2019-11-13 12:22:55 -08:00
|
|
|
# This file runs on a single g3.16xl or p3.16xl node. It is suggested
|
|
|
|
|
# to run these in a DLAMI / tensorflow_p36 env. Note that RL runs are
|
|
|
|
|
# inherently high variance, so you'll have to check to see if the
|
|
|
|
|
# rewards reached seem reasonably in line with previous results.
|
|
|
|
|
#
|
2020-01-18 23:30:09 -08:00
|
|
|
# You can find the reference results here:
|
|
|
|
|
# https://github.com/ray-project/ray/tree/master/doc/dev/release_logs
|
2019-11-13 12:22:55 -08:00
|
|
|
atari-impala:
|
|
|
|
|
env: BreakoutNoFrameskip-v4
|
|
|
|
|
run: IMPALA
|
|
|
|
|
num_samples: 4
|
|
|
|
|
stop:
|
|
|
|
|
time_total_s: 3600
|
|
|
|
|
config:
|
|
|
|
|
sample_batch_size: 50
|
|
|
|
|
train_batch_size: 500
|
|
|
|
|
num_workers: 10
|
|
|
|
|
num_envs_per_worker: 5
|
|
|
|
|
clip_rewards: True
|
|
|
|
|
lr_schedule: [
|
|
|
|
|
[0, 0.0005],
|
|
|
|
|
[20000000, 0.000000000001],
|
|
|
|
|
]
|
|
|
|
|
num_gpus: 1
|
|
|
|
|
atari-ppo:
|
|
|
|
|
env: BreakoutNoFrameskip-v4
|
|
|
|
|
run: PPO
|
|
|
|
|
num_samples: 4
|
|
|
|
|
stop:
|
|
|
|
|
time_total_s: 3600
|
|
|
|
|
config:
|
|
|
|
|
lambda: 0.95
|
|
|
|
|
kl_coeff: 0.5
|
|
|
|
|
clip_rewards: True
|
|
|
|
|
clip_param: 0.1
|
|
|
|
|
vf_clip_param: 10.0
|
|
|
|
|
entropy_coeff: 0.01
|
|
|
|
|
train_batch_size: 5000
|
|
|
|
|
sample_batch_size: 100
|
|
|
|
|
sgd_minibatch_size: 500
|
|
|
|
|
num_sgd_iter: 10
|
|
|
|
|
num_workers: 10
|
|
|
|
|
num_envs_per_worker: 5
|
|
|
|
|
batch_mode: truncate_episodes
|
|
|
|
|
observation_filter: NoFilter
|
|
|
|
|
vf_share_layers: true
|
|
|
|
|
num_gpus: 1
|
|
|
|
|
apex:
|
|
|
|
|
env: BreakoutNoFrameskip-v4
|
|
|
|
|
run: APEX
|
|
|
|
|
num_samples: 4
|
|
|
|
|
stop:
|
|
|
|
|
time_total_s: 3600
|
|
|
|
|
config:
|
|
|
|
|
double_q: false
|
|
|
|
|
dueling: false
|
|
|
|
|
num_atoms: 1
|
|
|
|
|
noisy: false
|
|
|
|
|
n_step: 3
|
|
|
|
|
lr: .0001
|
|
|
|
|
adam_epsilon: .00015
|
|
|
|
|
hiddens: [512]
|
|
|
|
|
buffer_size: 1000000
|
|
|
|
|
schedule_max_timesteps: 2000000
|
|
|
|
|
exploration_final_eps: 0.01
|
|
|
|
|
exploration_fraction: .1
|
|
|
|
|
prioritized_replay_alpha: 0.5
|
|
|
|
|
beta_annealing_fraction: 1.0
|
|
|
|
|
final_prioritized_replay_beta: 1.0
|
|
|
|
|
num_gpus: 1
|
|
|
|
|
num_workers: 8
|
|
|
|
|
num_envs_per_worker: 8
|
|
|
|
|
sample_batch_size: 20
|
|
|
|
|
train_batch_size: 512
|
|
|
|
|
target_network_update_freq: 50000
|
|
|
|
|
timesteps_per_iteration: 25000
|
|
|
|
|
atari-a2c:
|
|
|
|
|
env: BreakoutNoFrameskip-v4
|
|
|
|
|
run: A2C
|
|
|
|
|
num_samples: 4
|
|
|
|
|
stop:
|
|
|
|
|
time_total_s: 3600
|
|
|
|
|
config:
|
|
|
|
|
sample_batch_size: 20
|
|
|
|
|
clip_rewards: True
|
|
|
|
|
num_workers: 5
|
|
|
|
|
num_envs_per_worker: 5
|
|
|
|
|
num_gpus: 1
|
|
|
|
|
lr_schedule: [
|
|
|
|
|
[0, 0.0007],
|
|
|
|
|
[20000000, 0.000000000001],
|
|
|
|
|
]
|
|
|
|
|
atari-basic-dqn:
|
|
|
|
|
env: BreakoutNoFrameskip-v4
|
|
|
|
|
run: DQN
|
|
|
|
|
num_samples: 4
|
|
|
|
|
stop:
|
|
|
|
|
time_total_s: 3600
|
|
|
|
|
config:
|
|
|
|
|
double_q: false
|
|
|
|
|
dueling: false
|
|
|
|
|
num_atoms: 1
|
|
|
|
|
noisy: false
|
|
|
|
|
prioritized_replay: false
|
|
|
|
|
n_step: 1
|
|
|
|
|
target_network_update_freq: 8000
|
|
|
|
|
lr: .0000625
|
|
|
|
|
adam_epsilon: .00015
|
|
|
|
|
hiddens: [512]
|
|
|
|
|
learning_starts: 20000
|
|
|
|
|
buffer_size: 1000000
|
|
|
|
|
sample_batch_size: 4
|
|
|
|
|
train_batch_size: 32
|
|
|
|
|
schedule_max_timesteps: 2000000
|
|
|
|
|
exploration_final_eps: 0.01
|
|
|
|
|
exploration_fraction: .1
|
|
|
|
|
prioritized_replay_alpha: 0.5
|
|
|
|
|
beta_annealing_fraction: 1.0
|
|
|
|
|
final_prioritized_replay_beta: 1.0
|
|
|
|
|
num_gpus: 0.2
|
|
|
|
|
timesteps_per_iteration: 10000
|
