PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度
test: skip FP8 tests with transformers 5.16.x (#2241)
Wang, Changb4628f152d6b
diff --git a/test/unit/test_cuda/models/test_fp8_model.py b/test/unit/test_cuda/models/test_fp8_model.pyindex 4c6887fac..a21b51b36 100644--- a/test/unit/test_cuda/models/test_fp8_model.py+++ b/test/unit/test_cuda/models/test_fp8_model.py@@ -19,6 +19,10 @@ from auto_round.utils.weight_handler import (DEVICE_CAPABILITY = torch.cuda.get_device_capability()+ version.parse("5.16.0") <= transformers_version < version.parse("5.17.0"),+ reason="fails with transformers 5.16.x",+)class TestAutoRound:@pytest.fixture(autouse=True)