mirror of
https://github.com/wassname/ray.git
synced 2026-08-12 12:20:11 +08:00
22ccc43670
* Fix. * Rollback. * WIP. * WIP. * WIP. * WIP. * WIP. * WIP. * WIP. * WIP. * Fix. * Fix. * Fix. * Fix. * Fix. * WIP. * WIP. * Fix. * Test case fixes. * Test case fixes and LINT. * Test case fixes and LINT. * Rollback. * WIP. * WIP. * Test case fixes. * Fix. * Fix. * Fix. * Add regression test for DQN w/ param noise. * Fixes and LINT. * Fixes and LINT. * Fixes and LINT. * Fixes and LINT. * Fixes and LINT. * Comment * Regression test case. * WIP. * WIP. * LINT. * LINT. * WIP. * Fix. * Fix. * Fix. * LINT. * Fix (SAC does currently not support eager). * Fix. * WIP. * LINT. * Update rllib/evaluation/sampler.py Co-Authored-By: Eric Liang <ekhliang@gmail.com> * Update rllib/evaluation/sampler.py Co-Authored-By: Eric Liang <ekhliang@gmail.com> * Update rllib/utils/exploration/exploration.py Co-Authored-By: Eric Liang <ekhliang@gmail.com> * Update rllib/utils/exploration/exploration.py Co-Authored-By: Eric Liang <ekhliang@gmail.com> * WIP. * WIP. * Fix. * LINT. * LINT. * Fix and LINT. * WIP. * WIP. * WIP. * WIP. * Fix. * LINT. * Fix. * Fix and LINT. * Update rllib/utils/exploration/exploration.py * Update rllib/policy/dynamic_tf_policy.py Co-Authored-By: Eric Liang <ekhliang@gmail.com> * Update rllib/policy/dynamic_tf_policy.py Co-Authored-By: Eric Liang <ekhliang@gmail.com> * Update rllib/policy/dynamic_tf_policy.py Co-Authored-By: Eric Liang <ekhliang@gmail.com> * Fixes. * WIP. * LINT. * Fixes and LINT. * LINT and fixes. * LINT. * Move action_dist back into torch extra_action_out_fn and LINT. * Working SimpleQ learning cartpole on both torch AND tf. * Working Rainbow learning cartpole on tf. * Working Rainbow learning cartpole on tf. * WIP. * LINT. * LINT. * Update docs and add torch to APEX test. * LINT. * Fix. * LINT. * Fix. * Fix. * Fix and docstrings. * Fix broken RLlib tests in master. * Split BAZEL learning tests into cartpole and pendulum (reached the 60min barrier). * Fix error_outputs option in BAZEL for RLlib regression tests. * Fix. * Tune param-noise tests. * LINT. * Fix. * Fix. * test * test * test * Fix. * Fix. * WIP. * WIP. * WIP. * WIP. * LINT. * WIP. Co-authored-by: Eric Liang <ekhliang@gmail.com>
65 lines
2.2 KiB
Python
65 lines
2.2 KiB
Python
"""A more stable successor to TD3.
|
|
|
|
By default, this uses a near-identical configuration to that reported in the
|
|
TD3 paper.
|
|
"""
|
|
from ray.rllib.agents.ddpg.ddpg import DDPGTrainer, \
|
|
DEFAULT_CONFIG as DDPG_CONFIG
|
|
|
|
TD3_DEFAULT_CONFIG = DDPGTrainer.merge_trainer_configs(
|
|
DDPG_CONFIG,
|
|
{
|
|
# largest changes: twin Q functions, delayed policy updates, and target
|
|
# smoothing
|
|
"twin_q": True,
|
|
"policy_delay": 2,
|
|
"smooth_target_policy": True,
|
|
"target_noise": 0.2,
|
|
"target_noise_clip": 0.5,
|
|
"exploration_config": {
|
|
# TD3 uses simple Gaussian noise on top of deterministic NN-output
|
|
# actions (after a possible pure random phase of n timesteps).
|
|
"type": "GaussianNoise",
|
|
# For how many timesteps should we return completely random
|
|
# actions, before we start adding (scaled) noise?
|
|
"random_timesteps": 10000,
|
|
# Gaussian stddev of action noise for exploration.
|
|
"stddev": 0.1,
|
|
# Scaling settings by which the Gaussian noise is scaled before
|
|
# being added to the actions. NOTE: The scale timesteps start only
|
|
# after(!) any random steps have been finished.
|
|
# By default, do not anneal over time (fixed 1.0).
|
|
"initial_scale": 1.0,
|
|
"final_scale": 1.0,
|
|
"scale_timesteps": 1
|
|
},
|
|
|
|
# other changes & things we want to keep fixed:
|
|
# larger actor learning rate, no l2 regularisation, no Huber loss, etc.
|
|
"learning_starts": 10000,
|
|
"actor_hiddens": [400, 300],
|
|
"critic_hiddens": [400, 300],
|
|
"n_step": 1,
|
|
"gamma": 0.99,
|
|
"actor_lr": 1e-3,
|
|
"critic_lr": 1e-3,
|
|
"l2_reg": 0.0,
|
|
"tau": 5e-3,
|
|
"train_batch_size": 100,
|
|
"use_huber": False,
|
|
"target_network_update_freq": 0,
|
|
"num_workers": 0,
|
|
"num_gpus_per_worker": 0,
|
|
"worker_side_prioritization": False,
|
|
"buffer_size": 1000000,
|
|
"prioritized_replay": False,
|
|
"clip_rewards": False,
|
|
"use_state_preprocessor": False,
|
|
})
|
|
|
|
TD3Trainer = DDPGTrainer.with_updates(
|
|
name="TD3",
|
|
default_config=TD3_DEFAULT_CONFIG,
|
|
get_policy_class=None,
|
|
)
|