radiation/cloud ML: workflow

projects/cloud_ml/Makefile

@@ -0,0 +1,66 @@
+.PHONY: surface_reference_zarr
+surface_reference_zarr:
+	cd scripts && python ./surface_reference_zarr.py
+
+.PHONY: nudge_to_fine_baseline_run
+nudge_to_fine_baseline_run: deploy
+	cd argo/prognostic-run && ./run.sh nudge-to-fine-baseline
+
+.PHONY: extend_prognostic_run
+extend_prognostic_run: deploy
+	cd argo/prognostic-run && ./extend.sh $(URL) $(SEGMENTS)
+
+.PHONY: training_data_zarr
+training_data_zarr:
+	cd scripts && python ./training_data_zarr.py training_data_config.yaml
+
+.PHONY: prescribed_cloud_cc_decorr_run
+prescribed_cloud_cc_decorr_run: deploy
+	cd argo/prognostic-run && ./run.sh prescribed-cloud-cc-decorr
+
+.PHONY: prescribed_cloud_cc_max_random_run
+prescribed_cloud_cc_max_random_run: deploy
+	cd argo/prognostic-run && ./run.sh prescribed-cloud-cc-max-random
+
+.PHONY: prescribed_cloud_cc_random_run
+prescribed_cloud_cc_random_run: deploy
+	cd argo/prognostic-run && ./run.sh prescribed-cloud-cc-random
+
+.PHONY: prescribed_cloud_decorr_run
+prescribed_cloud_decorr_run: deploy
+	cd argo/prognostic-run && ./run.sh prescribed-cloud-decorr
+
+.PHONY: prescribed_cloud_max_random_run
+prescribed_cloud_max_random_run: deploy
+	cd argo/prognostic-run && ./run.sh prescribed-cloud-max-random
+
+.PHONY: prescribed_cloud_random_run
+prescribed_cloud_random_run: deploy
+	cd argo/prognostic-run && ./run.sh prescribed-cloud-random
+
+.PHONY: train_cloud_ml_dense_seed
+train_cloud_ml_dense_seed: deploy
+	cd argo/training && ./run_random_seed.sh dense 4
+
+.PHONY: upload_squashed_models
+upload_squashed_models:
+	cd scripts/upload-squashed-models && ./upload.sh 4 gs://vcm-ml-experiments/cloud-ml/2023-06-07/train-cloud-ml-dense
+
+.PHONY: prescribed_cloud_dense_seed_squash_threshold
+prescribed_cloud_dense_seed_squash_threshold: deploy
+	cd argo/prognostic-run && ./run_seed_squash_threshold.sh prescribed-cloud-dense 4 39
+
+.PHONY: offline_cloud_predictions_zarr
+offline_cloud_predictions_zarr:
+	cd scripts && python ./offline_predictions_zarr.py offline_predictions_config.yaml
+
+.PHONY: offline_cloud_predictions_squashed_zarr
+offline_cloud_predictions_squashed_zarr:
+	cd scripts && python ./offline_predictions_zarr.py offline_predictions_squashed_config.yaml
+
+.PHONY: deploy
+deploy: argo/kustomize
+	cd argo && ./kustomize build . | kubectl apply -f -
+
+argo/kustomize:
+	cd argo && ./install_kustomize.sh 3.10.0

projects/cloud_ml/argo/fine-restarts-to-zarr/pod.yaml

@@ -0,0 +1,51 @@
+apiVersion: v1
+kind: Pod
+metadata:
+  name: fine-restarts-to-zarr
+spec:
+  volumes:
+    - name: gcp-credentials-user-gcp-sa
+      secret:
+        secretName: gcp-key
+  containers:
+    - name: main
+      image: us.gcr.io/vcm-ml/fv3net:400c970a0af572973f5c6b8e970aa261b4baf8a2
+      command: ["bash", "-x", "-c"]
+      workingDir: /home/jovyan/fv3net/workflows/dataflow
+      env:
+        - name: GOOGLE_APPLICATION_CREDENTIALS
+          value: /secret/gcp-credentials/key.json
+        - name: CLOUDSDK_AUTH_CREDENTIAL_FILE_OVERRIDE
+          value: /secret/gcp-credentials/key.json
+        - name: POD_NAME
+          valueFrom:
+            fieldRef:
+              fieldPath: metadata.name
+      volumeMounts:
+        - mountPath: /secret/gcp-credentials
+          name: gcp-credentials-user-gcp-sa
+      args:
+        - |
+          ./dataflow.sh submit -m fv3net.pipelines.restarts_to_zarr \
+            gs://vcm-ml-raw-flexible-retention/2023-05-22-PIRE-like-C3072-ccnorm-true-simulation/C3072-to-C48-restarts \
+            gs://vcm-ml-intermediate/2023-05-22-PIRE-like-C3072-ccnorm-true-simulation-restarts.zarr \
+            --no-coarse-suffix \
+            --job_name $POD_NAME-$(openssl rand -hex 6) \
+            --project vcm-ml \
+            --region us-central1 \
+            --runner DataFlow \
+            --temp_location gs://vcm-ml-scratch/tmp_dataflow \
+            --num_workers 64 \
+            --autoscaling_algorithm=NONE
+      resources:
+        limits:
+          memory: 2Gi
+          cpu: "1000m"
+        requests:
+          memory: 1Gi
+          cpu: "1000m"
+  tolerations:
+    - effect: NoSchedule
+      key: dedicated
+      value: med-sim-pool
+  restartPolicy: Never

projects/cloud_ml/argo/kustomization.yaml

@@ -4,8 +4,10 @@ resources:
 kind: Kustomization
 images:
 - name: us.gcr.io/vcm-ml/fv3net
-  newTag: 762dfc1e945720f59f5b9e9ab68e10e82a0594a9
+  newTag: b9decb4be05133c569a79223b50879e2fd292bd0
 - name: us.gcr.io/vcm-ml/post_process_run
-  newTag: 762dfc1e945720f59f5b9e9ab68e10e82a0594a9
+  newTag: b9decb4be05133c569a79223b50879e2fd292bd0
 - name: us.gcr.io/vcm-ml/prognostic_run
-  newTag: 762dfc1e945720f59f5b9e9ab68e10e82a0594a9
+  newTag: b9decb4be05133c569a79223b50879e2fd292bd0
+- name: us.gcr.io/vcm-ml/fv3fit
+  newTag: b9decb4be05133c569a79223b50879e2fd292bd0