diff --git a/server/text_generation_server/models/custom_modeling/flash_llama_modeling.py b/server/text_generation_server/models/custom_modeling/flash_llama_modeling.py
index 43c5dfb4..20bab01b 100644
--- a/server/text_generation_server/models/custom_modeling/flash_llama_modeling.py
+++ b/server/text_generation_server/models/custom_modeling/flash_llama_modeling.py
@@ -647,7 +647,7 @@ class FlashLlamaForCausalLM(torch.nn.Module):
             )
         self.model = FlashLlamaModel(prefix, config, weights)
         if config.tie_word_embeddings:
-            suffix = f"model.embed_tokens"
+            suffix = "model.embed_tokens"
         else:
             suffix = "lm_head"