type: basic format_version: 1 maintainers: [mcore] loggers: [stdout] spec: name: '{test_case}_{environment}' model: multimodal-llava build: mcore-pyt-{environment} nodes: 1 gpus: 8 platforms: dgx_a100 time_limit: n_repeat: test_case: scope: script: |- ls cd /opt/megatron-lm ARGUMENTS=( "DATA_PATH='-'" "DATA_CACHE_PATH='-'" "OUTPUT_PATH={assets_dir}" "TENSORBOARD_PATH={assets_dir}/tensorboard" "CHECKPOINT_SAVE_PATH={artifacts_dir}/checkpoints" "CHECKPOINT_LOAD_PATH=/workspace/checkpoints/{name}/checkpoints" "TRAINING_SCRIPT_PATH=pretrain_vlm.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" "N_REPEAT={n_repeat}" ) bash ./tests/functional_tests/shell_test_utils/run_ci_test.sh ${{ARGUMENTS[@]}} products: - test_case: [multimodal_llava_mr_mcore_te_tp1_pp1_dgx_a100_1N8G] products: - environment: [dev] scope: [mr] - environment: [lts] scope: [nightly] - test_case: [multimodal_llava_mr_mcore_te_tp2_pp3_dgx_a100_1N8G] products: - environment: [dev] scope: [mr] - environment: [lts] scope: [nightly] - test_case: [multimodal_llava_mr_mcore_te_tp4_pp1_etp3_dgx_a100_1N7G] products: - environment: [dev] scope: [mr] - environment: [lts] scope: [nightly] - test_case: [multimodal_llava_mr_mcore_te_tp4_pp1_freeze_vit_freeze_lm_dgx_a100_1N8G] products: - environment: [dev] scope: [mr] - environment: [lts] scope: [nightly] - test_case: [multimodal_llava_mr_mcore_te_tp4_pp1_freeze_vit_freeze_lm_dist_opt_dgx_a100_1N8G] products: - environment: [dev] scope: [mr] - environment: [lts] scope: [nightly] - test_case: [multimodal_llava_mr_mcore_te_tp4_pp1_resume_torch_etp3_dgx_a100_1N7G] products: - environment: [dev] scope: [mr] - environment: [lts] scope: [nightly]