for k in preliminary_env.env_params.keys():
f.write('\t' + k + ' = ' + str(preliminary_env.env_params[k]) +
'\n')
device = "cpu"
env = wrap_openai_gym(
preliminary_env.App(always_render=False, verbose=False))
# env=wrap_openai_gym(minimal_preliminary.App())
if ag == 'ppo':
observation_processor, hidden_layers, action_mapper = default_network_components(
env)
agent = PPOAgent(observation_processor,
hidden_layers,
action_mapper,
save_path=agent_params['save_path'],
value_lr=agent_params['value_lr'],
policy_lr=agent_params['policy_lr']).to(device)
agent.train(env,
episodes=agent_params['episodes'],
train_freq=agent_params['train_freq'],
eval_freq=agent_params['eval_freq'],
render=agent_params['render'],
batch_size=agent_params['batch_size'],
gamma=agent_params['gamma'],
tau=agent_params['tau'],
clip=agent_params['clip'],
n_steps=agent_params['n_steps'],
entropy_coef=agent_params['entropy_coef'],
store_agent_every=agent_params['store_agent_every'])
f.write('env_params\n')
for k in gauss_env.env_params.keys():
f.write('\t' + k + ' = ' + str(gauss_env.env_params[k]) + '\n')
# device="cuda:0" if torch.cuda.is_available() else "cpu"
# # torch.cuda.set_device(0)
device = "cpu"
env = wrap_openai_gym(gauss_env.App(always_render=False, verbose=False))
observation_processor, hidden_layers, action_mapper = default_network_components(
env)
agent = PPOAgent(observation_processor,
hidden_layers,
action_mapper,
save_path=agent_params['save_path'],
value_lr=agent_params['value_lr'],
policy_lr=agent_params['policy_lr']).to(device)
if LOAD_AGENT_FROM is not None:
with open(LOAD_AGENT_FROM, 'rb') as f:
state_dict = pickle.load(f)
agent.load_state_dict(state_dict)
# agent.eval(env,eval_episodes=20,render=True, episodes=agent_params['episodes'],train_freq=agent_params['train_freq'],eval_freq=agent_params['eval_freq'], batch_size=agent_params['batch_size'],
# gamma=agent_params['gamma'],tau=agent_params['tau'],clip=agent_params['clip'],n_steps=agent_params['n_steps'],entropy_coef=agent_params['entropy_coef'])
# agent.train(env,episodes=1000,train_freq=2048,eval_freq=50,render=True, batch_size=128,gamma=.99,tau=.95,clip=.2,n_steps=32,entropy_coef=.01)
agent.train(env,
episodes=agent_params['episodes'],
from pyforce.env import wrap_openai_gym
from pyforce.nn import default_network_components
from pyforce.agents import PPOAgent
import gym
import torch
device="cuda:0" if torch.cuda.is_available() else "cpu"
env=wrap_openai_gym(gym.make("LunarLanderContinuous-v2"))
observation_processor,hidden_layers,action_mapper=default_network_components(env)
agent=PPOAgent(
observation_processor,
hidden_layers,
action_mapper,
save_path="./evals/ppo_example",
value_lr=5e-4,
policy_lr=5e-4
).to(device)
agent.train(env,episodes=1000,train_freq=2048,eval_freq=50,render=True, batch_size=128,gamma=.99,tau=.95,clip=.2,n_steps=32,entropy_coef=.01)