gpt-modelopt.yaml

type: basic
format_version: 1
maintainers: [mcore]
loggers: [stdout]
spec:
  name: '{test_case}_{environment}'
  model: gpt
  build: mcore-pyt-{environment}
  nodes: 1
  gpus: 2
  platforms: dgx_a100
  time_limit:
  n_repeat:
  artifacts:
    /workspace/data/gpt3_data: text/the_pile/shard00
    /workspace/checkpoints/teacher: model/gpt_dummy_pyt/ckpt/24.10.0_bf16_teacher
  script: |-
    ls
    cd /opt/megatron-lm

    ARGUMENTS=(
        "DATA_PATH=/workspace/data/gpt3_data"
        "DATA_CACHE_PATH=/workspace/data/cache"
        "OUTPUT_PATH={assets_dir}"
        "TENSORBOARD_PATH={assets_dir}/tensorboard"
        "CHECKPOINT_SAVE_PATH={artifacts_dir}/checkpoints"
        "CHECKPOINT_LOAD_PATH=/workspace/checkpoints"
        "TRAINING_SCRIPT_PATH=./examples/export/knowledge_distillation/pretrain_gpt_modelopt.py"
        "TRAINING_PARAMS_PATH=./tests/functional_tests/test_cases/{model}/{test_case}/model_config.yaml"
        "GOLDEN_VALUES_PATH=./tests/functional_tests/test_cases/{model}/{test_case}/golden_values_{environment}.json"
    )

    bash ./tests/functional_tests/shell_test_utils/run_ci_test.sh ${{ARGUMENTS[@]}}

products:
  - test_case: [gpt3_nightly_mcore_te_tp2_pp1_modelopt_distill_resume]
    products:
      - environment: [dev, lts]
        scope: [nightly]