trSet = ngsimDataset('data/TrainSet.mat')
valSet = ngsimDataset('data/ValSet.mat')
trDataloader = DataLoader(trSet,
batch_size=batch_size,
shuffle=True,
num_workers=8,
collate_fn=trSet.collate_fn)
valDataloader = DataLoader(valSet,
batch_size=batch_size,
shuffle=True,
num_workers=8,
collate_fn=valSet.collate_fn)
# Initialize network
net = highwayNet(params, newFeats=newFeats, behavFeats=behavFeats)
print(net)
if params.use_cuda:
net = net.cuda()
# Set the random seed for reproducible experiments
random.seed(30)
torch.manual_seed(230)
if params.use_cuda: torch.cuda.manual_seed(230)
# This corrects for the differences in dropout, batch normalization during training and testing.
# No dropout, batch norm so far; but it is a good default practice anyways
net.train()
## Initialize optimizer
pretrainEpochs = 0
args['use_cuda'] = True
args['encoder_size'] = 64 # lstm encoder hidden state size, adjustable
args['decoder_size'] = 128 # lstm decoder hidden state size, adjustable
args['in_length'] = 16
args['out_length'] = 5
args['grid_size'] = (13, 3)
args[
'input_embedding_size'] = 32 # input dimension for lstm encoder, adjustable
args['train_flag'] = True
start_time = datetime.datetime.now()
# Initialize network
net = highwayNet(args)
if args['use_cuda']:
net = net.cuda()
## Initialize optimizer
trainEpochs = 10
optimizer = torch.optim.Adam(net.parameters()) #lr = ...
batch_size = 128
crossEnt = torch.nn.BCELoss() # binary cross entropy
## Initialize data loaders
trSet = ngsimDataset('../../data/trajectory/TrainSet.mat')
valSet = ngsimDataset('../../data/trajectory/ValSet.mat')
trDataloader = DataLoader(trSet,
batch_size=batch_size,
shuffle=True,
utils.set_logger(os.path.join(args['model_dir'], 'evaluate.log'))
json_path = os.path.join(args['model_dir'], 'params.json')
assert os.path.isfile(
json_path), "No json configuration file found at {}".format(json_path)
params = utils.Params(json_path)
params.grid_size = (params.grid_size_lon, params.grid_size_lat)
# use GPU if available
params.use_cuda = torch.cuda.is_available()
params.train_flag = args['train_flag']
params.model_dir = args['model_dir']
# Initialize network
batch_size = 128
net = highwayNet(params)
net_path = os.path.join(args['model_dir'], 'best.pth.tar')
assert os.path.isfile(net_path), "No net file found at {}".format(net_path)
#net.load_state_dict(torch.load('trained_models/cslstm_m.tar'))
utils.load_checkpoint(net_path, net)
if params.use_cuda:
net = net.cuda()
net.eval()
# https://scipython.com/blog/visualizing-the-bivariate-gaussian-distribution/
def bivariate_normal(X,
