def training_pipeline(workers, config):
rollouts = ParallelRollouts(workers, mode="bulk_sync")
if config["microbatch_size"]:
num_microbatches = math.ceil(config["train_batch_size"] /
config["microbatch_size"])
# In microbatch mode, we want to compute gradients on experience
# microbatches, average a number of these microbatches, and then apply
# the averaged gradient in one SGD step. This conserves GPU memory,
# allowing for extremely large experience batches to be used.
train_op = (
rollouts.combine(
ConcatBatches(
min_batch_size=config["microbatch_size"])).for_each(
ComputeGradients(workers)) # (grads, info)
.batch(num_microbatches) # List[(grads, info)]
.for_each(AverageGradients()) # (avg_grads, info)
.for_each(ApplyGradients(workers)))
else:
# In normal mode, we execute one SGD step per each train batch.
train_op = rollouts \
.combine(ConcatBatches(
min_batch_size=config["train_batch_size"])) \
.for_each(TrainOneStep(workers))
return StandardMetricsReporting(train_op, workers, config)
def execution_plan(workers, config):
local_replay_buffer = ReplayBuffer(config["buffer_size"])
rollouts = ParallelRollouts(workers, mode="bulk_sync")
# We execute the following steps concurrently:
# (1) Generate rollouts and store them in our local replay buffer. Calling
# next() on store_op drives this.
store_op = rollouts.for_each(StoreToReplayBuffer(local_replay_buffer))
# (2) Read and train on experiences from the replay buffer. Every batch
# returned from the LocalReplay() iterator is passed to TrainOneStep to
# take a SGD step, and then we decide whether to update the target network.
replay_op = LocalReplay(local_replay_buffer, config["train_batch_size"]) \
.for_each(TrainOneStep(workers)) \
.for_each(UpdateTargetNetwork(
workers, config["target_network_update_freq"]))
# Alternate deterministically between (1) and (2).
train_op = Concurrently([store_op, replay_op], mode="round_robin")
return StandardMetricsReporting(train_op, workers, config)
def execution_plan(workers, config):
# Collects experiences in parallel from multiple RolloutWorker actors.
rollouts = ParallelRollouts(workers, mode="bulk_sync")
# Combine experiences batches until we hit `train_batch_size` in size.
# Then, train the policy on those experiences and update the workers.
train_op = rollouts \
.combine(ConcatBatches(
min_batch_size=config["train_batch_size"])) \
.for_each(TrainOneStep(workers))
# Add on the standard episode reward, etc. metrics reporting. This returns
# a LocalIterator[metrics_dict] representing metrics for each train step.
return StandardMetricsReporting(train_op, workers, config)
def execution_plan(workers, config):
# Create a number of replay buffer actors.
# TODO(ekl) support batch replay options
num_replay_buffer_shards = config["optimizer"]["num_replay_buffer_shards"]
replay_actors = create_colocated(ReplayActor, [
num_replay_buffer_shards,
config["learning_starts"],
config["buffer_size"],
config["train_batch_size"],
config["prioritized_replay_alpha"],
config["prioritized_replay_beta"],
config["prioritized_replay_eps"],
], num_replay_buffer_shards)
# Update experience priorities post learning.
def update_prio_and_stats(item):
actor, prio_dict, count = item
actor.update_priorities.remote(prio_dict)
metrics = LocalIterator.get_metrics()
metrics.counters[STEPS_TRAINED_COUNTER] += count
metrics.timers["learner_dequeue"] = learner_thread.queue_timer
metrics.timers["learner_grad"] = learner_thread.grad_timer
metrics.timers["learner_overall"] = learner_thread.overall_timer
# Update worker weights as they finish generating experiences.
class UpdateWorkerWeights:
def __init__(self, learner_thread, workers, max_weight_sync_delay):
self.learner_thread = learner_thread
self.workers = workers
self.steps_since_update = collections.defaultdict(int)
self.max_weight_sync_delay = max_weight_sync_delay
self.weights = None
def __call__(self, item):
actor, batch = item
self.steps_since_update[actor] += batch.count
if self.steps_since_update[actor] >= self.max_weight_sync_delay:
# Note that it's important to pull new weights once
# updated to avoid excessive correlation between actors.
if self.weights is None or self.learner_thread.weights_updated:
self.learner_thread.weights_updated = False
self.weights = ray.put(
self.workers.local_worker().get_weights())
actor.set_weights.remote(self.weights)
self.steps_since_update[actor] = 0
# Update metrics.
metrics = LocalIterator.get_metrics()
metrics.counters["num_weight_syncs"] += 1
# Start the learner thread.
learner_thread = LearnerThread(workers.local_worker())
learner_thread.start()
# We execute the following steps concurrently:
# (1) Generate rollouts and store them in our replay buffer actors. Update
# the weights of the worker that generated the batch.
rollouts = ParallelRollouts(workers, mode="async", async_queue_depth=2)
store_op = rollouts \
.for_each(StoreToReplayActors(replay_actors)) \
.zip_with_source_actor() \
.for_each(UpdateWorkerWeights(
learner_thread, workers,
max_weight_sync_delay=config["optimizer"]["max_weight_sync_delay"])
)
# (2) Read experiences from the replay buffer actors and send to the
# learner thread via its in-queue.
replay_op = ParallelReplay(replay_actors, async_queue_depth=4) \
.zip_with_source_actor() \
.for_each(Enqueue(learner_thread.inqueue))
# (3) Get priorities back from learner thread and apply them to the
# replay buffer actors.
update_op = Dequeue(
learner_thread.outqueue, check=learner_thread.is_alive) \
.for_each(update_prio_and_stats) \
.for_each(UpdateTargetNetwork(
workers, config["target_network_update_freq"],
by_steps_trained=True))
# Execute (1), (2), (3) asynchronously as fast as possible.
merged_op = Concurrently([store_op, replay_op, update_op], mode="async")
return StandardMetricsReporting(merged_op, workers, config)