Skip tests properly (#31308)

* Skip tests properly

* [test_all]

* Add 'reason' as kwarg for skipTest

* [test_all] Fix up

* [test_all]

tests/quantization/autoawq/test_awq.py

@@ -450,7 +450,7 @@ class AwqFusedTest(unittest.TestCase):
         outputs = model.generate(**inputs, max_new_tokens=12)
         self.assertEqual(tokenizer.decode(outputs[0], skip_special_tokens=True), self.EXPECTED_GENERATION_CUSTOM_MODEL)
 
-    @unittest.skip("Not enough GPU memory on CI runners")
+    @unittest.skip(reason="Not enough GPU memory on CI runners")
     @require_torch_multi_gpu
     def test_generation_mixtral_fused(self):
         """

tests/quantization/bnb/test_4bit.py

@@ -496,7 +496,7 @@ class Bnb4BitTestTraining(Base4bitTest):
 
     def test_training(self):
         if version.parse(importlib.metadata.version("bitsandbytes")) < version.parse("0.37.0"):
-            return
+            self.skipTest(reason="This test requires bitsandbytes >= 0.37.0")
 
         # Step 1: freeze all parameters
         model = AutoModelForCausalLM.from_pretrained(self.model_name, load_in_4bit=True)

tests/quantization/bnb/test_mixed_int8.py

@@ -826,7 +826,7 @@ class MixedInt8TestTraining(BaseMixedInt8Test):
 
     def test_training(self):
         if version.parse(importlib.metadata.version("bitsandbytes")) < version.parse("0.37.0"):
-            return
+            self.skipTest(reason="This test requires bitsandbytes>=0.37.0")
 
         # Step 1: freeze all parameters
         model = AutoModelForCausalLM.from_pretrained(self.model_name, load_in_8bit=True)

tests/quantization/quanto_integration/test_quanto.py

@@ -332,20 +332,23 @@ class QuantoQuantizationOffloadTest(QuantoQuantizationTest):
         "lm_head": 0,
     }
 
-    # the execution device is a gpu
+    @unittest.skip(reason="The execution device is a gpu")
     def test_generate_quality_cpu(self):
         pass
 
-    # we can't save offloaded values
+    @unittest.skip(reason="We can't save offloaded values")
     def test_serialization_bin(self):
         pass
 
+    @unittest.skip
     def test_serialization_safetensors(self):
         pass
 
+    @unittest.skip
     def test_compare_with_quanto(self):
         pass
 
+    @unittest.skip
     def test_load_from_quanto_saved(self):
         pass
 
@@ -370,7 +373,7 @@ class QuantoQuantizationOffloadTest(QuantoQuantizationTest):
             )
 
 
-@unittest.skip("Skipping test class because serialization is not supported yet")
+@unittest.skip(reason="Skipping test class because serialization is not supported yet")
 class QuantoQuantizationSerializationTest(QuantoQuantizationTest):
     """
     Perform the same tests as in QuantoQuantizationTest but with a serialized model.
@@ -403,7 +406,7 @@ class QuantoQuantizationSerializationTest(QuantoQuantizationTest):
         )
 
 
-@unittest.skip("Skipping test class because serialization is not supported yet")
+@unittest.skip(reason="Skipping test class because serialization is not supported yet")
 class QuantoQuantizationSerializationCudaTest(QuantoQuantizationTest):
     """
     Perform the same tests as in QuantoQuantizationTest but with model on cuda
@@ -422,7 +425,7 @@ class QuantoQuantizationQBitsTensorOffloadTest(QuantoQuantizationOffloadTest):
     weights = "int4"
 
 
-@unittest.skip("Skipping test class because serialization is not supported yet")
+@unittest.skip(reason="Skipping test class because serialization is not supported yet")
 class QuantoQuantizationQBitsTensorSerializationTest(QuantoQuantizationSerializationTest):
     EXPECTED_OUTPUTS = "Hello my name is Nils, I am a student of the University"
     weights = "int4"