diff --git a/test_runner.py b/test_runner.py index ddab7f0b..a1d3bf22 100755 --- a/test_runner.py +++ b/test_runner.py @@ -46,6 +46,21 @@ def build_test_list(): """ integration_tests_flavors = defaultdict(list) integration_tests_flavors["debug_model.toml"] = [ + OverrideDefinitions( + [ + [ + "--checkpoint.enable_checkpoint", + "--experimental.pipeline_parallel_degree 4", + "--experimental.pipeline_parallel_split_points layers.1,layers.2,layers.3,layers.4,layers.5,layers.6,layers.7", + "--experimental.pipeline_parallel_schedule flexible_interleaved_1f1b", + "--model.norm_type rmsnorm", # fused_rmsnorm throws cuda context error with pp + ], + ], + "PP looped flexible 1f1b test", + "pp_looped_flexible_1f1b", + requires_seed_checkpoint=True, + ngpu=4, + ), OverrideDefinitions( [ [ @@ -254,21 +269,6 @@ def build_test_list(): requires_seed_checkpoint=True, ngpu=4, ), - OverrideDefinitions( - [ - [ - "--checkpoint.enable_checkpoint", - "--experimental.pipeline_parallel_degree 4", - "--experimental.pipeline_parallel_split_points layers.1,layers.2,layers.3,layers.4,layers.5,layers.6,layers.7", - "--experimental.pipeline_parallel_schedule flexible_interleaved_1f1b", - "--model.norm_type rmsnorm", # fused_rmsnorm throws cuda context error with pp - ], - ], - "PP looped flexible 1f1b test", - "pp_looped_flexible_1f1b", - requires_seed_checkpoint=True, - ngpu=4, - ), OverrideDefinitions( [ [