ray/rllib/tuned_examples/pong-impala-vectorized.yaml
Eric Liang dd70720578
[rllib] Rename sample_batch_size => rollout_fragment_length (#7503)
* bulk rename

* deprecation warn

* update doc

* update fig

* line length

* rename

* make pytest comptaible

* fix test

* fi sys

* rename

* wip

* fix more

* lint

* update svg

* comments

* lint

* fix use of batch steps
2020-03-14 12:05:04 -07:00

11 lines
454 B
YAML

# This can reach 18-19 reward within 10 minutes on a Tesla M60 GPU (e.g., G3 EC2 node)
# with 32 workers and 10 envs per worker. This is more efficient than the non-vectorized
# configuration which requires 128 workers to achieve the same performance.
pong-impala-vectorized:
env: PongNoFrameskip-v4
run: IMPALA
config:
rollout_fragment_length: 50
train_batch_size: 500
num_workers: 32
num_envs_per_worker: 10