if __name__ == "__main__": args = parse_args() dataset = args.dataset mini_batch_size = args.mini_batch_size mini_batch_size_valid = args.mini_batch_size_valid time_length = args.time_length rnn_type = args.rnn_type # Make sure we don't have skip_connections with only one hidden layer assert(not(args.skip_connections and args.layers == 1)) # Prepare data train_stream, valid_stream = get_minibatch( dataset, mini_batch_size, mini_batch_size_valid, time_length, args.tot_num_char) # Build the model gate_values = None if rnn_type == "simple": (cost, unregularized_cost, updates, hidden_states) = build_model_vanilla(args) elif rnn_type == "clockwork": cost, unregularized_cost, updates, hidden_states = build_model_cw(args) elif rnn_type == "lstm": (cost, unregularized_cost, updates, gate_values, hidden_states) = build_model_lstm(args) elif rnn_type == "soft": (cost, unregularized_cost, updates, gate_values, hidden_states) = build_model_soft(args)
if __name__ == "__main__": args = parse_args() dataset = args.dataset mini_batch_size = args.mini_batch_size mini_batch_size_valid = args.mini_batch_size_valid time_length = args.time_length rnn_type = args.rnn_type # Make sure we don't have skip_connections with only one hidden layer assert (not (args.skip_connections and args.layers == 1)) # Prepare data train_stream, valid_stream = get_minibatch(dataset, mini_batch_size, mini_batch_size_valid, time_length, args.tot_num_char) # Build the model gate_values = None if rnn_type == "simple": (cost, unregularized_cost, updates, hidden_states) = build_model_vanilla(args) elif rnn_type == "clockwork": cost, unregularized_cost, updates, hidden_states = build_model_cw(args) elif rnn_type == "lstm": (cost, unregularized_cost, updates, gate_values, hidden_states) = build_model_lstm(args) elif rnn_type == "soft": (cost, unregularized_cost, updates, gate_values, hidden_states) = build_model_soft(args)
from rnn.datasets.dataset import get_minibatch from rnn.train import train_model from rnn.utils import parse_args from rnn.visualize import run_visualizations if __name__ == "__main__": args = parse_args() rnn_type = args.rnn_type # Make sure we don't have skip_connections with only one hidden layer assert (not (args.skip_connections and args.layers == 1)) # Prepare data train_stream, valid_stream = get_minibatch(args) # Build the model gate_values = None if rnn_type == "simple": (cost, unregularized_cost, updates, hidden_states) = build_model_vanilla(args) elif rnn_type == "clockwork": cost, unregularized_cost, updates, hidden_states = build_model_cw(args) elif rnn_type == "lstm": (cost, unregularized_cost, updates, gate_values, hidden_states) = build_model_lstm(args) elif rnn_type == "soft": (cost, unregularized_cost, updates, gate_values, hidden_states) = build_model_soft(args) elif rnn_type == "hard":
from rnn.datasets.dataset import get_minibatch from rnn.train import train_model from rnn.utils import parse_args from rnn.visualize import run_visualizations if __name__ == "__main__": args = parse_args() rnn_type = args.rnn_type # Make sure we don't have skip_connections with only one hidden layer assert(not(args.skip_connections and args.layers == 1)) # Prepare data train_stream, valid_stream = get_minibatch(args) # Build the model gate_values = None if rnn_type == "simple": (cost, unregularized_cost, updates, hidden_states) = build_model_vanilla(args) elif rnn_type == "clockwork": cost, unregularized_cost, updates, hidden_states = build_model_cw(args) elif rnn_type == "lstm": (cost, unregularized_cost, updates, gate_values, hidden_states) = build_model_lstm(args) elif rnn_type == "soft": (cost, unregularized_cost, updates, gate_values, hidden_states) = build_model_soft(args) elif rnn_type == "hard":