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

Add gh200 support and model #300

Merged
merged 2 commits into from
Nov 2, 2024
Merged

Add gh200 support and model #300

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
bd26bcd
add gh200 support and model
happytreees Oct 31, 2024
783795e
remove node label selector
happytreees Nov 1, 2024
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
14 changes: 14 additions & 0 deletions charts/kubeai/values.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,6 +20,7 @@ modelServers:
default: "vllm/vllm-openai:v0.6.3.post1"
cpu: "substratusai/vllm:v0.6.3.post1-cpu"
google-tpu: "substratusai/vllm:v0.6.3.post1-tpu"
gh200: "drikster80/vllm-gh200-openai:v0.6.3.post1"
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

the plan would be to remove this once the default image works with gh200. I do like how you created an image just for gh200 for now.

OLlama:
images:
default: "ollama/ollama:latest"
Expand Down Expand Up @@ -97,6 +98,19 @@ resourceProfiles:
operator: "Equal"
value: "present"
effect: "NoSchedule"
nvidia-gpu-gh200:
imageName: "gh200"
limits:
nvidia.com/gpu: "1"
requests:
nvidia.com/gpu: "1"
tolerations:
- key: "nvidia.com/gpu"
operator: "Equal"
value: "present"
effect: "NoSchedule"
nodeSelector:
happytreees marked this conversation as resolved.
Show resolved Hide resolved
nvidia.com/gpu.family: "hopper"
nvidia-gpu-a100-80gb:
imageName: "nvidia-gpu"
limits:
Expand Down
12 changes: 12 additions & 0 deletions charts/models/values.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -161,6 +161,18 @@ catalog:
# You can also use nvidia-gpu-a100-80gb:8
resourceProfile: nvidia-gpu-h100:8
targetRequests: 500
llama-3.1-70b-instruct-awq-int4-gh200:
enabled: false
features: [TextGeneration]
url: hf://hugging-quants/Meta-Llama-3.1-70B-Instruct-AWQ-INT4
engine: VLLM
args:
- --max-model-len=16384
- --max-num-batched-token=16384
- --enable-prefix-caching
- --disable-log-requests
resourceProfile: nvidia-gpu-gh200:1
targetRequests: 50
samos123 marked this conversation as resolved.
Show resolved Hide resolved
llama-3.1-405b-instruct-fp8-a100-80b:
features: [TextGeneration]
url: hf://neuralmagic/Meta-Llama-3.1-405B-Instruct-FP8
Expand Down
17 changes: 17 additions & 0 deletions manifests/models/llama-3.1-70b-instruct-awq-int4-gh200.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,17 @@
apiVersion: kubeai.org/v1
kind: Model
metadata:
name: llama-3.1-70b-instruct-awq-int4-gh200
spec:
features: [TextGeneration]
owner:
url: hf://hugging-quants/Meta-Llama-3.1-70B-Instruct-AWQ-INT4
engine: VLLM
args:
- --max-model-len=16384
- --max-num-batched-token=16384
- --enable-prefix-caching
- --disable-log-requests
targetRequests: 50
minReplicas: 1
resourceProfile: nvidia-gpu-gh200:1