add ml example

examples/tests/dlio-ml/README.md

@@ -0,0 +1,72 @@
+# DLIO ML Example
+
+This is an example of using the IO tool[DLIO](https://dlio-profiler.readthedocs.io/en/latest/build.html#build-dlio-profiler-with-pip-recommended) that can 
+be added on the fly with pip.
+
+## Usage
+
+Create a cluster and install JobSet to it.
+
+```bash
+kind create cluster
+VERSION=v0.2.0
+kubectl apply --server-side -f https://github.com/kubernetes-sigs/jobset/releases/download/$VERSION/manifests.yaml
+```
+
+Install the operator (from the development manifest here):
+
+```bash
+$ kubectl apply -f ../../dist/metrics-operator-dev.yaml
+```
+
+How to see metrics operator logs:
+
+```bash
+$ kubectl logs -n metrics-system metrics-controller-manager-859c66464c-7rpbw
+```
+
+Then create the metrics set. This is going to run a single run of LAMMPS over MPI.
+as lammps runs.
+
+```bash
+kubectl apply -f metrics.yaml
+```
+
+Wait until you see pods created by the job and then running (there should be two - a launcher and worker for LAMMPS):
+
+```bash
+kubectl get pods
+```
+```diff
+NAME                           READY   STATUS    RESTARTS   AGE
+metricset-sample-l-0-0-lt782   1/1     Running   0          3s
+metricset-sample-w-0-0-4s5p9   1/1     Running   0          3s
+```
+
+In the above, "l" is a launcher pod, and "w" is a worker node.
+If you inspect the log for the launcher you'll see a short sleep (the network isn't up immediately)
+and then LAMMPS running, and the log is printed to the console.
+
+```bash
+kubectl logs metricset-sample-l-0-0-lt782 -f
+```
+
+There is purposefully a sleep infinity at the end to give you a chance to copy over data.
+
+```bash
+mkdir -p ./data ./output
+# Only if you are interested in the ML data
+kubectl  cp metricset-sample-m-0-0-xfg6r:/dlio/data ./data/
+kubectl  cp metricset-sample-m-0-0-xfg6r:/dlio/output ./output
+```
+
+You can open the tiny file in [https://ui.perfetto.dev/](https://ui.perfetto.dev/).
+
+![img/trace.png](img/trace.png)
+
+Other applications of interest might be related to AI/ML - we will try more soon!
+Cleanup when you are done.
+
+```bash
+kubectl delete -f metrics.yaml
+```

examples/tests/dlio-ml/metrics.yaml

@@ -0,0 +1,37 @@
+apiVersion: flux-framework.org/v1alpha2
+kind: MetricSet
+metadata:
+  labels:
+    app.kubernetes.io/name: metricset
+    app.kubernetes.io/instance: metricset-sample
+  name: metricset-sample
+spec:
+  # kubectl apply -f metrics.yaml
+  # kubectl logs <launcher-pod> -f
+  pods: 1
+
+  metrics:
+   - name: io-ior
+     options:
+      command: mpirun --allow-run-as-root -np 10 dlio_benchmark workload=resnet50 ++workload.dataset.data_folder=/dlio/data ++workload.output.folder=/dlio/output
+      workdir: /dlio/data
+     addons:
+       - name: commands
+         options:
+           preBlock: |
+             apt-get update && apt-get install -y python3 python3-pip openmpi-bin openmpi-common libopenmpi-dev hwloc libhwloc-dev default-jre
+             #python3 -m pip install git+https://github.com/hariharan-devarajan/dlio-profiler.git
+             #python3 -m pip install git+https://github.com/argonne-lcf/dlio_benchmark.git
+             python3 -m pip install "dlio_benchmark[dlio_profiler] @ git+https://github.com/argonne-lcf/dlio_benchmark.git"
+             mkdir -p /dlio/data /dlio/output /dlio/logs
+             export DLIO_PROFILER_ENABLE=0
+             mpirun -np 10 --allow-run-as-root dlio_benchmark workload=resnet50 ++workload.dataset.data_folder=/dlio/data ++workload.output.folder=/dlio/output ++workload.workflow.generate_data=True ++workload.workflow.train=False
+             export DLIO_PROFILER_LOG_LEVEL=ERROR
+             export DLIO_PROFILER_ENABLE=1
+             export DLIO_PROFILER_INC_METADATA=1
+             cd /dlio/data
+           postBlock: |
+             gzip -d /dlio/output/.trace*.pfw.gz
+             cat /dlio/output/.trace*.pfw
+             sleep infinity
+