Recovered skipped w8a8 compression related tests (#1785)

SUMMARY:
Recovered skipped w8a8 compression/decompression tests now that
transformer side of code is merged. Added memory collection between test
instances.

TEST PLAN:
tested locally on transformers 4.56.0.dev0 they passed.

---------

Signed-off-by: shanjiaz <zsjwpianpian@gmail.com>

Author: shanjiaz

tests/llmcompressor/transformers/compression/decompression_configs_skipped/w8a8.yaml renamed to tests/llmcompressor/transformers/compression/decompression_configs/w8a8.yaml

@@ -84,9 +84,17 @@ def test_compressed_matches_decompressed(self):
     def tearDownClass(cls):
         if os.path.isdir(cls.test_dir):
             shutil.rmtree(cls.test_dir)
+
+        if hasattr(cls, "decompressed_model") and cls.decompressed_model is not None:
+            cls.decompressed_model.cpu()
+        if hasattr(cls, "uncompressed_model") and cls.uncompressed_model is not None:
+            cls.uncompressed_model.cpu()
         del cls.decompressed_model
         del cls.uncompressed_model
+        del cls.tokenizer
+
         torch.cuda.empty_cache()
+        torch.cuda.synchronize()
 
 
 @requires_gpu
@@ -168,6 +176,14 @@ def test_compressed_matches_decompressed__hf_quantizer(self):
     def tearDownClass(cls):
         if os.path.isdir(cls.test_dir):
             shutil.rmtree(cls.test_dir)
+
+        if hasattr(cls, "decompressed_model") and cls.decompressed_model is not None:
+            cls.decompressed_model.cpu()
+        if hasattr(cls, "compressed_model") and cls.compressed_model is not None:
+            cls.compressed_model.cpu()
         del cls.decompressed_model
         del cls.compressed_model
+        del cls.tokenizer
+
         torch.cuda.empty_cache()
+        torch.cuda.synchronize()