Update the pipeline to RHEL AI 1.5

[mprahl]

Description

Update the pipeline to RHEL AI 1.5.

How Has This Been Tested?

I uploaded the pipeline and ran it with the following parameters:

    "runtime_config": {
        "parameters": {
            "eval_gpu_identifier": "nvidia.com/gpu",
            "eval_judge_secret": "judge-eval-secret",
            "final_eval_batch_size": "auto",
            "final_eval_few_shots": 5,
            "final_eval_max_workers": "auto",
            "final_eval_merge_system_user_message": false,
            "k8s_storage_class_name": "nfs-csi",
            "k8s_storage_size": "100Gi",
            "mt_bench_max_workers": "auto",
            "mt_bench_merge_system_user_message": false,
            "output_model_name": "mprahl-test",
            "output_model_registry_api_url": "https://cluster.gcp.rh-ods.com",
            "output_model_version": "rhelai-1.5-full",
            "output_modelcar_base_image": "registry.access.redhat.com/ubi9-micro:latest",
            "output_oci_registry_secret": "output-oci-registry-secret",
            "sdg_base_model": "oci://registry.redhat.io/rhelai1/modelcar-granite-7b-starter:1.4",
            "sdg_batch_size": 128,
            "sdg_max_batch_len": 5000,
            "sdg_num_workers": 2,
            "sdg_pipeline": "/usr/share/instructlab/sdg/pipelines/agentic",
            "sdg_repo_branch": "main",
            "sdg_repo_pr": 0,
            "sdg_repo_secret": "taxonomy-repo-secret",
            "sdg_repo_url": "https://github.com/mprahl/taxonomy.git",
            "sdg_sample_size": 0.0002,
            "sdg_scale_factor": 2,
            "sdg_teacher_secret": "teacher-sdg-secret",
            "train_cpu_per_worker": "4",
            "train_effective_batch_size_phase_1": 128,
            "train_effective_batch_size_phase_2": 3840,
            "train_gpu_identifier": "nvidia.com/gpu",
            "train_gpu_per_worker": 2,
            "train_learning_rate_phase_1": 0.00002,
            "train_learning_rate_phase_2": 0.000006,
            "train_max_batch_len": 5000,
            "train_memory_per_worker": "250Gi",
            "train_node_selectors": {},
            "train_num_epochs_phase_1": 1,
            "train_num_epochs_phase_2": 1,
            "train_num_warmup_steps_phase_1": 800,
            "train_num_warmup_steps_phase_2": 800,
            "train_num_workers": 2,
            "train_save_samples": 0,
            "train_seed": 42,
            "train_tolerations": []
        }
    }

It passed:

Merge criteria:

• The commits are squashed in a cohesive manner and have meaningful messages.
• Testing instructions have been added in the PR body (for PRs involving changes that are not immediately obvious).
• The developer has manually tested the changes and verified that the changes work

[HumairAK]

/lgtm
/approve