wassname/DeepRL
1
0
Fork 0
mirror of https://github.com/wassname/DeepRL.git synced 2026-09-09 11:13:47 +08:00
Code Issues Packages Projects Releases Wiki Activity

Log wall time

Browse Source
This commit is contained in:
Shangtong Zhang
2017-10-14 16:38:51 -06:00
parent c04b2a395f
commit 3ed6f2a9db
4 changed files with 70 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
agent/async_agent.py
+7 -4
View File
@@ -29,6 +29,8 @@ def train(id, config, learning_network, extra):
def evaluate(config, task, learning_network, extra):
test_rewards = []
test_points = []
test_wall_times = []
initial_time = time.time()
worker = config.worker(config, learning_network, extra)
# config.logger = Logger('./evaluation_log', gym.logger)
while True:
@@ -45,10 +47,11 @@ def evaluate(config, task, learning_network, extra):
(steps, np.mean(rewards), np.std(rewards) / np.sqrt(config.test_repetitions)))
test_rewards.append(np.mean(rewards))
test_points.append(steps)
test_wall_times.append(time.time() - initial_time)
with open('data/%s-%s-statistics-%s.bin' % (
config.tag, config.worker.__name__, task.name), 'wb') as f:
pickle.dump([test_points, test_rewards], f)
if np.mean(rewards) > task.success_threshold:
pickle.dump([test_rewards, test_points, test_wall_times], f)
if np.mean(rewards) > task.success_threshold or (config.max_steps and steps >= config.max_steps):
config.stop_signal.value = True
break
@@ -80,8 +83,8 @@ class AsyncAgent:
extra = None
args = [(i, config, learning_network, extra) for i in range(config.num_workers)]
args.append((config, task, learning_network, extra))
procs = [mp.Process(target=train, args=args[i]) for i in range(config.num_workers)]
procs.append(mp.Process(target=evaluate, args=args[-1]))
procs = [mp.Process(target=evaluate, args=args[-1])]
procs.extend([mp.Process(target=train, args=args[i]) for i in range(config.num_workers)])
for p in procs: p.start()
while True:
time.sleep(1)