fix kv cache tests

Signed-off-by: Kyle Sayers <kylesayrs@gmail.com>

Author: kylesayrs

.github/workflows/test-check-transformers.yaml

@@ -98,4 +98,4 @@ jobs:
       - name: Running KV Cache Tests
         if: (success() || failure()) && steps.install.outcome == 'success'
         run: |
-          pytest -v tests/llmcompressor/transformers/kv_cache -k "not test_kv_cache_gptq_model_state_dict_attr"
+          pytest -v tests/llmcompressor/transformers/kv_cache

tests/e2e/vLLM/recipes/kv_cache/gptq.yaml

@@ -1,8 +1,5 @@
 quant_stage:
   quant_modifiers:
-    QuantizationModifier:
-      kv_cache_scheme:
-        {num_bits: 8, type: float, symmetric: true, strategy: tensor}
     GPTQModifier:
       sequential_update: false
       ignore: ["lm_head"]
@@ -15,3 +12,5 @@ quant_stage:
                   strategy: "channel"
                   actorder: False
               targets: ["Linear"]
+      kv_cache_scheme:
+        {num_bits: 8, type: float, symmetric: true, strategy: tensor}

tests/llmcompressor/transformers/kv_cache/test_kv_cache.py

@@ -214,13 +214,6 @@ def test_kv_cache_gptq_model_state_dict_attr(kv_cache_fixture, tmp_path):
     recipe = """
     quant_stage:
         quant_modifiers:
-            QuantizationModifier:
-                kv_cache_scheme:
-                    num_bits: {num_bits}
-                    type: {_type}
-                    strategy: {strategy}
-                    dynamic: {dynamic}
-                    symmetric: {symmetric}
             GPTQModifier:
                 ignore: ["lm_head"]
                 config_groups:
@@ -232,6 +225,12 @@ def test_kv_cache_gptq_model_state_dict_attr(kv_cache_fixture, tmp_path):
                             strategy: "channel"
                             actorder: False
                         targets: ["Linear"]
+                kv_cache_scheme:
+                    num_bits: {num_bits}
+                    type: {_type}
+                    strategy: {strategy}
+                    dynamic: {dynamic}
+                    symmetric: {symmetric}
     """
 
     output_dir, _ = next(kv_cache_fixture(recipe, tmp_path))