diff --git a/backends/exllamav2/model.py b/backends/exllamav2/model.py
index 65689f4..f7314a7 100644
--- a/backends/exllamav2/model.py
+++ b/backends/exllamav2/model.py
@@ -188,7 +188,7 @@ class ExllamaV2Container(BaseModelContainer):
         self.cache_mode = unwrap(kwargs.get("cache_mode"), "FP16")
 
         # Catch exllamav3 cache_mode
-        if not self.cache_mode.startswith("Q"):
+        if not self.cache.mode == "FP16" or not self.cache_mode.startswith("Q"):
             logger.warning(
                 f"Provided cache mode '{self.cache_mode}' is not a "
                 "valid choice for exllamav2, please check your settings. "
@@ -402,7 +402,10 @@ class ExllamaV2Container(BaseModelContainer):
             self.draft_cache_mode = unwrap(draft_args.get("draft_cache_mode"), "FP16")
 
             # Catch exllamav3 draft_cache_mode
-            if not self.draft_cache_mode.startswith("Q"):
+            if (
+                not self.draft_cache_mode == "FP16"
+                or not self.draft_cache_mode.startswith("Q")
+            ):
                 logger.warning(
                     f"Provided draft cache mode '{self.draft_cache_mode}' is not a "
                     "valid choice for exllamav2, please check your settings. "