Skip to content

Upgrade vllm image #119

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

Merged
AhmedSeemalK merged 4 commits into from
Jul 27, 2026
Merged

Upgrade vllm image #119

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
1d355c0
vLLM cpu image upgrade to v0.21.0 (#110)
sandeshk-intel May 20, 2026
f7a881c
Add vllm user to avoid get user issue during torch compilation (#111)
sandeshk-intel May 25, 2026
4fe496d
vLLM image update to v0.24.0 (#118)
sandeshk-intel Jul 8, 2026
66f457b
Remove SGL kernel flag in default config and update in docs (#120)
sandeshk-intel Jul 27, 2026
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 .github/workflows/code-scans.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -52,7 +52,7 @@ jobs:
continue-on-error: true
with:
scan-type: 'image'
image-ref: 'public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:v0.10.2'
image-ref: 'vllm/vllm-openai-cpu:v0.24.0'
severity: 'HIGH,CRITICAL'
exit-code: 0
format: 'table'
Expand Down
4 changes: 2 additions & 2 deletions core/helm-charts/vllm/values.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,11 +21,11 @@ accelDevice: ""
port: 2080
shmSize: 1Gi
image:
repository: public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo
repository: vllm/vllm-openai-cpu
# Uncomment the following line to set desired image pull policy if needed, as one of Always, IfNotPresent, Never.
# pullPolicy: ""
# Overrides the image tag whose default is the chart appVersion.
tag: "v0.10.2"
tag: "v0.24.0"

imagePullSecrets: []
nameOverride: ""
Expand Down
24 changes: 18 additions & 6 deletions core/helm-charts/vllm/xeon-values.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,8 +26,8 @@ max_model_len: 8192

# vLLM CPU image configuration
image:
repository: public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo
tag: "v0.10.2"
repository: vllm/vllm-openai-cpu
tag: "v0.24.0"
pullPolicy: IfNotPresent

# Node affinity for Xeon inference nodes
Expand Down Expand Up @@ -58,6 +58,8 @@ modelConfigs:
VLLM_CPU_NUM_OF_RESERVED_CPU: "0"
HF_HUB_DISABLE_XET: "1"
VLLM_CPU_SGL_KERNEL: "1"
USER: "vllm"
TRITON_CACHE_DIR: "/tmp/.triton"
extraCmdArgs:
[
"--block-size",
Expand All @@ -74,7 +76,6 @@ modelConfigs:
"2048",
"--max-num-seqs",
"256",
"--disable-log-requests",
"--enable-auto-tool-choice",
"--tool-call-parser", "llama3_json",
]
Expand All @@ -90,6 +91,8 @@ modelConfigs:
VLLM_CPU_NUM_OF_RESERVED_CPU: "0"
VLLM_CPU_SGL_KERNEL: "1"
HF_HUB_DISABLE_XET: "1"
USER: "vllm"
TRITON_CACHE_DIR: "/tmp/.triton"
extraCmdArgs:
[
"--block-size",
Expand Down Expand Up @@ -121,6 +124,8 @@ modelConfigs:
VLLM_CPU_NUM_OF_RESERVED_CPU: "0"
VLLM_CPU_SGL_KERNEL: "1"
HF_HUB_DISABLE_XET: "1"
USER: "vllm"
TRITON_CACHE_DIR: "/tmp/.triton"
extraCmdArgs:
[
"--block-size",
Expand All @@ -137,7 +142,6 @@ modelConfigs:
"2048",
"--max-num-seqs",
"256",
"--disable-log-requests",
]
tensor_parallel_size: "{{ .Values.tensor_parallel_size }}"
pipeline_parallel_size: "{{ .Values.pipeline_parallel_size }}"
Expand All @@ -151,6 +155,8 @@ modelConfigs:
VLLM_CPU_NUM_OF_RESERVED_CPU: "0"
VLLM_CPU_SGL_KERNEL: "1"
HF_HUB_DISABLE_XET: "1"
USER: "vllm"
TRITON_CACHE_DIR: "/tmp/.triton"
extraCmdArgs:
[
"--block-size",
Expand All @@ -167,7 +173,6 @@ modelConfigs:
"2048",
"--max-num-seqs",
"256",
"--disable-log-requests",
]
tensor_parallel_size: "{{ .Values.tensor_parallel_size }}"
pipeline_parallel_size: "{{ .Values.pipeline_parallel_size }}"
Expand All @@ -181,6 +186,8 @@ modelConfigs:
VLLM_CPU_NUM_OF_RESERVED_CPU: "0"
VLLM_CPU_SGL_KERNEL: "1"
HF_HUB_DISABLE_XET: "1"
USER: "vllm"
TRITON_CACHE_DIR: "/tmp/.triton"
extraCmdArgs:
[
"--block-size",
Expand Down Expand Up @@ -212,6 +219,8 @@ modelConfigs:
VLLM_CPU_NUM_OF_RESERVED_CPU: "0"
VLLM_CPU_SGL_KERNEL: "1"
HF_HUB_DISABLE_XET: "1"
USER: "vllm"
TRITON_CACHE_DIR: "/tmp/.triton"
extraCmdArgs:
[
"--block-size",
Expand Down Expand Up @@ -245,6 +254,8 @@ modelConfigs:
VLLM_CPU_SGL_KERNEL: "1"
HF_HUB_DISABLE_XET: "1"
LOGNAME: "vllm"
USER: "vllm"
TRITON_CACHE_DIR: "/tmp/.triton"
extraCmdArgs:
[
"--block-size",
Expand Down Expand Up @@ -274,8 +285,9 @@ defaultModelConfigs:
VLLM_ALLOW_LONG_MAX_MODEL_LEN: "1"
VLLM_ENGINE_ITERATION_TIMEOUT_S: "120"
VLLM_CPU_NUM_OF_RESERVED_CPU: "0"
VLLM_CPU_SGL_KERNEL: "1"
HF_HUB_DISABLE_XET: "1"
USER: "vllm"
TRITON_CACHE_DIR: "/tmp/.triton"
extraCmdArgs:
[
"--block-size",
Expand Down
2 changes: 1 addition & 1 deletion core/roles/inference-tools/tasks/main.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@
ansible.builtin.pip:
name: kubernetes
state: present
extra_args: "--break-system-packages"
extra_args: "--break-system-packages --ignore-installed"
become: true
when: pip_install_result is failed
tags: always
Expand Down
2 changes: 1 addition & 1 deletion core/scripts/vllm-quickstart/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -116,7 +116,7 @@ The `models.json` file contains all configuration:
```json
{
"docker": {
"image": "public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:v0.11.2",
"image": "vllm/vllm-openai-cpu:v0.24.0",
"port": "8000:8000",
"environment": { ... },
"volumes": [ ... ]
Expand Down
2 changes: 1 addition & 1 deletion core/scripts/vllm-quickstart/models.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
{
"docker": {
"image": "public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:v0.11.2",
"image": "vllm/vllm-openai-cpu:v0.24.0",
"port": "8000:8000",
"environment": {
"VLLM_CPU_SGL_KERNEL": "1",
Expand Down
23 changes: 23 additions & 0 deletions docs/deploy-llm-model-from-hugging-face.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,3 +20,26 @@ This option allows you to deploy any Hugging Face-hosted LLM on the Inference Cl
- **Tensor Parallel Size** (based on available Intel® AI Accelerator cards)

> **Note**: This deploys a model that has **not** been pre-validated. Make sure the tensor parallel size is configured correctly. An incorrect value can result in the model being stuck in a "not ready" state.

### Customizing Environment Variables and vLLM Arguments

Models deployed from Hugging Face use the `defaultModelConfigs` settings in `core/helm-charts/vllm/xeon-values.yaml`. You can customize environment variables and vLLM command-line arguments by editing the `defaultModelConfigs` section of that file before deploying.

**Example: enabling the SGL kernel, configuring KV cache, and other environment variables**

```yaml
defaultModelConfigs:
configMapValues:
VLLM_CPU_SGL_KERNEL: "1" # Enable the SGL kernel for improved CPU performance
VLLM_CPU_KVCACHE_SPACE: "40" # KV cache memory allocation in GB
# Add or override any other environment variables here
extraCmdArgs:
[
"--block-size", "128",
"--dtype", "bfloat16",
"--max-model-len", "8192",
# Add or override any vLLM CLI arguments here
]
```

You can set any vLLM-supported environment variable under `configMapValues` and any vLLM CLI flag under `extraCmdArgs`.
Loading