Files
kserve/docs/samples/llmisvc/e2e-gpt-oss/inference_pd_disaggregation.yaml
2026-04-02 18:32:37 -04:00

33 lines
729 B
YAML

---
apiVersion: serving.kserve.io/v1alpha2
kind: LLMInferenceService
metadata:
name: gpt-oss-20b
spec:
# Main/decode pool replica count
model:
uri: "pvc://gpt-oss-20b-pvc"
name: RedHatAI/gpt-oss-20b
replicas: 1
baseRefs:
- name: llmisvc-config-pd-disagg
template:
containers:
- name: main
resources:
limits:
nvidia.com/gpu: "1"
requests:
nvidia.com/gpu: "1"
prefill:
# Prefill pool replica count (higher for concurrent prefill requests)
replicas: 2
template:
containers:
- name: main
resources:
limits:
nvidia.com/gpu: "1"
requests:
nvidia.com/gpu: "1"