Skip to content
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

Enable multi-GPU inference in vLLM with tensor parallelism #105

Merged
merged 2 commits into from
Feb 11, 2025
+22 −3
Merged

Enable multi-GPU inference in vLLM with tensor parallelism #105

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ac5ee77
Adding TP>1 support
YuhanLiu11 Feb 11, 2025
55c6502
fixing bugs in template, and adding shm size as a configurable param
YuhanLiu11 Feb 11, 2025
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
2 changes: 1 addition & 1 deletion helm/Chart.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,7 +15,7 @@ type: application
# This is the chart version. This version number should be incremented each time you make changes
# to the chart and its templates, including the app version.
# Versions are expected to follow Semantic Versioning (https://semver.org/)
version: 0.0.3
version: 0.0.4

maintainers:
- name: apostac
20 changes: 18 additions & 2 deletions helm/templates/deployment-vllm-multi.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,6 +47,10 @@ spec:
- "--dtype"
- {{ .dtype | quote }}
{{- end }}
{{- if hasKey . "tensorParallelSize" }}
- "--tensor-parallel-size"
- {{ .tensorParallelSize | quote }}
{{- end }}
{{- if .extraArgs }}
{{- range .extraArgs }}
- {{ . | quote }}
Expand Down Expand Up @@ -108,12 +112,24 @@ spec:
volumeMounts:
- name: {{ .Release.Name }}-storage
mountPath: /data

{{- with $modelSpec.vllmConfig }}
{{- if hasKey $modelSpec.vllmConfig "tensorParallelSize"}}
- name: shm
mountPath: /dev/shm
{{- end}}
{{- end}}
volumes:
- name: {{ .Release.Name }}-storage
persistentVolumeClaim:
claimName: "{{ .Release.Name }}-{{$modelSpec.name}}-storage-claim"

{{- with $modelSpec.vllmConfig }}
{{- if hasKey $modelSpec.vllmConfig "tensorParallelSize"}}
- name: shm
emptyDir:
medium: Memory
sizeLimit: {{ default "20Gi" $modelSpec.shmSize }}
{{- end}}
{{- end}}
{{- if .Values.servingEngineSpec.tolerations }}
{{- with .Values.servingEngineSpec.tolerations }}
tolerations:
Expand Down
3 changes: 3 additions & 0 deletions helm/values.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,6 +29,7 @@ servingEngineSpec:
# - enableChunkedPrefill: (optional, bool) Enable chunked prefill, e.g., false
# - maxModelLen: (optional, int) The maximum model length, e.g., 16384
# - dtype: (optional, string) The data type, e.g., "bfloat16"
# - tensorParallelSize: (optional, int) The degree of tensor parallelism, e.g., 2
# - extraArgs: (optional, list) Extra command line arguments to pass to vLLM, e.g., ["--disable-log-requests"]
#
# - lmcacheConfig: (optional, map) The configuration of the LMCache for KV offloading, supported options are:
Expand All @@ -41,6 +42,8 @@ servingEngineSpec:
#
# - nodeSelectorTerms: (optional, list) The node selector terms to match the nodes
#
# - shmSize: (optional, string) The size of the shared memory, e.g., "20Gi"
#
# Example:
# modelSpec:
# - name: "mistral"
Expand Down