skip layers if already fused (#322)

Author: dsikka

src/compressed_tensors/quantization/lifecycle/initialize.py

@@ -304,6 +304,9 @@ def _valid_fp4_quant(layer_list: List[torch.nn.Linear]):
     ):
 
         if _is_attention_module(submodule):
+            # already fused/treated as one layer
+            if hasattr(submodule, "qkv_proj"):
+                continue
 
             if not _valid_fp4_quant(
                 [submodule.q_proj, submodule.v_proj, submodule.k_proj]