2020-02-24 21:18:53 -08:00
|
|
|
cluster_name: ray-rllib-stress-tests
|
2019-01-16 02:05:16 -08:00
|
|
|
|
2020-02-24 21:18:53 -08:00
|
|
|
min_workers: 9
|
|
|
|
|
max_workers: 9
|
2019-01-16 02:05:16 -08:00
|
|
|
|
2020-12-08 19:15:07 +01:00
|
|
|
idle_timeout_minutes: 15
|
2019-01-16 02:05:16 -08:00
|
|
|
|
2020-12-08 19:15:07 +01:00
|
|
|
docker:
|
|
|
|
|
image: anyscale/ray-ml:latest-gpu
|
|
|
|
|
container_name: ray_container
|
|
|
|
|
pull_before_run: True
|
2019-01-16 02:05:16 -08:00
|
|
|
|
|
|
|
|
provider:
|
|
|
|
|
type: aws
|
2020-01-13 15:01:03 -06:00
|
|
|
region: us-west-2
|
|
|
|
|
availability_zone: us-west-2a
|
2020-02-24 21:18:53 -08:00
|
|
|
cache_stopped_nodes: False
|
2019-01-16 02:05:16 -08:00
|
|
|
|
2021-05-04 23:10:04 +02:00
|
|
|
available_node_types:
|
|
|
|
|
gpu_ondemand:
|
|
|
|
|
node_config:
|
|
|
|
|
InstanceType: p3.16xlarge
|
|
|
|
|
resources: {"CPU": 56, "GPU": 8} # Keep one CPU free for each GPU
|
|
|
|
|
min_workers: 0
|
|
|
|
|
max_workers: 0
|
|
|
|
|
cpu_ondemand:
|
|
|
|
|
node_config:
|
|
|
|
|
InstanceType: m5.16xlarge
|
|
|
|
|
resources: {"CPU": 64}
|
|
|
|
|
min_workers: 9
|
|
|
|
|
max_workers: 9
|
|
|
|
|
|
2019-01-16 02:05:16 -08:00
|
|
|
auth:
|
|
|
|
|
ssh_user: ubuntu
|
|
|
|
|
|
2021-05-04 23:10:04 +02:00
|
|
|
head_node_type: gpu_ondemand
|
|
|
|
|
worker_default_node_type: cpu_ondemand
|
2019-01-16 02:05:16 -08:00
|
|
|
|
2020-12-08 19:15:07 +01:00
|
|
|
setup_commands:
|
2021-05-04 23:10:04 +02:00
|
|
|
- sudo apt update
|
|
|
|
|
- sudo apt-get install -y libglib2.0-0 libcudnn7=7.6.5.32-1+cuda10.1
|
|
|
|
|
# libcudnn8=8.0.5.39-1+cuda11.0
|
|
|
|
|
- pip install tensorflow-gpu==2.3.0
|
|
|
|
|
- pip install -U {{env["RAY_WHEEL"]}}
|
