1# Task name (optional), used for display purposes.
2name: my-task
3# Working directory (optional)
4workdir: ~/my-task-code
5resources:
6  cloud: aws  # The cloud to use (optional).
7  # The region to use (optional)
8  region: us-east-1
9  # The zone to use (optional)
10  zone: us-east-1a
11  # 1x A10/A10G GPU
12  accelerators: {A10G:1, A10:1}
13  # Open port 8000
14  ports: [8000]
15
16env:
17  # set the environment variable MY_ENV_VAR to my-value
18  MY_ENV_VAR: my-value
19
20# Copy the contents of the current directory onto the remote machine
21workdir: .
22
23# Typical use: pip install -r requirements.txt
24# Invoked under the workdir (ie. can use its files)
25setup: |
26  echo "Running setup operations"
27
28# Typical uses:
29#    torchserve ..
30#    python -u -m vllm.entrypoints.openai.api_server ...
31# Invoked under the workdir (ie. can use its files)
32run: |
33  python -m http.server --port 8000
34
35service:
36  replica_policy:
37    min_replicas: 1
38    max_replicas: 5
39    # When the average QPS (queries per second) per replica goes above this number,
40    # Outpost will dynamically scale up the number of replicas running your service,
41    # up to a maximum of max_replicas.
42    # Similarly, when the average QPS per replica goes below this number, Outpost will
43    # scale down the number of replicas running your service, down to a minimum of min_replicas.
44    target_qps_per_replica: 10
45  
46  readiness_probe:
47    # this is the endpoint path within your service that Outpost uses to check if your service is running
48    path: /