def train(args):
traj, n_traj, translator = traj_prepare(
TRAJ_PATH,
CODE_PATH,
NUM_SAMPLES,
use_cols=args.begin_index +
np.arange(args.num_per_day) * args.time_interval,
add_idx=ADD_IDX)
cond, n_categ_cond = cond_prepare(COND_PATH, NUM_SAMPLES)
cond = cond[:, 1] # only use job here
n_categ_cond = n_categ_cond[1]
cond = trans_one_hot(cond.flatten(), n_categ_cond).astype(np.float32)
NUM_CLASS = n_traj
# add SOS
num_training_sample = int(NUM_SAMPLES * 0.8)
ix_sos = translator.trans_code2ix(["0/SOS"])[0]
traj = np.concatenate([np.ones((traj.shape[0], 1), dtype=int), traj],
axis=1)
training_data, validation_data = traj[:num_training_sample, :-1], traj[
num_training_sample:, :-1]
training_label, validation_label = traj[:num_training_sample,
1:], traj[num_training_sample:, 1:]
training_cond, validation_cond = cond[:num_training_sample], cond[
num_training_sample:]
training_loader = torch.utils.data.DataLoader(TrajDatasetGeneral(
[training_data, training_label, training_cond]),
batch_size=args.batch_size,
shuffle=True,
num_workers=4,
drop_last=False)
validation_loader = torch.utils.data.DataLoader(TrajDatasetGeneral(
[validation_data, validation_label, validation_cond]),
batch_size=2000,
shuffle=False,
num_workers=4,
drop_last=False)
seq_vae = CondVAE(num_class=NUM_CLASS,
embedding_size=args.embed_size,
embedding_drop_out=args.emb_dropout,
hid_size=args.hid_size,
hid_layers=args.hid_layer,
latent_z_size=args.latent_z_size,
cond_size=n_categ_cond,
word_dropout_rate=args.word_dropout,
gru_drop_out=args.gru_dropout,
anneal_k=args.anneal_k,
anneal_x0=args.anneal_x0,
anneal_function=args.anneal_func,
data_keys=DATA_KEYS,
learning_rate=args.lr,
sos_idx=ix_sos,
unk_idx=None)
test_batch = wrap_data(next(iter(training_loader)), DATA_KEYS)
model_summary = torchsummaryX.summary(seq_vae.model, test_batch['data'],
test_batch['cond'])
seq_vae.model.to(DEVICE)
seq_vae.fit(training_loader=training_loader,
validation_loader=validation_loader,
epochs=args.epochs,
device=DEVICE)
return seq_vae, translator
default=5e-3,
metavar='LR',
help='learning rate')
args, unknown = parser.parse_known_args()
torch.manual_seed(3)
np.random.seed(3)
source_folder = "/home/ubuntu/pflow_data/filtered"
traj_path = f"{source_folder}/traj.csv"
code_path = f"{source_folder}/code.csv"
cond_path = f"{source_folder}/cond.csv"
num_samples = 180000
traj, n_traj, translator = traj_prepare(
traj_path, code_path, num_samples,
use_cols=360 + np.arange(18 * 10) * 6) #, add_idx=["0/SOS"])
#TODO check here
cond, n_categ = cond_prepare(cond_path, num_samples)
NUM_CLASS = n_traj
num_training_sample = int(num_samples * 0.8)
training_data, validation_data = traj[:num_training_sample, :-1], traj[
num_training_sample:, :-1]
training_label, validation_label = traj[:num_training_sample,
1:], traj[num_training_sample:, 1:]
training_cond, validation_cond = cond[:num_training_sample], cond[
num_training_sample:]
training_loader = torch.utils.data.DataLoader(TrajDatasetCond(
default=5e-3,
metavar='LR',
help='learning rate')
args, unknown = parser.parse_known_args()
torch.manual_seed(3)
np.random.seed(3)
source_folder = "/home/ubuntu/pflow_data/filtered"
traj_path = f"{source_folder}/traj.csv"
code_path = f"{source_folder}/code.csv"
num_samples = 50000
traj, n_traj, translator = traj_prepare(
traj_path,
code_path,
num_samples,
use_cols=args.begin_index +
np.arange(args.num_per_day) * args.time_interval)
#add_idx=["0/SOS"])
NUM_CLASS = n_traj
DATA_KEYS = ["data", "label"]
num_training_sample = int(num_samples * 0.8)
training_data, validation_data = traj[:num_training_sample, :-1], traj[
num_training_sample:, :-1]
training_label, validation_label = traj[:num_training_sample,
1:], traj[num_training_sample:, 1:]
training_loader = torch.utils.data.DataLoader(TrajDatasetGeneral(
[training_data, training_label]),