[https://nvbugs/5478151][fix] Add missing spec for Llama-3.3 70B (NVIDIA#7267)

Signed-off-by: Balaram Buddharaju <[email protected]>
Co-authored-by: Larry <[email protected]>
Signed-off-by: Wangshanshan <[email protected]>

Author: 2 people

tests/integration/defs/accuracy/references/gpqa_diamond.yaml

@@ -8,6 +8,9 @@ meta-llama/Llama-3.3-70B-Instruct:
     accuracy: 48.03
   - quant_algo: FP8
     accuracy: 48.03
+  - quant_algo: FP8
+    kv_cache_quant_algo: FP8
+    accuracy: 48.03
 deepseek-ai/DeepSeek-R1:
   - quant_algo: NVFP4
     accuracy: 70.45