Advanced Model Training Example: Run ResNet Multi-node Cluster

The below are instructions for setting up a ResNet dataset and performing distributed training. ResNet Model, from Model References, requires dependencies to be installed.

Build and Store Custom Docker Image

1. Create a Dockerfile with the below content:

   FROM vault.habana.ai/gaudi-docker/1.21.0/ubuntu22.04/habanalabs/pytorch-installer-2.6.0:latest
   
   # Clones Model References and installs ResNet dependencies
   RUN git clone -b 1.21.0 https://github.com/HabanaAI/Model-References.git /Model-References && \
   python3 -m pip install -r /Model-References/PyTorch/computer_vision/classification/torchvision/requirements.txt
   

2. Build and push the image to AWS’s Elastic Container Registry (ECR) for ease of access on EC2 instances. For further information on how to build and push an image to ECR, refer to Amazon ECR Getting Started Guide.

Upload Data to Elastic File System (EFS)

Follow the instructions for ResNet Data Generation and upload that to Amazon EFS. For more details on how to use EFS, refer to Amazon EFS Getting Started Guide. This will allow the cluster ease of access to the training dataset.

Enable EFS CSI Driver on EKS

To set up EFS on EKS, refer to EKS EFS CSI Driver Installation Guide.

Launch EFS on EKS

1. Create storage.yaml file. This file creates a file system that can be accessed from multiple pods at once. For more information on how Persistent Volume functions, refer to Kubernetes Persistent Volume Guide.

   ---
   kind: StorageClass
   apiVersion: storage.k8s.io/v1
   metadata:
     name: efs-sc
   provisioner: efs.csi.aws.com
   ---
   apiVersion: v1
   kind: PersistentVolume
   metadata:
     name: efs-pv
   spec:
     capacity:
       storage: 150Gi
     volumeMode: Filesystem
     accessModes:
       - ReadWriteMany
     persistentVolumeReclaimPolicy: Retain
     storageClassName: efs-sc
     csi:
       driver: efs.csi.aws.com
       volumeHandle: fs-05af1ea276164472d
   ---
   apiVersion: v1
   kind: PersistentVolumeClaim
   metadata:
     name: efs-claim
   spec:
     accessModes:
       - ReadWriteMany
     storageClassName: efs-sc
     resources:
       requests:
         storage: 150Gi
   ---
   

2. Update the Elastic File System ID using the volumeHandle parameter to run the desired configuration.

3. Create the Persistent Volume:

   kubectl apply -f storage.yaml
   

Launch ResNet Training

1. Create resnet.yaml file with the content below. Check the model code README for details on how to run ResNet Multi-Server Training:

   apiVersion: kubeflow.org/v2beta1
   kind: MPIJob
   metadata:
     name: resnet50-2wkr
   spec:
     slotsPerWorker: 8
     runPolicy:
       cleanPodPolicy: Running
     mpiReplicaSpecs:
       Launcher:
         replicas: 1
         template:
           spec:
             imagePullSecrets:
               - name: private-registry
             terminationGracePeriodSeconds: 0
             containers:
               - image: <Custom Docker Image>
                 imagePullPolicy: Always
                 name: resnet-launcher
                 command: ["/bin/bash", "-c"]
                 args:
                   - >-
                     /usr/bin/ssh-keygen -A;
                     /usr/sbin/sshd;
   
                     HOSTSFILE=$OMPI_MCA_orte_default_hostfile;
                     MASTER_ADDR="$(head -n 1 $HOSTSFILE | sed -n s/[[:space:]]slots.*//p)";
   
                     NUM_NODES=$(wc -l < $HOSTSFILE);
                     CARDS_PER_NODE=8;
                     N_CARDS=$((NUM_NODES*CARDS_PER_NODE));
   
                     MODEL_PATH=/Model-References/PyTorch/computer_vision/classification/torchvision/;
                     DATA_DIR=/data/pytorch/imagenet/ILSVRC2012;
   
                     CMD="python $MODEL_PATH/train.py \
                       --data-path=${DATA_DIR} \
                       --model=resnet50 \
                       --device=hpu \
                       --batch-size=256 \
                       --epochs=90 \
                       --print-freq=1 \
                       --output-dir=. \
                       --seed=123 \
                       --autocast \
                       --custom-lr-values 0.275 0.45 0.625 0.8 0.08 0.008 0.0008 \
                       --custom-lr-milestones 1 2 3 4 30 60 80 \
                       --deterministic \
                       --dl-time-exclude=False";
   
                       mpirun -np ${N_CARDS} \
                           --allow-run-as-root \
                           --prefix $MPI_ROOT \
                           --map-by ppr:4:socket:PE=6 \
                           --bind-to core \
                           -x PYTHONPATH="/usr/lib/habanalabs:/Model-References" \
                           -x RDMAV_FORK_SAFE=1 \
                           -x FI_EFA_USE_DEVICE_RDMA=1 \
                           -x MASTER_ADDR=$MASTER_ADDR \
                           $CMD;
                 resources:
                   requests:
                     cpu: "100m"
                 volumeMounts:
                   - mountPath: /data
                     name: persistent-storage
             volumes:
             - name: persistent-storage
               persistentVolumeClaim:
                 claimName: efs-claim
       Worker:
         replicas: 2
         template:
           spec:
             imagePullSecrets:
               - name: private-registry
             hostIPC: true
             containers:
               - image: <Custom Docker Image>
                 name: resnet-worker
                 resources:
                   requests:
                     habana.ai/gaudi: 8
                     hugepages-2Mi: 42000Mi
                     cpu: 90
                     vpc.amazonaws.com/efa: 4
                   limits:
                     habana.ai/gaudi: 8
                     hugepages-2Mi: 42000Mi
                     cpu: 90
                     vpc.amazonaws.com/efa: 4
                 volumeMounts:
                   - mountPath: /data
                     name: persistent-storage
                 command: ["/bin/bash", "-c"]
                 args:
                   - >-
                     /usr/bin/ssh-keygen -A;
                     /usr/sbin/sshd;
                     sleep 365d;
             volumes:
             - name: persistent-storage
               persistentVolumeClaim:
                 claimName: efs-claim
   

2. Update the parameters listed below to run the desired configuration:

   Parameter	Description
   <Custom Docker Image>	Image with ResNet dependencies installed
   replicas: 2	Number of DL1s for training.

3. Run the job:

   kubectl apply -f resnet.yaml
   

4. Check the job status:

   kubectl get pods -A
   

5. Retrieve the name of the created pod and run the following command to see the results:

   kubectl logs <pod-name>