latest fixes needed for GA in latest transformers

This commit is contained in:
Wing Lian
2025-01-13 13:36:47 -05:00
parent 49b5501fc2
commit 5b5ba49c46
5 changed files with 72 additions and 17 deletions

View File

@@ -102,9 +102,5 @@ class TestMixtral(unittest.TestCase):
cli_args = TrainerCliArgs()
dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
model, _ = train(cfg=cfg, dataset_meta=dataset_meta)
assert (
"MixtralFlashAttention2"
in model.model.layers[0].self_attn.__class__.__name__
)
train(cfg=cfg, dataset_meta=dataset_meta)
check_model_output_exists(temp_dir, cfg)