def main(): # Load Experiment Specifications precursor = sys.argv[1] ID = int(sys.argv[2]) factor = int(sys.argv[3]) ID += factor * 1000 opt = inst.gen_tune_exp(precursor)[ID] opt.reinitialize_paths() print(opt) #TRAIN tf.reset_default_graph() if opt.mode == "train" or opt.mode == 'both': train_model = model.Autoencoder(opt) with tf.Session() as sess: train_model.train(sess) #TEST tf.reset_default_graph() if opt.mode == "test" or opt.mode == 'both': test_model = model.Autoencoder(opt) with tf.Session() as sess: test_model.tester(sess)
def __init__(self, model_name="model.pt", data=None): if model_name is None: import model self.enc = model.Autoencoder() else: self.enc = torch.load(os.path.join(dir_path, model_name), map_location='cpu') self.enc.eval()
def main(): m = model.Autoencoder() m.load_state_dict(torch.load("autoencoder_save.pth")) decoder = m.decoder.eval() for i in range(100): # 100 random samples rnd_latent_vector = torch.rand(1, 64, 1, 1) * 2 - 1 output = decoder(rnd_latent_vector) image = ld.resize_from_3x32x32_to_32x32x3( np.squeeze(output.data.numpy(), 0)) image = (image * 255).astype(np.uint8) ld.save_img(image, "generated data/cifar10_generated_image" + str(i))
def train_and_evaluate(args): Xgbmodel = models.Xgb(args) AEmodel = models.Autoencoder(args) bst, train_seen_bidids = train(args, Xgbmodel, AEmodel) auc, valid_seen_bidids = evaluate(args, Xgbmodel, AEmodel, bst, train_seen_bidids) print('Validation AUC: {:.5f}'.format(auc)) f = open(args.filepath, 'a') f.write('Validation AUC: %.5f\n' % (auc)) f.close() if not args.cv: auc, _ = evaluate(args, Xgbmodel, AEmodel, bst, train_seen_bidids, valid_seen_bidids, False) print('Test AUC: {:.5f}'.format(auc)) f = open(args.filepath, 'a') f.write('Test AUC: %.5f\n' % (auc)) f.close()
the games of the masters from the website "https://www.pgnmentor.com/players/" """ def get_move(line): line = str(line) game = [] # list of moves in pgn notation for move in re.findall(r"[^{\[.}\]]+ ", line.replace("?", "").replace( "!", "")): # Extract all moves without comments game.extend(move.strip().split( " ")) # Sometimes one string contains two moves with space between return game coder = model.Autoencoder(settings.BOARD_SHAPE, settings.LATENT_SIZE).to(settings.DEVICE) coder.load_state_dict( torch.load(settings.CODER_PATH, map_location=settings.DEVICE)) coder = coder.coder coder.eval() inf = Inference(settings.DEVICE, coder) csv_name = "positions_lite.csv" ID = 0 with open(csv_name, "w", newline="") as file: writer = csv.writer(file, delimiter=";") writer.writerow(["ID", "Author", "Number", "Move", "Embeding"]) games_csv = open(os.path.join(os.getcwd(), 'games_lite.csv')) for row in games_csv: try: data = row[:-2].split(";")
# Normalize in [0, 1] r = df['rating'].values.astype(float) min_max_scaler = preprocessing.MinMaxScaler() x_scaled = min_max_scaler.fit_transform(r.reshape(-1, 1)) df_normalized = pd.DataFrame(x_scaled) df['rating'] = df_normalized # Convert DataFrame in user-item matrix matrix = df.pivot(index='user_id', columns='item_id', values='rating') matrix.fillna(0, inplace=True) # Users and items ordered as they are in matrix users = matrix.index.tolist() items = matrix.columns.tolist() matrix = matrix.values print("Matrix shape: {}".format(matrix.shape)) # num_users = matrix.shape[0] # num_items = matrix.shape[1] # print("USERS: {} ITEMS: {}".format(num_users, num_items)) #%% Define and train model mymodel = model.Autoencoder(input_size=len(items), hidden_layer_size=100) mymodel.fit(X=matrix, epochs=200)
#!/usr/bin/env python3 import numpy as np import sys, os import pickle, gzip import pandas as pd import gc, gzip import model import torch enc = model.Autoencoder() enc.eval() def encode(x): x = torch.from_numpy(x) emb = enc.autoencode_1.encode(x[None, None, :])[0, :, 0].detach().numpy() return emb test_labels = pd.read_csv("test_labels.csv.zip") test_labels.set_index("id", inplace=True) frame_length = 2**11 + 1 filename = "" labels = None results = [] #f= open("test_emb.csv","w+") f = gzip.open('test_emb_enc.csv.gz', 'wt') f.write("sample, segment, frame," + ",".join(map(str, range(frame_length))) + " \n")
import model import mxnet as mx #implementation #dataset = MNIST or FashionMNIST result = model.Autoencoder(epoch=1, batch_size=128, save_period=100, load_period=100, weight_decay=0.0001, learning_rate=0.001, dataset="MNIST", ctx=mx.gpu(0)) print("///" + result + "///")
import model import mxnet as mx #implementation #dataset = MNIST or FashionMNIST result = model.Autoencoder(epoch=0, batch_size=128, save_period=100, load_period=100, optimizer="adam", learning_rate=0.001, dataset="FashionMNIST", ctx=mx.gpu(0)) print("///" + result + "///")
''' Import ''' import idx2numpy as idx import numpy as np import scipy.misc import img_dataset as ig import tensorflow as tf import model as md print('beginning training') minst = ig.ImgDataset("train-images.idx3-ubyte") ds = md.Autoencoder() ds.train() print('training complete')