def main():
#Data.highlighted_image.browse(Data.dirs.images.images, Data.dirs.images.cell)
#return
images = Data.ImageIterableDataset(
(Data.dirs.images.images, Data.dirs.images.cell),
Data.data_aug.flip_scale_pipeline,
Data.CropGenerator(
(256, 256),
validate_crop=lambda t: .1 < t[1].mean(), # cell prominence
),
n_streams=8,
)
images = torch.utils.data.DataLoader(images, batch_size=64, drop_last=True)
ae = Models.Autoencoders.VAE(2048, 2)
Models.LayerInfo.on()
print(ae)
out = ae(next(iter(images)))
Models.LayerInfo.off()
Training.VAE.train(ae, images, epochs=256)
import os
from timeit import default_timer as timer
from src import Plotting as p
import pickle
n = 1000 # n: Numero de agentes
s = 8 # s: Numero de estrategias por agente
g_a = 60 # g_a: periodo de DGA
g_m = 0.5 # g_m: ratio de mutacion
g_p = 8 # g_p: tamao de la 'communication pool'
p_max = 10 # p_max: pos max de cada agente
ncycles = 100 # ncycles: Numero de ciclos en los que el algoritmo recorre el periodo de optimizacion
max_t = 20 # Valor max que pueden alcanzar los valores Tm, Td, Ty dentro de las estragias
stock = 'SP 500'
df = Data.get_pandas_dataframe(r'datasets/SP500.csv',
'Open').drop(columns=['index'])
title = 'SP 500'
prices = list(df.Price.values)
dates = list(df.Date.values)
Mt = list(df.Mt.values)
Dt = list(df.Dt.values)
Yt = list(df.Yt.values)
initial_date = '1990-01-02'
cut_date = '2005-05-12'
end_date = '2012-02-10'
initial_idx = dates.index(initial_date)
cut_idx = dates.index(cut_date)
end_idx = dates.index(end_date)
elif message.content.startswith(Constants.status):
await statusPrompt.print_current_statuses(message)
elif message.content.startswith(Constants.my_status):
await statusPrompt.print_my_status(message)
elif message.content.startswith(Constants.heroAction):
await actionPrompt.do_action(message)
elif message.content.startswith(Constants.partyAction):
await roomChange.parse_action(message)
data = Data.Data({
Constants.items: [],
Constants.characters: [],
Constants.rooms: copy.copy(Constants.room_names),
Constants.doors: copy.copy(Constants.door_names),
})
data.current_room.objects.append(
Character(
{Constants.health: 500,
Constants.value: 50,
Constants.attack: 0,
Constants.speed: 0,
Constants.mana: 0,
Constants.crit: 0,
Constants.name: "Target_Dummy",
Constants.description: "An unassuming target dummy",
Constants.inventory: [Item({
Constants.name: "Head",
Constants.description: "The head of your conquered foe, the target dummy",
from __future__ import print_function import keras.backend as K import numpy as np from keras.utils import plot_model import src.Data as Data import src.net as net ''' the file can draw the structure of the neural network ''' (x_train, y_train),(x_test, y_test),input_shape, batch_size, num_classes, epoches = Data.getMiniData() np.random.seed(14343) cnn = net.cnnOneLayer(input_shape, 2, (4, 4), batch_size=128) plot_model(cnn, show_shapes=True, show_layer_names=True,to_file='./logs/cnn.pdf') np.random.seed(14343) fcnn = net.fcOneLayer((13 * 13, 16), batch_size=128) plot_model(fcnn, show_shapes=True, show_layer_names=True,to_file='./logs/fc.pdf') K.clear_session()
type=str,
default='model_files/MIE',
help='Location to save.')
args = parser.parse_args()
# import pdb; pdb.set_trace()
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu_id
dictionary = Dictionary()
dictionary.load('./data/dictionary.txt')
ontology = Ontology(dictionary)
ontology.add_raw('./data/ontology.json', '状态')
ontology.add_examples('./data/example_dict.json')
data = Data(100, dictionary, ontology)
data.add_raw('train', './data/train.json', 'window')
data.add_raw('test', './data/test.json', 'window')
data.add_raw('dev', './data/dev.json', 'window')
# params of the model.
params = {
"add_global": args.add_global,
"num_units": args.hidden_size,
"num_layers": args.mlp_layer_num,
"keep_p": args.keep_p
}
# Initialize the model.
model = MIE(data, ontology, params=params)
def main():
print("================================")
print(" AIStocksAgents-v1 ")
print("================================")
print("Select source .csv file (/full/path/to/file):")
file = input().strip()
print(
"Select initial date of optimization period (same format as in .csv files):"
)
initial_date = input().strip()
print(
"Select cut date of optimization period (same format as in .csv files):"
)
cut_date = input().strip()
print("Select end date of test period (same format as in .csv files):")
end_date = input().strip()
print("Select number of optimzation cycles:")
ncycles = int(input().strip())
print("Select foder where results will be stored (/full/path/to/folder):")
results = input().strip()
n = 1000 # n: Number of agents
s = 8 # s: Number of strategies per agent
g_a = 60 # g_a: DGA rounds length
g_m = 0.5 # g_m: Mutation rate
g_p = 8 # g_p: Size of the communication pool
p_max = 10 # p_max: Max position held by agent
max_t = 20
df = dat.get_pandas_dataframe(file, 'Open').drop(columns=['index'])
prices = list(df.Price.values)
dates = list(df.Date.values)
Mt = list(df.Mt.values)
Dt = list(df.Dt.values)
Yt = list(df.Yt.values)
initial_idx = dates.index(initial_date)
cut_idx = dates.index(cut_date)
end_idx = dates.index(end_date)
print('Initial date (index in series) is ' + initial_date + '(' +
str(initial_idx) + ')')
print('Cut date (index in series) is ' + cut_date + '(' + str(cut_idx) +
')')
print('End date (index in series) is ' + end_date + '(' + str(end_idx) +
')')
print('Lenght in timesteps of the optimization period: ' +
str(cut_idx - initial_idx))
print('Lenght in timesteps of the test period: ' + str(end_idx - cut_idx))
alg = Alg.Algorithm(prices, Mt, Dt, Yt, n, s, g_a, g_m, g_p, p_max,
ncycles, initial_idx, cut_idx, end_idx, max_t)
start = timer()
print('Starting the simulation...')
pool, all_agents, statistics_opt, avg_wealth_t_test, avg_position_t_test, max_opt, min_opt = alg.run(
)
end = timer() - start
print('Duration of the execution ', f.normalize_seconds(end))
tmp = str(datetime.datetime.now())
directory = results + '/' + tmp
os.mkdir(directory)
last_price_opt = prices[cut_idx]
last_price_test = prices[end_idx]
p.save_histogram(all_agents,
'Number of agents by wealth - Histogram - Optimization',
25, directory + '/histogram-optimization-period.png')
p.save_histogram([a.get_wealth(last_price_test) for a in pool],
'Number of agents by wealth - Histogram - Test', 25,
directory + '/histogram-test-period.png')
# Position + Price chart
plt.rcParams["figure.figsize"] = (12, 8)
fig, ax1 = plt.subplots()
color = 'black'
ax1.set_xlabel('Day')
ax1.set_ylabel('Stcok')
ax1.plot(prices[cut_idx:end_idx], color=color)
ax2 = ax1.twinx()
color = 'orange'
ax2.set_ylabel('Average Position', color=color)
ax2.plot(avg_position_t_test, color=color, linewidth=0.8, alpha=0.7)
ax2.tick_params(axis='y', labelcolor=color)
fig.tight_layout()
plt.savefig(directory + '/position-price-test-chart.png')
# Wealth + Price chart
fig, ax1 = plt.subplots()
color = 'black'
ax1.set_xlabel('Day')
ax1.set_ylabel('Stock')
ax1.plot(prices[cut_idx:end_idx], color=color)
ax2 = ax1.twinx()
color = 'red'
ax2.set_ylabel('Mean Wealth', color=color)
ax2.plot(avg_wealth_t_test, color=color)
ax2.tick_params(axis='y', labelcolor=color)
fig.tight_layout()
plt.savefig(directory + '/wealth-price-test-chart.png')
test_result = [agent.get_wealth(last_price_test) for agent in pool]
file = open(directory + '/report.txt', 'w')
p.out_report(file, alg, statistics_opt, max_opt, min_opt, test_result,
'Stock')
parser.add_argument('-lr',
'--lrate',
type=float,
help='learning rate',
required=True)
args = parser.parse_args()
layers = [int(item) for item in args.layers.split(',')]
sim_id = args.id
lr = args.lrate
myModel = CNN2DModel(num_gpus=1, sim_id=sim_id)
myModel.build_model(blocks=layers)
myModel.compile_model(lr=lr, verbose=True)
myModel.build_callbacks('./logs' + str(sim_id))
#witout generators
data = Data()
X_data, Y_data = data.load_XY(pathX='./train/X_data.npz',
pathY='./train/Y_data.npz')
#load checkpoint
myModel.load_checkpoint()
myModel.fit_model(X_data, Y_data, epochs=80)
import os
from timeit import default_timer as timer
from src import Plotting as p
import pickle
n = 1000 # n: Numero de agentes
s = 8 # s: Numero de estrategias por agente
g_a = 60 # g_a: periodo de DGA
g_m = 0.5 # g_m: ratio de mutacion
g_p = 8 # g_p: tamao de la 'communication pool'
p_max = 10 # p_max: pos max de cada agente
ncycles = 100 # ncycles: Numero de ciclos en los que el algoritmo recorre el periodo de optimizacion
max_t = 20 # Max value
stock = 'NKY 225'
df = Data.get_pandas_dataframe(r'datasets/NKY225.csv',
'Close').drop(columns=['index'])
title = 'NKY 225'
prices = list(df.Price.values)
dates = list(df.Date.values)
Mt = list(df.Mt.values)
Dt = list(df.Dt.values)
Yt = list(df.Yt.values)
initial_date = '1993-03-17'
cut_date = '2005-05-12'
end_date = '2012-05-17'
initial_idx = dates.index(initial_date)
cut_idx = dates.index(cut_date)
end_idx = dates.index(end_date)