train.yaml

apiVersion: v1
kind: Pod
metadata:
  generateName: aggrmean-seq10-pred2-
  labels:
    type: interactive
    research-group: rl-team
spec:
  hostIPC: true
  restartPolicy: Never
  containers:
  - name: nri-train
    command: ["python", "-u", "train.py", 
    --name="fulltraining_lr1e-7_sparsity",
    --epochs=200,
    --batch_size=2,
    --seq_len=49,
    --predicted_timesteps=10,
    --lr=1e-7,
    --batch_limit=0,
    --sparsity_prior, 
    ]
    # command: ["sleep", "infinity"]
    image: amr-registry.caas.intel.com/aipg/mpettee-pytorch-geometric
    workingDir: /workspace/choreo-graph/
    resources:
      limits:
        nvidia.com/gpu: 1
        cpu: 4
        memory: 50Gi
    volumeMounts:
    - mountPath: "/workspace"
      name: workspace 
    envFrom:
    - configMapRef:
        name: proxy-config
    env:
    - name: LD_LIBRARY_PATH
      value: $LD_LIBRARY_PATH:/.mujoco/mjpro150/bin
    - name: OMP_NUM_THREADS
      value: "1"
  securityContext:
    fsGroup: 17685
    runAsUser: 11964579
  volumes:
  - name: workspace
    persistentVolumeClaim:
      claimName: workdisk