amago/examples/14_d4rl.py at main · UT-Austin-RPL/amago · GitHub

Repository navigation

History

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

from argparse import ArgumentParser

import random

import torch

import gym as og_gym

import d4rl

import gymnasium as gym

import wandb

import numpy as np

import amago

from amago.envs import AMAGOEnv

from amago import cli_utils

from amago.loading import RLData, RLDataset, DiskTrajDataset, MixtureOfDatasets

from amago.nets.policy_dists import TanhGaussian, GMM, Beta

from amago.nets.actor_critic import ResidualActor, Actor

from amago.agent import binary_filter, exp_filter

def add_cli(parser):

parser.add_argument(

"--env", type=str, required=True, help="Environment/Dataset name"

)

parser.add_argument(

"--max_seq_len", type=int, default=32, help="Policy sequence length."

)

parser.add_argument(

"--policy_dist",

type=str,

default="Beta",

help="Policy distribution type",

choices=["TanhGaussian", "GMM", "Beta"],

)

parser.add_argument(

"--actor_type",

type=str,

default="Actor",

help="Actor head type",

choices=["ResidualActor", "Actor"],

)

parser.add_argument(

"--online_after_epoch",

type=int,

default=float("inf"),

help="Number of epochs after which to start collecting online data",

)

parser.add_argument(

"--eval_timesteps",

type=int,

default=1000,

help="Number of timesteps to evaluate for each actor. Will be overridden if the environment has a known time limit.",

)

return parser

class D4RLDataset(RLDataset):

def __init__(self, d4rl_dset: dict[str, np.ndarray]):

super().__init__()

self.d4rl_dset = d4rl_dset

self.episode_ends = np.where(d4rl_dset["terminals"] | d4rl_dset["timeouts"])[0]

self.ep_lens = self.episode_ends[1:] - self.episode_ends[:-1]

self.max_ep_len = self.ep_lens.max()

def get_description(self) -> str:

return "D4RL"

@property

def save_new_trajs_to(self):

# disables saving new amago trajectories to disk

return None

def sample_random_trajectory(self):

episode_idx = random.randrange(0, len(self.episode_ends) - 1)

return self._sample_trajectory(episode_idx)

def _sample_trajectory(self, episode_idx: int):

# pick a random episode

s = self.episode_ends[episode_idx] + 1

e = self.episode_ends[episode_idx + 1] + 1

traj_len = e - s

obs_np = self.d4rl_dset["observations"][s : e + 1]

actions_np = self.d4rl_dset["actions"][s:e]

rewards_np = self.d4rl_dset["rewards"][s:e]

terminals_np = self.d4rl_dset["terminals"][s:e]

timeouts_np = self.d4rl_dset["timeouts"][s:e]

# convert to torch, adding time_idxs

obs = {"observation": torch.from_numpy(obs_np)}

actions = torch.from_numpy(actions_np).float()

rewards = torch.from_numpy(rewards_np).float().unsqueeze(-1)

time_idxs = torch.arange(traj_len).unsqueeze(-1).long()

dones = torch.from_numpy(terminals_np).bool().unsqueeze(-1)

return RLData(

obs=obs,

actions=actions,

rews=rewards,

dones=dones,

time_idxs=time_idxs,

)

from amago.envs.env_utils import space_convert

from amago.envs.amago_env import AMAGO_ENV_LOG_PREFIX

class D4RLGymEnv(gym.Env):

"""

Light wrapper that logs the D4RL normalized return and handles

the gym/gymnasium conversion while we're at it.

"""

def __init__(self, env_name: str):

# hack fix seeding for parallel envs

np.random.seed(random.randrange(1e6))

self.env_name = env_name

self.env = og_gym.make(env_name)

self.action_space = space_convert(self.env.action_space)

self.observation_space = space_convert(self.env.observation_space)

if isinstance(self.env, og_gym.wrappers.TimeLimit):

# this time limit is apparently not consistent with the datasets

self.time_limit = self.env._max_episode_steps

else:

self.time_limit = None

self.max_return = d4rl.infos.REF_MAX_SCORE[self.env_name]

self.min_return = d4rl.infos.REF_MIN_SCORE[self.env_name]

@property

def dset(self):

return self.env.get_dataset()

def reset(self, *args, **kwargs):

self.episode_return = 0

return self.env.reset(), {}

def step(self, action):

s, r, d, i = self.env.step(action)

truncated = i.get("TimeLimit.truncated", False)

terminated = d and not truncated

self.episode_return += r

if terminated or truncated:

i[f"{AMAGO_ENV_LOG_PREFIX} D4RL Normalized Return"] = (

d4rl.get_normalized_score(self.env_name, self.episode_return)

)

return s, r, terminated, truncated, i

if __name__ == "__main__":

parser = ArgumentParser()

cli_utils.add_common_cli(parser)

add_cli(parser)

args = parser.parse_args()

# ues env to set some args

env_name = args.env

example_env = D4RLGymEnv(args.env)

assert isinstance(

example_env.action_space, gym.spaces.Box

), "Only supports continuous action spaces"

args.eval_timesteps = example_env.time_limit + 1

args.timesteps_per_epoch = example_env.time_limit

# setup environment

make_train_env = lambda: AMAGOEnv(

D4RLGymEnv(args.env),

env_name=env_name,

batched_envs=1,

)

# agent architecture: drop everything down to standard small sizes

config = {

"amago.nets.actor_critic.NCritics.d_hidden": 128,

"amago.nets.actor_critic.NCriticsTwoHot.d_hidden": 128,

"amago.nets.actor_critic.NCriticsTwoHot.output_bins": 64,

"amago.nets.actor_critic.Actor.d_hidden": 128,

"amago.nets.actor_critic.Actor.continuous_dist_type": eval(args.policy_dist),

"amago.nets.actor_critic.ResidualActor.feature_dim": 128,

"amago.nets.actor_critic.ResidualActor.residual_ff_dim": 256,

"amago.nets.actor_critic.ResidualActor.residual_blocks": 2,

"amago.nets.actor_critic.ResidualActor.continuous_dist_type": eval(

args.policy_dist

}

tstep_encoder_type = cli_utils.switch_tstep_encoder(

config,

arch="ff",

d_hidden=128,

d_output=128,

n_layers=1,

)

exploration_wrapper_type = cli_utils.switch_exploration(

config,

strategy="egreedy",

eps_start=0.05,

eps_end=0.01,

steps_anneal=15_000,

)

traj_encoder_type = cli_utils.switch_traj_encoder(

config,

arch=args.traj_encoder,

memory_size=args.memory_size,

layers=args.memory_layers,

)

agent_type = cli_utils.switch_agent(

config,

args.agent_type,

online_coeff=0.0,

offline_coeff=1.0,

gamma=0.997,

reward_multiplier=100.0 if example_env.max_return <= 10.0 else 1,

num_actions_for_value_in_critic_loss=3,

num_actions_for_value_in_actor_loss=5,

num_critics=5,

actor_type=eval(args.actor_type),

fbc_filter_func=exp_filter,

)

cli_utils.use_config(config, args.configs)

group_name = f"{args.run_name}_{env_name}"

for trial in range(args.trials):

run_name = group_name + f"_trial_{trial}"

# create dataset

d4rl_dataset = D4RLDataset(d4rl_dset=example_env.dset)

online_dset = DiskTrajDataset(

dset_root=args.buffer_dir,

dset_name=run_name,

dset_min_size=250,

dset_max_size=args.dset_max_size,

)

combined_dset = MixtureOfDatasets(

datasets=[d4rl_dataset, online_dset],

# skew sampling towards the demos 80/20

sampling_weights=[0.8, 0.2],

# gradually increase the weight of the online dset

# over the first 100 epochs *after online collection starts*

smooth_sudden_starts=50,

)

experiment = cli_utils.create_experiment_from_cli(

args,

make_train_env=make_train_env,

make_val_env=make_train_env,

max_seq_len=args.max_seq_len,

run_name=run_name,

tstep_encoder_type=tstep_encoder_type,

traj_encoder_type=traj_encoder_type,

agent_type=agent_type,

group_name=group_name,

val_timesteps_per_epoch=args.eval_timesteps,

learning_rate=1e-4,

dataset=combined_dset,

padded_sampling="right",

start_collecting_at_epoch=args.online_after_epoch,

stagger_traj_file_lengths=False,

traj_save_len=args.eval_timesteps + 1,

sample_actions_train=False,

sample_actions_val=False,

exploration_wrapper_type=exploration_wrapper_type,

)

# save a copy of this script at the time of the run

experiment = cli_utils.switch_async_mode(experiment, args.mode)

experiment.start()

if args.ckpt is not None:

experiment.load_checkpoint(args.ckpt)

experiment.learn()

experiment.evaluate_test(make_train_env, timesteps=10_000, render=False)

wandb.finish()

amago/examples/14_d4rl.py at main · UT-Austin-RPL/amago · GitHub

Repository navigation

Expand file tree

Breadcrumbs

14_d4rl.py

Latest commit

History

Breadcrumbs

14_d4rl.py

File metadata and controls