wassname/ray
1
0
Fork 0
mirror of https://github.com/wassname/ray.git synced 2026-08-13 12:30:18 +08:00
Code Issues Packages Projects Releases Wiki Activity

[RLlib] Issue 8319 DDPG (MA or num_envs_per_worker > 1) broken. (#8324)

Browse Source
This commit is contained in:
Sven Mika
2020-05-08 08:26:32 +02:00
committed by GitHub
parent 5f278c6411
commit d7eaacb5fe
5 changed files with 100 additions and 55 deletions
Download Patch File Download Diff File Expand all files Collapse all files
rllib/tests/test_multi_agent_pendulum.py
+33 -28
View File
@@ -2,9 +2,10 @@
import unittest
import ray
from ray.rllib.examples.env.multi_agent import MultiAgentPendulum
from ray.tune import run_experiments
from ray.tune.registry import register_env
from ray.rllib.examples.env.multi_agent import MultiAgentPendulum
from ray.rllib.utils.test_utils import framework_iterator
class TestMultiAgentPendulum(unittest.TestCase):
@@ -17,34 +18,38 @@ class TestMultiAgentPendulum(unittest.TestCase):
def test_multi_agent_pendulum(self):
register_env("multi_agent_pendulum",
lambda _: MultiAgentPendulum({"num_agents": 1}))
trials = run_experiments({
"test": {
"run": "PPO",
"env": "multi_agent_pendulum",
"stop": {
"timesteps_total": 500000,
"episode_reward_mean": -200,
},
"config": {
"train_batch_size": 2048,
"vf_clip_param": 10.0,
"num_workers": 0,
"num_envs_per_worker": 10,
"lambda": 0.1,
"gamma": 0.95,
"lr": 0.0003,
"sgd_minibatch_size": 64,
"num_sgd_iter": 10,
"model": {
"fcnet_hiddens": [128, 128],
# Test for both torch and tf.
for fw in framework_iterator(frameworks=["torch", "tf"]):
trials = run_experiments({
"test": {
"run": "PPO",
"env": "multi_agent_pendulum",
"stop": {
"timesteps_total": 500000,
"episode_reward_mean": -300.0,
},
"batch_mode": "complete_episodes",
},
}
})
if trials[0].last_result["episode_reward_mean"] < -200:
raise ValueError("Did not get to -200 reward",
trials[0].last_result)
"config": {
"train_batch_size": 2048,
"vf_clip_param": 10.0,
"num_workers": 0,
"num_envs_per_worker": 10,
"lambda": 0.1,
"gamma": 0.95,
"lr": 0.0003,
"sgd_minibatch_size": 64,
"num_sgd_iter": 10,
"model": {
"fcnet_hiddens": [128, 128],
},
"batch_mode": "complete_episodes",
"use_pytorch": fw == "torch",
},
}
})
if trials[0].last_result["episode_reward_mean"] < -300.0:
raise ValueError("Did not get to -200 reward",
trials[0].last_result)
if __name__ == "__main__":