only support fp8 e4m3 on nmz

vllm/v1/attention/backends/flash_attn.py

@@ -142,7 +142,10 @@ class FlashAttentionBackend(AttentionBackend):
     @staticmethod
     def get_fp8_dtype_for_flashattn(kv_cache_dtype: str) -> torch.dtype:
         if kv_cache_dtype in ("fp8", "fp8_e4m3"):
-            return torch.float8_e4m3fn
+            if torch.cuda.get_device_properties("cuda").gcnArchName.split(':')[0] == "gfx938":
+                return torch.float8_e4m3fn
+            else:
+                raise ValueError(f"Unsupported FP8 dtype: {kv_cache_dtype}")
         elif kv_cache_dtype in ("fp8_e5m2"):
             return torch.float8_e5m2
         else: