Skip to content

add CPU BKMs #20

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
+29 −2
Open

add CPU BKMs #20

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
24 changes: 23 additions & 1 deletion helm/templates/servingruntime.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,6 +24,28 @@ spec:
- {{ .Values.model.maxModelLen | quote }}
- '--served-model-name'
- {{ .Values.model.name }}
- '--tensor-parallel-size'
- {{ .Values.resources.inference.tp }}
- '--dtype'
- {{ .Values.model.dtype }}
- '--enforce-eager'
- '--distributed-executor-backend'
- "mp"
- '--block-size'
- {{ .Values.model.block_size }}
- '--max-num-batched-tokens'
- {{ .Values.model.max_num_batched_tokens }}
- '--max-num-seqs'
- {{ .Values.model.max_num_seqs }}
env:
- name: VLLM_CPU_KVCACHE_SPACE
value: "40"
- name: VLLM_CPU_SGL_KERNEL
value: "1"
- name: VLLM_ALLOW_LONG_MAX_MODEL_LEN
value: "1"
- name: VLLM_RPC_TIMEOUT
value: "100000"
image: {{ .Values.images.vllmRuntime.repository }}:{{ .Values.images.vllmRuntime.tag }}
name: kserve-container
ports:
Expand All @@ -33,4 +55,4 @@ spec:
multiModel: false
supportedModelFormats:
- autoSelect: true
name: vLLM
name: vLLM
7 changes: 6 additions & 1 deletion helm/values.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,14 +40,19 @@ model:
storageUri: "oci://quay.io/rh-aiservices-bu/tinyllama:1.0"
name: "tinyllama"
maxModelLen: 2048
dtype: "bfloat16"
block_size: "128"
max_num_batched_tokens: "2048"
max_num_seqs: "256"

resources:
inference:
requests:
cpu: "2"
memory: "4Gi"
tp: "1"
limits:
cpu: "8"
memory: "8Gi"

storageClassName: gp3-csi
storageClassName: gp3-csi