3-llama3-finetune-trn1-rayjob-submit-finetuning-job.yaml

# ----------------------------------------------------------------------------
# RayJob: 3-llama3-finetune-trn1-rayjob-submit-finetuning-job
#
# Description:
# This RayJob is responsible for the main finetune step of the Llama3 model. It runs a
# Python script (`ray_train_llama3.py`) to perform the fine-tuning using AWS Neuron devices.
# This step is critical for training the language model with the prepared dataset.

# Usage:
# Apply this configuration to your Kubernetes cluster using `kubectl apply -f 3-llama3-finetune-trn1-rayjob-submit-finetuning-job.yaml`.
# Ensure that the Ray cluster (`kuberay-trn1`) is running and accessible in the specified namespace.
# ----------------------------------------------------------------------------

---
apiVersion: ray.io/v1
kind: RayJob
metadata:
  name: 3-llama3-finetune-trn1-rayjob-submit-finetuning-job
spec:
  submissionMode: K8sJobMode
  entrypoint: "NEURON_NUM_DEVICES=32 bash tp_zero1_llama3_8b_hf_finetune_ptl.sh"
  runtimeEnvYAML: |
    working_dir: /llama3_finetune
  clusterSelector:
    ray.io/cluster: kuberay-trn1
    rayClusterNamespace: default  # Replace with the namespace where your RayCluster is deployed
  shutdownAfterJobFinishes: true
  ttlSecondsAfterFinished: 60  # Time to live for the pod after completion (in seconds)