Fix venv guidance

scripts/cluster_k8s.yaml

@@ -0,0 +1,73 @@
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: sft-trainer-config-alex
+data:
+  config.json: |
+    {
+        "model_name_or_path": "/granite/granite-13b-base-v2/step_300000_ckpt",
+        "training_data_path": "/data/anhuong/twitter_complaints.json",
+        "output_dir": "/data/anhuong/tuning_output/some-name",
+        "num_train_epochs": 20.0,
+        "per_device_train_batch_size": 2,
+        "gradient_accumulation_steps": 1,
+        "evaluation_strategy": "no",
+        "save_strategy": "epoch",
+        "learning_rate": 1e-5,
+        "response_template": "\n### Label:",
+        "dataset_text_field": "output",
+        "tokenizer_name_or_path": "/granite/granite-13b-base-v2/step_300000_ckpt"
+    }
+---
+apiVersion: v1
+kind: Pod
+metadata:
+  name: sft-trainer-test-alex-granite-13b
+spec:
+  securityContext:
+    runAsUser: 1000850000
+    runAsGroup: 0
+  containers:
+    - env:
+        - name: SFT_TRAINER_CONFIG_JSON_PATH
+          value: /config/config.json
+        - name: LOG_LEVEL
+          value: info
+      image: docker-na-public.artifactory.swg-devops.com/wcp-ai-foundation-team-docker-virtual/sft-trainer:7f32893_ubi9_py311
+      imagePullPolicy: IfNotPresent
+      name: train-conductor-training
+      command: ["sleep", "99999"]
+      resources:
+        limits:
+          nvidia.com/gpu: "2"
+          memory: 200Gi
+          cpu: "10"
+        requests:
+          memory: 80Gi
+          cpu: "5"
+      volumeMounts:
+        - mountPath: /data
+          name: input-data
+        - mountPath: /config
+          name: sft-trainer-config
+        - mountPath: /llama
+          name: llama-eval-pvc
+        - mountPath: /granite
+          name: ibm-granite-pvc
+  imagePullSecrets:
+    - name: artifactory-docker-anh
+  restartPolicy: Never
+  terminationGracePeriodSeconds: 30
+  volumes:
+    - name: input-data
+      persistentVolumeClaim:
+        claimName: fms-tuning-pvc
+    - name: sft-trainer-config
+      configMap:
+        name: sft-trainer-config
+    - name: ibm-granite-pvc
+      persistentVolumeClaim:
+        claimName: ibm-granite-pvc
+    - name: llama-eval-pvc
+      persistentVolumeClaim:
+        claimName: llama-eval-pvc