{}-L1Prune'.format(args.model))
# setting
args.save_path = log_dir
args.save_tbx_log = args.save_path + '/tbx_log'
writer = SummaryWriter(args.save_tbx_log)
if args.manualSeed is None:
args.manualSeed = random.randint(1, 10000)
np.random.seed(args.manualSeed)
torch.manual_seed(args.manualSeed)
args.use_cuda = args.gpus > 0 and torch.cuda.is_available()
args.device = torch.device('cuda' if args.use_cuda else 'cpu')
if args.use_cuda:
torch.cuda.manual_seed(args.manualSeed)
cudnn.benchmark = True
setting = {k: v for k, v in args._get_kwargs()}
logger.info(setting)
train_loader=get_dataloder(args,split_flag="train")
val_loader=get_dataloder(args,split_flag="valid")
# init loss
if args.loss == 'bce':
criterion = nn.BCELoss()
elif args.loss == 'bcelog':
criterion = nn.BCEWithLogitsLoss()
elif args.loss == "dice":
criterion = DiceLoss()
elif args.loss == "softdice":
criterion = SoftDiceLoss()
elif args.loss == 'bcedice':
criterion = BCEDiceLoss()
else:
criterion = nn.CrossEntropyLoss()
if args.use_cuda:
logger.info("load model and criterion to gpu !")
model = model.to(args.device)
criterion = criterion.to(args.device)
logger.info("Original trained model performance test: ")
infer(args, model, criterion, val_loader,logger)
# Pruning
# Pruning Configuration, in paper 'PRUNING FILTERS FOR EFFICIENT CONVNETS',
configure_list = [{
'sparsity': 0.5,
'op_types': ['Conv2d'],
'op_names': ['Conv1.conv.0','Conv1.conv.3','Conv2.conv.0','Conv2.conv.3','Conv3.conv.0','Conv3.conv.3',
'Conv4.conv.0','Conv4.conv.3','Conv5.conv.0','Conv5.conv.3',
'Up5.up.1','Up_conv5.conv.0','Up_conv5.conv.3',
'Up4.up.1','Up_conv4.conv.0','Up_conv4.conv.3',
'Up3.up.1','Up_conv3.conv.0','Up_conv3.conv.3',
'Up2.up.1','Up_conv2.conv.0','Up_conv2.conv.3',
]}
]
# Prune model and test accuracy without fine tuning.
logger.info('=' * 10 + 'Test on the pruned model before fine tune' + '=' * 10)
pruner = L1FilterPruner(model, configure_list)
# change the forward func (mul pruning mask )
model = pruner.compress()
# test performance without finetuning
logger.info("Pruning trained model performance test: ")
infer(args, model, criterion, val_loader,logger)
# Fine tune the pruned model for 40 epochs and test accuracy
logger.info('=' * 10 + 'Fine tuning' + '=' * 10)
#torch.optim.SGD(parametetrs,lr=args.lr,weight_decay=args.weight_decay,momentum=args.momentum)
optimizer=torch.optim.SGD(model.parameters(),lr=args.lr,weight_decay=args.weight_decay,momentum=args.momentum)
# init schedulers Steplr
scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer,args.epoch)
max_value = 0
for epoch in range(0, args.epoch):
# lr=adjust_learning_rate(args,optimizer,epoch)
scheduler.step()
logger.info('Epoch: %d lr %e', epoch, scheduler.get_lr()[0])
# update mask
pruner.update_epoch(epoch)
# train
train(args, model, criterion, train_loader,optimizer, epoch, logger)
# val
vmr, vms, vmp, vmf, vmjc, vmd, vmacc,vloss = infer(args, model, criterion, val_loader,logger)
writer.add_scalar('Val/Loss', vloss, epoch)
writer.add_scalar('Val/mAcc', vmacc, epoch)
writer.add_scalar('Val/Recall', vmr, epoch)
writer.add_scalar('Val/Specifi', vms, epoch)
writer.add_scalar('Val/Precision', vmp, epoch)
writer.add_scalar('Val/F1', vmf, epoch)
writer.add_scalar('Val/Jc', vmjc, epoch)
writer.add_scalar('Val/Dice', vmd, epoch)
is_best = True if (vmjc >= max_value) else False
max_value = max(max_value, vmjc)
if is_best:
pruner.export_model(model_path=os.path.join(args.save_path,"best_prune_unet.pth"), mask_path=os.path.join(args.save_path,'mask_prune_indexs.pth'))
state = {
'epoch': epoch,
'optimizer': optimizer.state_dict(),
'state_dict': model.state_dict(),
'scheduler': model.state_dict(),
}
logger.info("epoch:{} best:{} max_value:{}".format(epoch, is_best, max_value))
torch.save(state, os.path.join(args.save_path, "checkpoint.pth.tar"))
writer.close()
# test the best_prune_unet.pth
args.model='unet'
model_weight_path=os.path.join(args.save_path,"best_prune_unet.pth")
model=get_models(args)
model.load_state_dict(torch.load(model_weight_path, map_location='cpu'))
model = model.to(args.device)
logger.info("Final saved pruned model performance test: ")
infer(args, model, criterion, val_loader,logger)
from django.contrib.gis import admin
from models import *
from django.db import models
for model in models.get_models():
try:
admin.site.register(model, admin.OSMGeoAdmin)
except Exception:
continue
def start():
system('clear')
print ''' ... MMM
. .,MMMMMMMMMD. . MMM.
.DMMMMMMMMMMMMMM . MMM
.MMMMMMM..MMMM? MMMM . MMM+. . .
.IMMMO?: ?$ ,. MMM . NMMM .8MM.
. ...MMM. MMM .. MMM. MMM
.OMM~ MMM8 OMM~. MMM
.7. .MMN .MMMM. MMM..... MMM .
.ZMM+ MMM7MMMM: ~MM~ MMMM . MMM .
OMMZ MMMMMMM MMM. MMN. . .IMMMMMMMMMMMMM
.?MMN .MMMMZ MMM ...DMM. +MMMMMMMMMMMMMMMM
.MMM MMM .. MMM MM8 OMMZ, MMM. .
..MMM. MMM MMM .IMM MMM.
..MMMN. 8MM .7MM, MMM. .DMM
MMM. .OMM MMMMMM . MM.
.ZMM 7MM.. ...
.. . ~ . . . .
..
.
... . ,,.. .MMM
ZMMMMMMMMMMMMMM, .. .. . .NMM7 .$MM . .
+MMMMMMMMM?. .. ... ?NZ DMMMMMMM MM$.. . ZD~
MMMMM? .. MM=. .MM . MMMMM8.. NMMMM:.+MM~ ... ... ...MMM. . . . DMMMM? MMM ..MMMMM,
MMM MM$ MMMMM ,MMMM .MM8. . MMMMMMM+ MMM . 7MMM. MMMMMMMMM. MMM MMMMZ
MMM MMM MMMM.. MM.. MMM. . MMMM..NMM. MMM $MMMMM= MMM.....$MM DMM= OMMMM
MMM MMM MMM$ .MMM MMM... . . MMMIOMMMMM, .MMM+$MMMMMMM :MMM.MMM.
.MMM IMM~MMZ . .$O: MMM .NMM ... . . $MMMMMM7. MMMMMMMMM~ . MMMMMM
MMM MMMMM . . MMMMMMMMMM MMM .MMMMMMI DMM..... . . . MMMMM7.
MMM. MMMM= . MMMM. MMMMM. .MMM . MD MMMMMMMMN . MMM. . . MM ..OMMMM.
MMM. .MMMM. . MMM= ...=MMMM...$MMM....$MMMMM .MMM. ,MMMMM. .,MMMM?7MMMM+ . MMMM
NMM. $MMM. ..MMM. ..MMMMMMI...OMMMMMMMMMM. NMM MMMMM .. MMMMMMMM MMMM
?MM MMM MMMMMMMMMM .:NMMI ~MM MMM . .MM=.
... .. DMMMMM: . MM+ ... ..
... . Z
'''
print 'Welcome to the iPhone 6+ CLI tracking tool v1.1.2, from Risto Keravuori (www.risto.io).\n' \
'Please enjoy, and good luck in your search!'
mode = get_mode()
if mode == 4:
with open(join(FILE_PATH, 'previous.pickle'), 'rb') as f:
data = load(f)
zip_code = data['zip_code']
target_stores = data['target_stores']
alert_models = data['alert_models']
beep_models = data['beep_models']
print '\nRestarting monitoring with the following settings:'
print '-- zip code: %s' % zip_code
print '-- target_stores: ',
for store in target_stores[:-1]:
print '%s, ' % store,
print target_stores[-1]
print '-- alert models: ',
for model in alert_models[:-1]:
print '%s, ' % nice_model_name(model),
print nice_model_name(alert_models[-1]) if alert_models else '--'
print '-- beep models: ',
for model in beep_models[:-1]:
print '%s, ' % nice_model_name(model),
print nice_model_name(beep_models[-1]) if beep_models else '--'
else:
zip_code = get_zip_code()
target_stores = get_target_stores(zip_code)
alert_models, beep_models = get_models(mode)
with open(join(FILE_PATH, 'previous.pickle'), 'wb') as f:
dump({'zip_code': zip_code,
'target_stores': target_stores,
'alert_models': alert_models,
'beep_models': beep_models}, f)
start_monitoring(zip_code, target_stores, alert_models, beep_models)
from django.contrib import admin
from models import *
from django.db import models
for model in models.get_models():
try:
admin.site.register(model)
except Exception:
continue
def models(self):
from models import get_models
return get_models()
