diff --git a/crates/llm-local/src/llama.rs b/crates/llm-local/src/llama.rs
index 84f218c8e..66976ab3c 100644
--- a/crates/llm-local/src/llama.rs
+++ b/crates/llm-local/src/llama.rs
@@ -15,7 +15,7 @@ use tokenizers::Tokenizer;
 const TOKENIZER_FILENAME: &str = "tokenizer.json";
 const CONFIG_FILENAME: &str = "config.json";
 const EOS_TOKEN: &str = "</s>";
-const MODEL_SAFETENSORS_INDEX: &str = "model.safetensors.index.json";
+const MODEL_SAFETENSORS_INDEX_FILE: &str = "model.safetensors.index.json";
 
 pub fn auto_device() -> Result<Device> {
     if utils::cuda_is_available() {
@@ -51,7 +51,7 @@ impl LlamaModels {
         let config = config.into_config(false);
         let cache = llama::Cache::new(true, dtype, &config, &device)?;
 
-        let safetensor_files = load_safetensors(&model_dir, MODEL_SAFETENSORS_INDEX)?;
+        let safetensor_files = load_safetensors(model_dir, MODEL_SAFETENSORS_INDEX_FILE)?;
 
         // TODO: Check if there is a safe way to load the model from the file
         let vb = unsafe { VarBuilder::from_mmaped_safetensors(&safetensor_files, dtype, &device)? };