diff --git a/server/text_generation_server/models/__init__.py b/server/text_generation_server/models/__init__.py
index 92f3c51c..4b506532 100644
--- a/server/text_generation_server/models/__init__.py
+++ b/server/text_generation_server/models/__init__.py
@@ -380,6 +380,8 @@ def get_model(
     transformers_model_class = getattr(transformers, modeling_auto.MODEL_FOR_CAUSAL_LM_MAPPING_NAMES[model_type])
     if transformers_model_class.is_backend_compatible():
         transformers_causal_lm_class = TransformersFlashCausalLM
+        if not FLASH_ATTENTION and lora_adapter_ids is not None and len(lora_adapter_ids) > 0:
+            raise ValueError("Transformers backend AutoModel do not support `lora_adapter_ids`.")
 
     quantization_config = config_dict.get("quantization_config", None)
     if quantization_config is None:
diff --git a/server/text_generation_server/models/transformers_flash_causal_lm.py b/server/text_generation_server/models/transformers_flash_causal_lm.py
index 42ec1b3f..98fbf9a2 100644
--- a/server/text_generation_server/models/transformers_flash_causal_lm.py
+++ b/server/text_generation_server/models/transformers_flash_causal_lm.py
@@ -48,7 +48,7 @@ def tgi_flash_attention_forward(
     softmax_scale: Optional[float] = None,
     sliding_window: Optional[int] = None,
     softcap: Optional[float] = None,
-    **kwargs,  # This is needed to "absorb" other args passed by Transformers modeling
+    **_kwargs,  # This is needed to "absorb" other args passed by Transformers modeling
 ):
 
     kv_cache = kv_cache[module.layer_idx]