add drq example in sim with lift cube env

examples/async_drq_sim/async_drq_sim.py

@@ -0,0 +1,355 @@
+#!/usr/bin/env python3
+
+# NOTE: this requires jaxrl_m to be installed:
+#       https://github.com/rail-berkeley/jaxrl_minimal
+
+import time
+from functools import partial
+import jax
+import jax.numpy as jnp
+import numpy as np
+import tqdm
+from absl import app, flags
+
+import gymnasium as gym
+from gymnasium.wrappers.record_episode_statistics import RecordEpisodeStatistics
+
+from jaxrl_m.agents.continuous.drq import DrQAgent
+from jaxrl_m.common.evaluation import evaluate
+from jaxrl_m.utils.timer_utils import Timer
+from jaxrl_m.envs.wrappers.chunking import ChunkingWrapper
+
+from edgeml.trainer import TrainerServer, TrainerClient, TrainerTunnel
+from edgeml.data.data_store import QueuedDataStore
+
+from serl_launcher.utils.jaxrl_m_common import (
+    MemoryEfficientReplayBufferDataStore,
+    make_drq_agent,
+    make_trainer_config,
+    make_wandb_logger,
+)
+from serl_launcher.wrappers.serl_obs_wrappers import SERLObsWrapper
+
+FLAGS = flags.FLAGS
+
+flags.DEFINE_string("env", "HalfCheetah-v4", "Name of environment.")
+flags.DEFINE_string("agent", "drq", "Name of agent.")
+flags.DEFINE_string("exp_name", None, "Name of the experiment for wandb logging.")
+flags.DEFINE_integer("max_traj_length", 1000, "Maximum length of trajectory.")
+flags.DEFINE_integer("seed", 42, "Random seed.")
+flags.DEFINE_bool("save_model", False, "Whether to save model.")
+flags.DEFINE_integer("batch_size", 256, "Batch size.")
+flags.DEFINE_integer("utd_ratio", 4, "UTD ratio.")
+
+flags.DEFINE_integer("max_steps", 1000000, "Maximum number of training steps.")
+flags.DEFINE_integer("replay_buffer_capacity", 200000, "Replay buffer capacity.")
+
+flags.DEFINE_integer("random_steps", 300, "Sample random actions for this many steps.")
+flags.DEFINE_integer("training_starts", 300, "Training starts after this step.")
+flags.DEFINE_integer("steps_per_update", 10, "Number of steps per update the server.")
+
+flags.DEFINE_integer("log_period", 10, "Logging period.")
+flags.DEFINE_integer("eval_period", 2000, "Evaluation period.")
+flags.DEFINE_integer("eval_n_trajs", 5, "Number of trajectories for evaluation.")
+
+# flag to indicate if this is a leaner or a actor
+flags.DEFINE_boolean("learner", False, "Is this a learner or a trainer.")
+flags.DEFINE_boolean("actor", False, "Is this a learner or a trainer.")
+flags.DEFINE_boolean("render", False, "Render the environment.")
+flags.DEFINE_string("ip", "localhost", "IP address of the learner.")
+
+flags.DEFINE_boolean(
+    "debug", False, "Debug mode."
+)  # debug mode will disable wandb logging
+
+
+def print_green(x):
+    return print("\033[92m {}\033[00m".format(x))
+
+
+##############################################################################
+
+
+def actor(agent: DrQAgent, data_store, env, sampling_rng, tunnel=None):
+    """
+    This is the actor loop, which runs when "--actor" is set to True.
+    NOTE: tunnel is used the transport layer for multi-threading
+    """
+    if tunnel:
+        client = tunnel
+    else:
+        client = TrainerClient(
+            "actor_env",
+            FLAGS.ip,
+            make_trainer_config(),
+            data_store,
+            wait_for_server=True,
+        )
+
+    # Function to update the agent with new params
+    def update_params(params):
+        nonlocal agent
+        agent = agent.replace(state=agent.state.replace(params=params))
+
+    client.recv_network_callback(update_params)
+
+    eval_env = gym.make(FLAGS.env)
+    if FLAGS.env == "PandaPickCubeVision-v0":
+        eval_env = SERLObsWrapper(eval_env)
+        eval_env = ChunkingWrapper(eval_env, obs_horizon=1, act_exec_horizon=None)
+    eval_env = RecordEpisodeStatistics(eval_env)
+
+    obs, _ = env.reset()
+    done = False
+
+    # training loop
+    timer = Timer()
+    running_return = 0.0
+
+    for step in tqdm.tqdm(range(FLAGS.max_steps), dynamic_ncols=True):
+        timer.tick("total")
+
+        with timer.context("sample_actions"):
+            if step < FLAGS.random_steps:
+                actions = env.action_space.sample()
+            else:
+                sampling_rng, key = jax.random.split(sampling_rng)
+                actions = agent.sample_actions(
+                    observations=jax.device_put(obs),
+                    seed=key,
+                    deterministic=False,
+                )
+                actions = np.asarray(jax.device_get(actions))
+
+        # Step environment
+        with timer.context("step_env"):
+
+            next_obs, reward, done, truncated, info = env.step(actions)
+            reward = np.asarray(reward, dtype=np.float32)
+            info = np.asarray(info)
+            running_return += reward
+            transition = dict(
+                observations=obs,
+                actions=actions,
+                next_observations=next_obs,
+                rewards=reward,
+                masks=1.0 - done,
+                dones=done,
+            )
+            data_store.insert(transition)
+
+            obs = next_obs
+            if done or truncated:
+                running_return = 0.0
+                obs, _ = env.reset()
+
+        if step % FLAGS.steps_per_update == 0:
+            client.update()
+
+        if step % FLAGS.eval_period == 0:
+            with timer.context("eval"):
+                evaluate_info = evaluate(
+                    policy_fn=partial(agent.sample_actions, argmax=True),
+                    env=eval_env,
+                    num_episodes=FLAGS.eval_n_trajs,
+                )
+            stats = {"eval": evaluate_info}
+            client.request("send-stats", stats)
+
+        timer.tock("total")
+
+        if step % FLAGS.log_period == 0:
+            stats = {"timer": timer.get_average_times()}
+            client.request("send-stats", stats)
+
+
+##############################################################################
+
+
+def learner(
+    rng, agent: DrQAgent, replay_buffer, replay_iterator, wandb_logger=None, tunnel=None
+):
+    """
+    The learner loop, which runs when "--learner" is set to True.
+    NOTE: tunnel is used the transport layer for multi-threading
+    """
+    # To track the step in the training loop
+    update_steps = 0
+
+    def stats_callback(type: str, payload: dict) -> dict:
+        """Callback for when server receives stats request."""
+        assert type == "send-stats", f"Invalid request type: {type}"
+        if wandb_logger is not None:
+            wandb_logger.log(payload, step=update_steps)
+        return {}  # not expecting a response
+
+    # Create server
+    if tunnel:
+        tunnel.register_request_callback(stats_callback)
+        server = tunnel
+    else:
+        server = TrainerServer(make_trainer_config(), request_callback=stats_callback)
+        server.register_data_store("actor_env", replay_buffer)
+        server.start(threaded=True)
+
+    # Loop to wait until replay_buffer is filled
+    pbar = tqdm.tqdm(
+        total=FLAGS.training_starts,
+        initial=len(replay_buffer),
+        desc="Filling up replay buffer",
+        position=0,
+        leave=True,
+    )
+    while len(replay_buffer) < FLAGS.training_starts:
+        pbar.update(len(replay_buffer) - pbar.n)  # Update progress bar
+        time.sleep(1)
+    pbar.update(len(replay_buffer) - pbar.n)  # Update progress bar
+    pbar.close()
+
+    # send the initial network to the actor
+    server.publish_network(agent.state.params)
+    print_green("sent initial network to actor")
+
+    # wait till the replay buffer is filled with enough data
+    timer = Timer()
+    for step in tqdm.tqdm(range(FLAGS.max_steps), dynamic_ncols=True, desc="learner"):
+        # run n-1 critic updates and 1 critic + actor update.
+        # This makes training on GPU faster by reducing the large batch transfer time from CPU to GPU
+        for critic_step in range(FLAGS.utd_ratio - 1):
+            with timer.context("sample_replay_buffer"):
+                batch = next(replay_iterator)
+
+            with timer.context("train_critics"):
+                agent, critics_info = agent.update_critics(
+                    batch,
+                )
+                agent = jax.block_until_ready(agent)
+
+        with timer.context("train"):
+            batch = next(replay_iterator)
+            agent, update_info = agent.update_high_utd(batch, utd_ratio=1)
+            agent = jax.block_until_ready(agent)
+
+        # publish the updated network
+        if step > 0 and step % (FLAGS.steps_per_update) == 0:
+            server.publish_network(agent.state.params)
+
+        if update_steps % FLAGS.log_period == 0 and wandb_logger:
+            wandb_logger.log(update_info, step=update_steps)
+            wandb_logger.log({"timer": timer.get_average_times()}, step=update_steps)
+
+        update_steps += 1
+
+
+##############################################################################
+
+
+def main(_):
+    devices = jax.local_devices()
+    num_devices = len(devices)
+    sharding = jax.sharding.PositionalSharding(devices)
+    assert FLAGS.batch_size % num_devices == 0
+
+    # seed
+    rng = jax.random.PRNGKey(FLAGS.seed)
+
+    # create env and load dataset
+    if FLAGS.render:
+        env = gym.make(FLAGS.env, render_mode="human")
+    else:
+        env = gym.make(FLAGS.env)
+
+    if FLAGS.env == "PandaPickCube-v0":
+        env = gym.wrappers.FlattenObservation(env)
+    if FLAGS.env == "PandaPickCubeVision-v0":
+        env = SERLObsWrapper(env)
+        env = ChunkingWrapper(env, obs_horizon=1, act_exec_horizon=None)
+
+    image_keys = [key for key in env.observation_space.keys() if key != "state"]
+
+    rng, sampling_rng = jax.random.split(rng)
+    agent: DrQAgent = make_drq_agent(
+        seed=FLAGS.seed,
+        sample_obs=env.observation_space.sample(),
+        sample_action=env.action_space.sample(),
+        image_keys=image_keys,
+        encoder_type="mobilenet",
+    )
+
+    # replicate agent across devices
+    # need the jnp.array to avoid a bug where device_put doesn't recognize primitives
+    agent: DrQAgent = jax.device_put(
+        jax.tree_map(jnp.array, agent), sharding.replicate()
+    )
+
+    def create_replay_buffer_and_wandb_logger():
+        replay_buffer = MemoryEfficientReplayBufferDataStore(
+            env.observation_space,
+            env.action_space,
+            capacity=FLAGS.replay_buffer_capacity,
+            image_keys=image_keys,
+        )
+        # set up wandb and logging
+        wandb_logger = make_wandb_logger(
+            project="serl_dev",
+            description=FLAGS.exp_name or FLAGS.env,
+            debug=FLAGS.debug,
+        )
+        return replay_buffer, wandb_logger
+
+    if FLAGS.learner:
+        sampling_rng = jax.device_put(sampling_rng, device=sharding.replicate())
+        replay_buffer, wandb_logger = create_replay_buffer_and_wandb_logger()
+        replay_iterator = replay_buffer.get_iterator(
+            sample_args={
+                "batch_size": FLAGS.batch_size,
+                "pack_obs_and_next_obs": True,
+            },
+            device=sharding.replicate(),
+        )
+
+        # learner loop
+        print_green("starting learner loop")
+        learner(
+            sampling_rng,
+            agent,
+            replay_buffer,
+            replay_iterator=replay_iterator,
+            wandb_logger=wandb_logger,
+            tunnel=None,
+        )
+
+    elif FLAGS.actor:
+        sampling_rng = jax.device_put(sampling_rng, sharding.replicate())
+        data_store = QueuedDataStore(50000)  # the queue size on the actor
+
+        # actor loop
+        print_green("starting actor loop")
+        actor(agent, data_store, env, sampling_rng, tunnel=None)
+
+    else:
+        print_green("starting actor and learner loop with multi-threading")
+
+        # In this example, the tunnel acts as the transport layer for the
+        # trainerServer and trainerClient. Also, both actor and learner shares
+        # the same replay buffer.
+        replay_buffer, wandb_logger = create_replay_buffer_and_wandb_logger()
+
+        tunnel = TrainerTunnel()
+        sampling_rng = jax.device_put(sampling_rng, sharding.replicate())
+
+        import threading
+
+        # Start learner thread
+        learner_thread = threading.Thread(
+            target=learner, args=(agent, replay_buffer, wandb_logger, tunnel)
+        )
+        learner_thread.start()
+
+        # Start actor in main process
+        actor(agent, replay_buffer, env, sampling_rng, tunnel=tunnel)
+        learner_thread.join()
+
+
+if __name__ == "__main__":
+    app.run(main)

examples/async_drq_sim/run_actor.sh

@@ -0,0 +1,14 @@
+export XLA_PYTHON_CLIENT_PREALLOCATE=false && \
+export XLA_PYTHON_CLIENT_MEM_FRACTION=.1 && \
+python async_drq_sim.py \
+    --actor \
+    --render \
+    --env PandaPickCubeVision-v0 \
+    --exp_name=serl_dev_drq_sim_test \
+    --seed 0 \
+    --random_steps 1000 \
+    --training_starts 1000 \
+    --utd_ratio 4 \
+    --batch_size 256 \
+    --eval_period 2000 \
+    --debug

examples/async_drq_sim/run_learner.sh

@@ -0,0 +1,13 @@
+export XLA_PYTHON_CLIENT_PREALLOCATE=false && \
+export XLA_PYTHON_CLIENT_MEM_FRACTION=.1 && \
+python async_drq_sim.py \
+    --learner \
+    --env PandaPickCubeVision-v0 \
+    --exp_name=serl_dev_drq_sim_test_mobilenet_s_stopgrad \
+    --seed 0 \
+    --random_steps 1000 \
+    --training_starts 1000 \
+    --utd_ratio 4 \
+    --batch_size 256 \
+    --eval_period 2000 \
+    --debug