# encoding: utf-8 """ @author : zhirui zhou @contact: [email protected] @time : 2020/6/1 16:37 """ from deepseries.data import create_seq2seq_data_loader import numpy as np enc_len = 12 dec_len = 8 series = np.random.rand(1000, 8, 100) data_loader = create_seq2seq_data_loader(series, enc_len, dec_len, time_idx=np.arange(series.shape[2]), batch_size=32, seq_last=True) for i in data_loader: pass i[0]['enc_x'].shape i[1].shape
series, series_mean, series_std = F.normalize(series[:, np.newaxis,
DROP_BEFORE:],
axis=2)
series_lags = F.normalize(series_lags[:, :, DROP_BEFORE:])[0]
series_lags = Value(series_lags, 'xy_lags')
time_idxes = np.arange(series.shape[2])
trn_idx, val_idx = forward_split(time_idxes, ENC_LEN, VALID_LEN + TEST_LEN)
val_idx, test_idx = forward_split(val_idx, ENC_LEN, TEST_LEN)
trn_dl = create_seq2seq_data_loader(series,
enc_len=ENC_LEN,
dec_len=DEC_LEN,
time_idx=trn_idx,
batch_size=BATCH_SIZE,
features=[series_lags, series_lags_corr],
seq_last=True,
device='cuda',
mode='train',
num_workers=0,
pin_memory=False)
val_dl = create_seq2seq_data_loader(series,
enc_len=ENC_LEN,
dec_len=DEC_LEN,
time_idx=val_idx,
batch_size=BATCH_SIZE,
features=[series_lags, series_lags_corr],
seq_last=True,
device='cuda',
mode='valid')
import numpy as np
from torch.optim import Adam
batch_size = 16
enc_len = 36
dec_len = 12
series = np.sin(np.arange(0, 1000))
series = series.reshape(1, 1, -1)
train_idx, valid_idx = forward_split(np.arange(series.shape[2]),
enc_len=14,
valid_size=200)
train_dl = create_seq2seq_data_loader(series,
enc_len=14,
dec_len=7,
time_idx=train_idx,
batch_size=12,
sampling_rate=1.,
seq_last=True)
valid_dl = create_seq2seq_data_loader(series,
enc_len=14,
dec_len=7,
time_idx=valid_idx,
batch_size=12,
sampling_rate=1.,
seq_last=True)
model = Wave2Wave(1, debug=False, num_layers=5, num_blocks=1)
model.cuda()
opt = Adam(model.parameters(), 0.001)
learner = Learner(model, opt, ".")
learner.fit(100, train_dl, valid_dl, early_stopping=False)
train_idx, valid_idx = forward_split(np.arange(series_len),
enc_len=enc_len,
valid_size=valid_size + test_size)
valid_idx, test_idx = forward_split(valid_idx, enc_len, test_size)
# mask test, will not be used for calculating mean/std.
mask = np.zeros_like(series).astype(bool)
mask[:, :, test_idx] = False
series, mu, std = F.normalize(series, axis=2, fillna=True, mask=mask)
# wave2wave train
train_dl = create_seq2seq_data_loader(series[:, :, train_idx],
enc_len,
dec_len,
sampling_rate=0.1,
batch_size=batch_size,
seq_last=True,
device='cuda')
valid_dl = create_seq2seq_data_loader(series[:, :, valid_idx],
enc_len,
dec_len,
batch_size=batch_size,
seq_last=True,
device='cuda')
wave = Wave2Wave(target_size=1,
num_layers=6,
num_blocks=1,
dropout=0.1,
loss_fn=RMSE())