def train_agent(config, agent):
device = torch.device('cuda' if config.USE_CUDA else 'cpu')
torch.manual_seed(config.SEED)
env = neo.NeodroidFormalWrapper(
environment_name=config.ENVIRONMENT_NAME,
connect_to_running=config.CONNECT_TO_RUNNING)
env.seed(config.SEED)
agent.build(env, device)
listener = U.add_early_stopping_key_combination(agent.stop_training)
listener.start()
try:
_trained_model, running_signals, running_lengths, *training_statistics = agent.train(
env, config.ROLLOUTS, render=config.RENDER_ENVIRONMENT)
finally:
listener.stop()
U.save_statistic(running_signals,
'running_signals',
LOG_DIRECTORY=C.LOG_DIRECTORY)
U.save_statistic(running_lengths,
'running_lengths',
LOG_DIRECTORY=C.LOG_DIRECTORY)
U.save_model(_trained_model, config)
env.close()
def test_ddpg_agent(config):
'''
:rtype: object
'''
import gym
device = torch.device('cuda' if config.USE_CUDA else 'cpu')
env = gym.make(config.ENVIRONMENT_NAME)
# env = NormaliseActionsWrapper(env)
# env = neo.make('satellite',connect_to_running=False)
agent = DDPGAgent(config)
agent.build(env, device)
listener = U.add_early_stopping_key_combination(agent.stop_training)
listener.start()
try:
(actor_model, critic_model), stats = agent.train(
env, config.ROLLOUTS, render=config.RENDER_ENVIRONMENT
)
finally:
listener.stop()
U.save_model(actor_model, config, name='actor')
U.save_model(critic_model, config, name='critic')
def train_agent(config, agent):
device = torch.device('cuda' if config.USE_CUDA else 'cpu')
torch.manual_seed(config.SEED)
env = U.BinaryActionEncodingWrapper(
environment_name=config.ENVIRONMENT_NAME,
connect_to_running=config.CONNECT_TO_RUNNING)
env.seed(config.SEED)
agent.build(env, device)
listener = U.add_early_stopping_key_combination(agent.stop_training)
listener.start()
try:
_trained_model, running_signals, running_lengths, *training_statistics = agent.train(
env, config.ROLLOUTS, render=config.RENDER_ENVIRONMENT)
except ValueError:
running_signals = None
running_lengths = None
_trained_model = None
print('Training procedure did not return as excepted')
finally:
listener.stop()
U.save_statistic(running_signals,
'running_signals',
LOG_DIRECTORY=C.LOG_DIRECTORY)
U.save_statistic(running_lengths,
'running_lengths',
LOG_DIRECTORY=C.LOG_DIRECTORY)
U.save_model(_trained_model, config)
env.close()
def train(event, context):
"""Trains our ML model."""
utilities.download_directory(uri=f"s3://{DATA_BUCKET}/{DATA_PREFIX}",
dst=DATA_TMP_DST)
df = utilities.read_csv_directory(DATA_TMP_DST)
print(f"SHAPE: {df.shape}")
train, test = utilities.train_test_split(df)
X_train, y_train = utilities.Xy_split(train, target='y')
X_test, y_test = utilities.Xy_split(test, target='y')
X_train = utilities.preprocessing.preprocess(X_train)
X_test = utilities.preprocessing.preprocess(X_test)
model = utilities.Model()
model.fit(X_train, y_train)
y_hat = model.predict(X_test)
eval_results = utilities.evaluate(y_actual=y_test, y_predict=y_hat)
utilities.save_model(obj=model, uri=f"s3://{DATA_BUCKET}/{MODEL_PREFIX}")
return {
"status": "success",
"results": eval_results,
}
def main():
dataset = "context_finetuned"
# dataset = "context_pretrained"
data = getattr(utilities, f"read_{dataset}")()
data["vocab"] = sorted(
list(
set([
w for sent in data["train_sen"] + data["dev_sen"] +
data["test_sen"] for w in sent
])))
data["concept"] = sorted(
list(
set([
concept for seq in data["train_concept"] +
data["dev_concept"] + data["test_concept"] for concept in seq
])))
data["classes"] = sorted(list(set(data["train_class"])))
data["word_to_idx"] = {w: i for i, w in enumerate(data["vocab"])}
data["idx_to_word"] = {i: w for i, w in enumerate(data["vocab"])}
data["concept_to_idx"] = {w: i for i, w in enumerate(data["concept"])}
data["idx_to_concept"] = {i: w for i, w in enumerate(data["concept"])}
MAX_SENT_LEN = max([
len(sent)
for sent in data["train_sen"] + data["dev_sen"] + data["test_sen"]
])
listFilter = [[3, 4, 5], [9, 10, 11], [15, 16, 17], [21, 22, 23],
[27, 28, 29]]
FILTER_NUM = [100, 100, 100]
FILTER_PRE = 3
FILTER_NUM_PRE = 100
listModality = ["N", "D", "ND", "NK", "DK", "NDK"]
for modality in listModality:
for FILTERS in listFilter:
params = {
"MODALITY": modality,
"DATASET": dataset,
"SAVE_MODEL": True,
"EARLY_STOPPING": True,
"EPOCH": 500,
"LEARNING_RATE": 0.001,
"MAX_SENT_LEN": MAX_SENT_LEN,
"BATCH_SIZE": 50,
"WORD_DIM": 200,
"VOCAB_SIZE": len(data["vocab"]),
"CLASS_SIZE": len(data["classes"]),
"CONCEPT_SIZE": len(data["concept"]),
"FILTER_PRE": FILTER_PRE,
"FILTER_NUM_PRE": FILTER_NUM_PRE,
"FILTERS": FILTERS,
"FILTER_NUM": FILTER_NUM,
"DROPOUT_PROB": 0.5,
"NORM_LIMIT": 3,
"OPTIMIZATION": "adam"
}
model, dev_acc, dev_acc_each, listLog = train(data, params)
if params["SAVE_MODEL"]:
utilities.save_model(model, params)
torch.cuda.empty_cache()
def main(args: Optional[argparse.Namespace] = None):
return_value = True
if args is None:
args = parse_arguments()
return_value = False
caffe_model_path = os.path.join(args.model_dir, args.caffe_model)
pytorch_model_out_path = os.path.join(args.model_dir, args.out_name)
tmp_path = os.path.join(args.model_dir, 'temp.pt')
print('Converting normalized Caffe weights to a PyTorch state_dict...')
caffe_args = SimpleNamespace(model_caffemodel=caffe_model_path,
output_path=tmp_path,
caffe_proto='')
caffemodel2pytorch.main(caffe_args)
state_dict = torch.load(tmp_path)
# reshape caffe bias to match the PyTorch one
for learnables_key in state_dict:
if 'bias' in learnables_key:
state_dict[learnables_key] = state_dict[learnables_key].squeeze()
print('Loading VGG19 PyTorch model...')
# only features are needed
net = torchvision.models.vgg19(pretrained=True).features
# rename VGG19's feature layers, so that we can refer to them easily later
new_layer_names = [
'conv1_1', 'relu1_1', 'conv1_2', 'relu1_2', 'pool1', 'conv2_1',
'relu2_1', 'conv2_2', 'relu2_2', 'pool2', 'conv3_1', 'relu3_1',
'conv3_2', 'relu3_2', 'conv3_3', 'relu3_3', 'conv3_4', 'relu3_4',
'pool3', 'conv4_1', 'relu4_1', 'conv4_2', 'relu4_2', 'conv4_3',
'relu4_3', 'conv4_4', 'relu4_4', 'pool4', 'conv5_1', 'relu5_1',
'conv5_2', 'relu5_2', 'conv5_3', 'relu5_3', 'conv5_4', 'relu5_4',
'pool5'
]
net = torch.nn.Sequential(OrderedDict(zip(new_layer_names, net)))
# replace max pooling by average pooling
for i in range(len(net)):
if isinstance(net[i], torch.nn.MaxPool2d):
net[i] = torch.nn.AvgPool2d(2, stride=2)
print('Loading normalized weights into the PyTorch model...')
net.load_state_dict(state_dict)
# remove intermediate files
os.remove(tmp_path)
if return_value:
return net
else:
print('Saving the converted model..')
utilities.save_model(net, pytorch_model_out_path)
return None
def main():
print("Reading in the training data")
train = util.get_train_df()
#clean data
#print("Extracting features and training model")
#classifier = get_pipeline()
classifier = RandomForestClassifier(n_estimators = 100)
classifier.fit(train[0::,1::],train[0::,0])
#print("Saving the classifier")
util.save_model(classifier)
def regular_train_agent_procedure(agent_type, config, environment=None):
if not environment:
if '-v' in config.ENVIRONMENT_NAME:
environment = gym.make(config.ENVIRONMENT_NAME)
else:
environment = BinaryActionEncodingWrapper(
name=config.ENVIRONMENT_NAME,
connect_to_running=config.CONNECT_TO_RUNNING)
U.set_seeds(config.SEED)
environment.seed(config.SEED)
agent = agent_type(config)
device = torch.device('cuda' if config.USE_CUDA else 'cpu')
agent.build(environment, device)
listener = U.add_early_stopping_key_combination(agent.stop_training)
listener.start()
try:
models, stats = agent.train(environment,
config.ROLLOUTS,
render=config.RENDER_ENVIRONMENT)
finally:
listener.stop()
identifier = count()
if isinstance(models, list):
for model in models:
U.save_model(model,
config,
name=f'{type(agent)}-{identifier.__next__()}')
else:
U.save_model(models,
config,
name=f'{type(agent)}-{identifier.__next__()}')
stats.save()
environment.close()
def main():
"""
This is the main program
"""
# Initiating variables with parsed command line arguments
in_arg = get_input_args()
data_dir = in_arg.data_dir
train_dir = data_dir + '/train'
valid_dir = data_dir + '/valid'
test_dir = data_dir + '/test'
save_dir = in_arg.save_dir
arch = in_arg.arch.lower()
epochs = in_arg.epochs
hidden_units = in_arg.hidden_units
learning_rate = in_arg.learning_rate
gpu = in_arg.gpu
# Correct the variables if necessary to avoid incorrect calculations
# Collect error messages what variables have been changed to what values
error_messages = []
if (epochs <= 0):
epochs = 1
error_messages.append("epochs was corrected to 1")
elif (epochs > 10):
epochs = 10
error_messages.append("epochs was corrected to 10")
if (learning_rate <= 0.000001):
learning_rate = 0.00001
error_messages.append("learning_rate was corrected to 0.00001")
elif (learning_rate >= 0.1):
learning_rate = 0.01
error_messages.append("learning_rate was corrected to 0.01")
if (hidden_units < 4):
hidden_units = 4
error_messages.append("hidden_units was corrected to 4")
if not save_dir:
save_dir = os.getcwd()
save = False
elif save_dir == "/": # slash means that the new trained classified should be stored in the current directory
save = True
save_dir = os.getcwd()
else:
save = True
if path.exists(data_dir) and path.exists(train_dir) and path.exists(valid_dir) \
and path.exists(test_dir) and path.exists(save_dir): # check if all paths are correct
if (arch in "alexnet,vgg16,densenet161"
): # check if the stated architecture is supported
# define the data transforms of the train data
data_transforms_train = transforms.Compose([
transforms.RandomRotation(30),
transforms.RandomResizedCrop(224),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406],
[0.229, 0.224, 0.225])
])
# define the data transforms of the validation data
data_transforms_valid = transforms.Compose([
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406],
[0.229, 0.224, 0.225])
])
# load the image data of the train and validation set and perform transforms
train_data = datasets.ImageFolder(train_dir,
transform=data_transforms_train)
valid_data = datasets.ImageFolder(valid_dir,
transform=data_transforms_valid)
# load the transformed image data into loader variables by batches
trainloader = DataLoader(train_data, batch_size=64, shuffle=True)
validloader = DataLoader(valid_data, batch_size=64, shuffle=True)
# download the pretrained version of the selected model defined by the parser variable 'arch'
model = getattr(models, arch)
model = model(pretrained=True)
# freeze the parameters of the model as only classifier will be updated
for param in model.parameters():
param.requires_grad = False
# classifier layer will be fully connected so get the input units first
if (arch == "vgg16"):
num_in_features = model.classifier[0].in_features
elif (arch == "densenet161"):
num_in_features = model.classifier.in_features
elif (arch == "alexnet"):
num_in_features = model.classifier[1].in_features
# define the new classifier und replace the current one
new_classifier = nn.Sequential(
nn.Linear(num_in_features, hidden_units), nn.ReLU(),
nn.Dropout(0.5), nn.Linear(hidden_units,
int(hidden_units / 4)), nn.ReLU(),
nn.Dropout(0.5), nn.Linear(int(hidden_units / 4), 102),
nn.LogSoftmax(dim=1))
model.classifier = new_classifier
# use GPU power if available and model to it
if gpu:
device = torch.device(
"cuda" if torch.cuda.is_available() else "cpu")
else:
device = "cpu"
model.to(device)
# define the loss function and the optimizer
criterion = nn.NLLLoss()
optimizer = optim.Adam(model.classifier.parameters(),
lr=learning_rate)
print_every = 50
train_len = len(trainloader)
valid_len = len(validloader)
# start training and validation epochs
for epoch in range(epochs):
epoch += 1
last_print = 0
running_loss = 0
for batch, (inputs, labels) in enumerate(trainloader):
batch += 1
inputs, labels = inputs.to(device), labels.to(device)
# for each batch gradients should be zeroed
optimizer.zero_grad()
# perform feed-forward and calculate loss through back propagation
logps = model.forward(inputs)
loss = criterion(logps, labels)
loss.backward()
optimizer.step()
running_loss += loss.item()
print(f"Epoch {epoch}/{epochs}, batch {batch}/{train_len}")
# do validation on the test set after defined set of batches
if (((batch) % print_every == 0) &
(batch >= print_every)) or (batch == train_len):
valid_loss = 0
accuracy = 0
# put into evaluation mode and deactivate gradients for feed-forward validation
model.eval()
with torch.no_grad():
# iterate through the valid data set
for inputs, labels in validloader:
inputs, labels = inputs.to(device), labels.to(
device)
logps = model.forward(inputs)
# calculate the losses
batch_loss = criterion(logps, labels)
valid_loss += batch_loss.item()
# calculate accuracy and return the category with the top probability
# then compare with the labels and calculate the mean of the right matches
ps = torch.exp(logps)
top_p, top_class = ps.topk(1, dim=1)
equals = top_class == labels.view(
*top_class.shape)
accuracy += torch.mean(
equals.type(torch.FloatTensor)).item()
# after each train and validation circle defined by the number of batches before print the statistics
if (batch == train_len):
print(
f"Train loss: {running_loss/(train_len - last_print):.3f}.. ",
end='')
else:
last_print += print_every # print statistics after each epoch
print(
f"Train loss: {running_loss/print_every:.3f}.. ",
end='')
print(f"Valid loss: {valid_loss/valid_len:.3f}.. ",
end='')
print(f"Accuracy: {accuracy/valid_len:.3f}")
running_loss = 0
# switch back to train mode
model.train()
# if save path is defined
if (save == True):
save_model(model, save_dir, train_data.class_to_idx, arch,
new_classifier)
else:
print("Architecture chosen not supported!")
else:
print("Incorrect directories - please check!")
# print out error messages if any
if (len(error_messages)):
for v in error_messages:
print(v)
logging.info("*** Describe training data ***")
ut.describe_data(X_train, Y_train)
logging.info("*** Describe validation data ***")
ut.describe_data(X_val, Y_val)
## Create model
tstart = time.time()
Y_train_c = to_categorical(Y_train, num_classes=2)
Y_val_c = to_categorical(Y_val, num_classes=2)
if args.model_name == "mnist":
model, history = mtt.mnist_cnn(X_train,
Y_train_c,
X_val,
Y_val_c,
batch_size=128,
epochs=int(args.epochs))
elif args.model_name == "too_simple":
model, history = mtt.too_simple(X_train,
Y_train_c,
X_val,
Y_val_c,
batch_size=128,
epochs=int(args.epochs))
tend = time.time()
logging.info("Time to fit model: {0:.1f} s.".format(tend - tstart))
## Save model
ut.save_model(args.save_dir, model, history, X_val, Y_val_c)
logging.info("Finished cnn.py")
def save(self, C):
U.save_model(self._policy, C)
train_accuracy = sess.run(acc, feed_dict=feed_dict_train)
# train_accuracy, summary = sess.run([acc_1, merged_train], feed_dict=feed_dict_train)
# writer.add_summary(summary, epoch)
print_and_log_timestamp("epoch {}/{}: train accuracy is {}",
epoch + 1, NUM_EPOCHS, train_accuracy)
if train_accuracy < MIN_ACCURACY_DISCARD and epoch >= MIN_EPOCHS_DISCARD:
print_and_log_timestamp("model is not learning...exit")
exit(1)
if train_accuracy > best_test_accuracy:
test_accuracy, test_summary = sess.run(
[acc, merged_test], feed_dict=feed_dict_test)
writer.add_summary(test_summary, epoch)
print_and_log_timestamp("epoch {}/{}: test accuracy is {}",
epoch + 1, NUM_EPOCHS, test_accuracy)
if test_accuracy > best_test_accuracy:
best_test_accuracy = test_accuracy # updating best test accuracy
save_model(saver, sess) # save model
if train_accuracy - test_accuracy > DROPOUT_UPDATE_MIN_GAP and train_dropout > MIN_DROPOUT:
train_dropout = train_dropout * DROPOUT_UPDATE_FACTOR # updating dropout
print_and_log_timestamp(
"**** updating dropout value to {} ****",
train_dropout)
writer.close()
print_and_log('\n')
print_and_log_timestamp(" END session! {}",
datetime.datetime.now().strftime("%H:%M:%S"))
copyfile(CM.TEMP_LOG_FILE_PATH, CM.LOG_FILE_PATH)
def save(self, C): U.save_model(self._actor, C, 'actor') U.save_model(self._critic, C, 'policy')
return -1.0
try:
#def calculate():
if cuda.is_available():
print('cuda is available!')
# Build the model
# self.conv1=nn.DataParallel(self.conv1)
encoder=torch.load('model8/encoder')
# encoder=EncoderRNN()
# encoder =nn.DataParallel(encoder)
# decoder = DecoderRNN()
decoder=torch.load('model8/decoder')
if cuda.is_available():
encoder.cuda()
decoder.cuda()
# encoder = nn.DataParallel(encoder)
#encoder=nn.DataParallel(encoder)
# decoder=nn.DataParallel(decoder,dim=1,device_ids=[0,1])
train_iters(encoder, decoder)
save_model(encoder, decoder)
writer.export_scalars_to_json("./all_scalars.json")
writer.close()
except Exception as e: # если вдруг произошла какя-то ошибка при обучении, сохраняем модель
print(e)
save_model(encoder, decoder)
#calculate()
n_classes = 10
epochs = 20 # int(input("Epochs: "))
batch_size = 1000 # int(input("batch_size: "))
lr = 0.01 # float(input("lr: "))
model = Classifier(input_shape, conv, fc, n_classes).to(device)
criterion = nn.CrossEntropyLoss().to(device)
# optimizer = optim.SGD(model.parameters(), lr=0.001, momentum=0.9)
optimizer = optim.Adagrad(model.parameters(), lr=lr)
tr.train(model, dataset, criterion, optimizer, device, epochs, batch_size)
print("Saving Black Box:")
utilities.save_model(model)
if args['sub']:
bb = input("Black box model name: ")
oracle_model = torch.load("saved_models/" + bb)
test_dataset = dataset["eval"]
dataset = dataset["train"]
num_points = 150 # int(input("Num Points Initial: "))
idxs = []
for _ in range(num_points):
while True:
idx = random.randint(0, len(dataset) - 1)
if idx not in idxs:
def save(self, C):
U.save_model(self._actor_critic, C)
def save(self, C): U.save_model(self._value_model, C)
def main():
#print(f'datetime.datetime={str(datetime.datetime)}, datetime.date={str(datetime.date)}, strftime():{datetime.datetime.now().strftime("%Y%d%H%M%S")}')
thrash = True
print('tf version:{0}'.format(tf.VERSION))
print('tf.keras version:{0}'.format(tf.keras.__version__))
start_time = datetime.datetime.now().strftime("%Y%d%H%M%S")
flags, unparsed = ut.parseArgs()
print(flags)
SAMPLE_FILE = flags.train_data_path + flags.sample + '.' + flags.img_file_extension
img = ut.read_image(filename=SAMPLE_FILE, show=False)
img = np.array(img)
if thrash == True:
img = ut.thrash_img(img)
IMG_SHAPE=img.shape
(x_train, y_train), (x_test, y_test)=ut.load_data(numclasses=flags.numclasses, train_path=flags.train_data_path, test_path=flags.test_data_path, onehot=True, extension=flags.img_file_extension)
print('IMG_SHAPE:{0}, y_train shape:{1}'.format(IMG_SHAPE,y_train[0].shape))
if flags.load_model:
model = ut.load_stored_model(name=flags.model_dir + flags.model_name)
elif flags.model == 'dense':
model = ut.make_dense_model(flags=flags)
elif flags.model == 'conv2d':
model = ut.make_convnet_model(flags=flags, shape=IMG_SHAPE)
else:
print('No model, no hope. Quitting...')
return
if flags.load_data:
model = ut.load_stored_data(model=model, date_file_name=flags.data_dir + flags.data_name)
print('Saving in {0}'.format(flags.tb_dir + start_time))
tensorboard = TensorBoard(log_dir=flags.tb_dir + '{0}'.format(start_time))
adam=tf.keras.optimizers.Adam(lr=flags.learning_rate)
model.compile(optimizer=adam,
loss=flags.loss,
metrics=[flags.metric]
)
if flags.train == True:
print('Training...')
scores = []
for epoch in range(flags.epochs):
print('Epoch:{0} of {1}'.format(epoch+1, flags.epochs))
n = len(x_train)
for batch in range(0,len(x_train), flags.batch_size):
print('Batch {0} of {1}, epoch {2} of {3}.'.format(batch+1,n+1, epoch+1, flags.epochs))
bunch_x, bunch_y = x_train[batch:batch+flags.batch_size], y_train[batch:batch+flags.batch_size]
if len(bunch_x) < flags.batch_size: # skip partial batches
print('Skipping {0} samples..'.format(len(bunch_x)))
continue
xs = []
ys = []
print("Iterating {0} samples".format(len(bunch_x)))
for datum in range(len(bunch_x)):
file = bunch_x[datum]
img = ut.read_image(filename=flags.train_data_path+file, show=False)
img=np.array(img)
if thrash == True:
img = ut.thrash_img(img)
xs.append(img)
ys.append(bunch_y[datum])
X= np.stack(xs, axis=0)
Y= np.stack(ys, axis=0)
score_before = model.evaluate(x=X,y=Y, batch_size=flags.batch_size)
_ = model.fit(x=X, y=Y, shuffle=flags.shuffle, callbacks=[tensorboard])
score_after = model.evaluate(x=X,y=Y, batch_size=flags.batch_size)
if score_before == score_after:
print("Scores before and after training are identical")
scores.append(score_after)
if epoch == 0 and batch == 0:
model.summary()
print('Score:{0}'.format(score_after))
loss,acc = np.array([s[0] for s in scores]), np.array([s[1] for s in scores])
print("Average loss:{0} Average accuracy:{1}%".format(np.mean(loss), 100*np.mean(acc)))
if flags.save_model:
model_name = flags.model_name if flags.model_name != None else start_time
ut.save_model(model, flags.model_dir+model_name)
print('Saved model to disk, json in {0}'.format(flags.model_dir + model_name + ".json"))
if flags.save_data:
data_name = flags.data_name if flags.data_name != None else start_time
model.save_weights(flags.data_dir + data_name + ".h5")
print('Saved data to disk in {0}'.format(flags.model_dir + data_name + ".h5"))
test_scores = []
predictions = []
if flags.evaluate or flags.predict:
n = len(x_test)
nTotal = 0
sums_array = None
for batch in range(0, len(x_test), flags.batch_size):
print('Batch {0} of {1}.'.format(batch+1, n+1))
bunch_x, bunch_y = x_test[batch:batch + flags.batch_size], y_test[batch:batch + flags.batch_size]
if len(bunch_x) < flags.batch_size: # skip partial batches
print('Skipping {0} samples..'.format(len(bunch_x)))
continue
xs = []
ys = []
for d in range(len(bunch_x)):
file = bunch_x[d]
img = ut.read_image(filename=flags.test_data_path + file, show=False)
img = np.array(img)
if thrash == True:
img = ut.thrash_img(img)
xs.append(img)
ys.append(bunch_y[d])
X = np.stack(xs, axis=0)
Y = np.stack(ys, axis=0)
if flags.evaluate:
score = model.evaluate(x=X, y=Y, batch_size=flags.batch_size)
test_scores.append(score)
print('Test score:{0}'.format(score))
if flags.predict:
prediction = model.predict(X, verbose=2)
processed_predictions = ut.process_predictions(prediction, Y)
for pp in processed_predictions:
if sums_array is None:
sums_array = np.zeros_like(pp)
sums_array = np.add(sums_array, pp)
nTotal = nTotal+1
pass
if flags.predict:
sums_array /= nTotal
if predictions != None:
pass
print('Average score:{0},{1}'.format(np.mean([test_scores[i][0] for i in range(len(test_scores))]),np.mean([test_scores[i][1] for i in range(len(test_scores))])))
if flags.show_results:
y_axis = np.arange(0, 1.0, 1.0/float(len(sums_array)))
plt.plot(y_axis,sums_array)
plt.show()
pass
