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

fill in top-level TODO items #156

Merged
merged 4 commits into from
Mar 20, 2025
Merged

fill in top-level TODO items #156

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
7b8a4c9
fill in top-level TODO items
dgrove-oss Mar 20, 2025
9552b28
tweaks
dgrove-oss Mar 20, 2025
191b633
tweaks
dgrove-oss Mar 20, 2025
8ee775b
fix indent
dgrove-oss Mar 20, 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
18 changes: 10 additions & 8 deletions setup.KubeConEU25/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -121,10 +121,10 @@ cd mlbatch
# Setup priority classes
kubectl apply -f setup.k8s/mlbatch-priorities.yaml

# Deploy scheduler plugins
# Deploy scheduler-plugins
helm install scheduler-plugins --namespace scheduler-plugins --create-namespace scheduler-plugins/manifests/install/charts/as-a-second-scheduler/ --set-json pluginConfig='[{"args":{"scoringStrategy":{"resources":[{"name":"nvidia.com/GPU","weight":1}],"requestedToCapacityRatio":{"shape":[{"utilization":0,"score":0},{"utilization":100,"score":10}]},"type":"RequestedToCapacityRatio"}},"name":"NodeResourcesFit"},{"args":{"permitWaitingTimeSeconds":300},"name":"Coscheduling"}]'

# Wait for scheduler-plugins pods to be running
# Wait for scheduler-plugins pods to be ready
while [[ $(kubectl get pods -n scheduler-plugins -o 'jsonpath={..status.conditions[?(@.type=="Ready")].status}' | tr ' ' '\n' | sort -u) != "True" ]]
do
echo -n "." && sleep 1;
Expand Down Expand Up @@ -154,8 +154,6 @@ do
done
echo ""

kubectl get pods -n mlbatch-system

# Deploy AppWrapper
kubectl apply --server-side -k setup.k8s/appwrapper/coscheduling

Expand Down Expand Up @@ -496,7 +494,8 @@ kubectl label servicemonitors.monitoring.coreos.com -n nvidia-GPU-operator nvidi

## Workload Management

TODO
We will now demonstrate the queueing, quota management, and fault recovery capabilities
of MLBatch using synthetic workloads.

<details>

Expand All @@ -506,7 +505,8 @@ TODO

## Example Workloads

We now run a few example workloads.
We now will now run some sample workloads that are representative of what is run on
an AI GPU Cluster.

### Batch Inference with vLLM

Expand Down Expand Up @@ -627,7 +627,8 @@ The two containers are synchronized as follows: `load-generator` waits for

### Pre-Training with PyTorch

TODO
In this example, `alice` uses the [Kubeflow Training Operator](https://github.com/kubeflow/training-operator)
to run a job that uses [PyTorch](https://pytorch.org) to train a machine learning model.

<details>

Expand All @@ -637,7 +638,8 @@ TODO

### Fine-Tuning with Ray

TODO
In this example, `alice` uses [KubeRay](https://github.com/ray-project/kuberay) to run a job that
uses [Ray](https://github.com/ray-project/ray) to fine tune a machine learning model.

<details>

Expand Down