from agent import DDPGAgent, DDPGArgs
from trainer import DistributedTrainer
from numeric_env import MultiEnv
import torch
torch.set_num_threads(1)
env = MultiEnv(2, 2)
args = DDPGArgs(state_dim=env.STATE_DIM, action_dim=2)
agents = [DDPGAgent(args) for _ in range(2)]
trainer = DistributedTrainer(agents,
env,
parameter_share=False,
log_dir='../logs/ddpg_d')
trainer.train(1000000)
from agent import *
from trainer import DistributedTrainer
from numeric_env import MultiEnv
import torch
torch.set_num_threads(1)
env = MultiEnv(2, 2, car_policy='wjf')
args = DDPGArgs(state_dim=env.STATE_DIM,
action_dim=2,
action_scale=1.0,
)
agents = [DDPGAgent(args) for _ in range(2)]
normalizer_path = '../models/debug/norm.pkl'
save_pathlists = [['../models/debug/actor.pkl', '../models/debug/critic.pkl'] for _ in range(2)]
trainer = DistributedTrainer(agents,
env,
parameter_share=False,
log_dir='../logs/debug',
save_pathlists=None,
load_pathlists=None,
normalizer_path=None
)
trainer.train(-1)
# torch.set_num_threads(1)
#
#
# env = MultiEnv(2, 2, car_policy='wjf')