From 0f129b28e522b9da7415a092b76ae48a202f15e6 Mon Sep 17 00:00:00 2001 From: Chen Cui Date: Fri, 8 May 2026 13:44:33 -0700 Subject: [PATCH] [ci, test] chore: Split L0 converter launch script into conversion and generation Split L0_Launch_converter.sh into two scripts to reduce per-job runtime and isolate the heavier generation tests. Also drops an existing duplication where test_hf_fsdp_conversion.py was running both in the catch-all converter script and in the dedicated L0_Launch_converter_fsdp.sh. - L0_Launch_converter.sh -> test_checkpoint_conversion.py + test_multi_gpu_conversion.py - L0_Launch_converter_generate.sh -> test_generate_from_hf.py + test_generate_vlm_from_hf.py (new) - L0_Launch_converter_fsdp.sh -> test_hf_fsdp_conversion.py (unchanged) The CI matrix is generated dynamically by scanning the launch_scripts directory, so no workflow changes are required. Co-Authored-By: Claude Opus 4.7 (1M context) Signed-off-by: Chen Cui --- .../h100/active/L0_Launch_converter.sh | 3 ++- .../active/L0_Launch_converter_generate.sh | 24 +++++++++++++++++++ 2 files changed, 26 insertions(+), 1 deletion(-) create mode 100755 tests/functional_tests/launch_scripts/h100/active/L0_Launch_converter_generate.sh diff --git a/tests/functional_tests/launch_scripts/h100/active/L0_Launch_converter.sh b/tests/functional_tests/launch_scripts/h100/active/L0_Launch_converter.sh index 886cfef1eb..4bf2b3c53f 100755 --- a/tests/functional_tests/launch_scripts/h100/active/L0_Launch_converter.sh +++ b/tests/functional_tests/launch_scripts/h100/active/L0_Launch_converter.sh @@ -19,5 +19,6 @@ export CUDA_VISIBLE_DEVICES="0,1" uv run coverage run --data-file=/opt/Megatron-Bridge/.coverage --source=/opt/Megatron-Bridge/ --parallel-mode -m pytest \ -o log_cli=true -o log_cli_level=INFO -v -s -x -m "not pleasefixme" --tb=short -rA \ - tests/functional_tests/test_groups/converter + tests/functional_tests/test_groups/converter/test_checkpoint_conversion.py \ + tests/functional_tests/test_groups/converter/test_multi_gpu_conversion.py coverage combine -q diff --git a/tests/functional_tests/launch_scripts/h100/active/L0_Launch_converter_generate.sh b/tests/functional_tests/launch_scripts/h100/active/L0_Launch_converter_generate.sh new file mode 100755 index 0000000000..757b6aba76 --- /dev/null +++ b/tests/functional_tests/launch_scripts/h100/active/L0_Launch_converter_generate.sh @@ -0,0 +1,24 @@ +# Copyright (c) 2026, NVIDIA CORPORATION. All rights reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +#!/bin/bash +set -xeuo pipefail # Exit immediately if a command exits with a non-zero status + +export CUDA_VISIBLE_DEVICES="0,1" + +uv run coverage run --data-file=/opt/Megatron-Bridge/.coverage --source=/opt/Megatron-Bridge/ --parallel-mode -m pytest \ + -o log_cli=true -o log_cli_level=INFO -v -s -x -m "not pleasefixme" --tb=short -rA \ + tests/functional_tests/test_groups/converter/test_generate_from_hf.py \ + tests/functional_tests/test_groups/converter/test_generate_vlm_from_hf.py +coverage combine -q