[RLlib] Make checkpointing test have multiple nodes, make node for dq…

…n test larger (ray-project#37127) Signed-off-by: Avnish <[email protected]>
krfricke · Jul 6, 2023 · 509184a · 509184a
1 parent 8f17b73
commit 509184a
Show file tree

Hide file tree

Showing 4 changed files with 22 additions and 5 deletions.
diff --git a/release/release_tests.yaml b/release/release_tests.yaml
@@ -4190,7 +4190,7 @@
  team: rllib
  cluster:
  cluster_env: app_config.yaml
- cluster_compute: 1gpu_16cpus.yaml
+ cluster_compute: 1gpu_32cpus.yaml
 
  run:
  timeout: 18000

diff --git a/release/rllib_tests/1gpu_32cpus.yaml b/release/rllib_tests/1gpu_32cpus.yaml
@@ -0,0 +1,17 @@
+cloud_id: {{env["ANYSCALE_CLOUD_ID"]}}
+region: us-west-2
+
+max_workers: 0
+
+head_node_type:
+ name: head_node
+ instance_type: g5.8xlarge
+
+worker_node_types: []
+
+aws:
+ BlockDeviceMappings:
+ - DeviceName: /dev/sda1
+ Ebs:
+ DeleteOnTermination: true
+ VolumeSize: 500
diff --git a/release/rllib_tests/multi_node_checkpointing_compute_config.yaml b/release/rllib_tests/multi_node_checkpointing_compute_config.yaml
@@ -9,9 +9,9 @@ head_node_type:
 
 worker_node_types:
  - name: worker_node
- instance_type: g3.8xlarge
- min_workers: 1
- max_workers: 1
+ instance_type: g3s.xlarge
+ min_workers: 2
+ max_workers: 2
  use_spot: false
 
 aws:

diff --git a/rllib/core/learner/torch/torch_learner.py b/rllib/core/learner/torch/torch_learner.py
@@ -69,7 +69,7 @@ def __init__(
  # Will be set during build.
  self._device = None
 
- # Whether to compile the RL Module of this learner. This implies that the
+ # Whether to compile the RL Module of this learner. This implies that the.
  # forward_train method of the RL Module will be compiled. Further more,
  # other forward methods of the RL Module will be compiled on demand.
  # This is assumed to not happen, since other forwrad methods are not expected