Merge branch 'skip_incompatibile_cuda_awq_ut' of https://github.com/intel/auto-round into skip_incompatibile_cuda_awq_ut

Author: WeiweiZhang1

auto_round/inference/backend.py

@@ -1025,4 +1025,3 @@ def build_pip_commands(gptq_req, other_reqs):
         log(joined_cmds)
         if logger_level == "error":
             exit(-1)
-

auto_round/testing_utils.py

@@ -15,9 +15,9 @@
 import importlib.util
 import unittest
 from functools import wraps
-from typing import Literal
+from typing import Callable, Literal
+
 import torch
-from typing import Callable
 from transformers.utils.versions import require_version
 
 from auto_round.logger import logger
@@ -218,9 +218,7 @@ def require_vlm_env(test_case):
 
 
 def require_package_version(
-    package: str,
-    version_spec: str,
-    on_fail: Literal["skip", "warn", "error"] = "skip"
+    package: str, version_spec: str, on_fail: Literal["skip", "warn", "error"] = "skip"
 ) -> bool:
     """
     Check if a package satisfies a version requirement.
@@ -264,10 +262,9 @@ def require_package_version_ut(package: str, version_spec: str) -> Callable:
     Returns:
         Callable: A decorator to wrap unittest test methods.
     """
+
     def decorator(test_func: Callable) -> Callable:
         reason = f"Test requires {package}{version_spec}"
-        return unittest.skipUnless(
-            require_package_version(package, version_spec, on_fail="skip"),
-            reason
-        )(test_func)
+        return unittest.skipUnless(require_package_version(package, version_spec, on_fail="skip"), reason)(test_func)
+
     return decorator

test/test_cuda/test_auto_round_format.py

@@ -330,4 +330,3 @@ def test_load_gptq_model_3bits(self):
 
 if __name__ == "__main__":
     unittest.main()
-

test/test_cuda/test_export.py

@@ -405,4 +405,3 @@ def test_nvfp4_llmcompressor_format(self):
 
 if __name__ == "__main__":
     unittest.main()
-

test/test_cuda/test_main_func.py

@@ -83,7 +83,6 @@ def test_backend_awq(self):
         assert accuracy > 0.35
         shutil.rmtree("./saved", ignore_errors=True)
 
-
     @unittest.skipIf(torch.cuda.is_available() is False, "Skipping because no cuda")
     @require_gptqmodel
     def test_fp_layers(self):
@@ -108,7 +107,6 @@ def test_fp_layers(self):
         assert accuracy > 0.35
         shutil.rmtree("./saved", ignore_errors=True)
 
-
     @unittest.skipIf(torch.cuda.is_available() is False, "Skipping because no cuda")
     @require_awq
     @require_package_version_ut("transformers", "<4.57.0")
@@ -134,7 +132,6 @@ def test_fp_layers_awq(self):
         assert accuracy > 0.35
         shutil.rmtree("./saved", ignore_errors=True)
 
-
     @unittest.skipIf(torch.cuda.is_available() is False, "Skipping because no cuda")
     def test_undivided_group_size_tuning(self):
         model_name = "/models/opt-125m"
@@ -185,4 +182,3 @@ def test_autoround_asym(self):  ##need to install false
 
 if __name__ == "__main__":
     unittest.main()
-

test/test_cuda/test_support_vlms.py

@@ -388,4 +388,3 @@ def test_deepseek_vl2(self):
 
 if __name__ == "__main__":
     unittest.main()
-