diff --git a/unsloth/models/llama.py b/unsloth/models/llama.py
index 3340b89eae..39880cc5f1 100644
--- a/unsloth/models/llama.py
+++ b/unsloth/models/llama.py
@@ -1890,8 +1890,8 @@ class FastLlamaModel:
             gc.collect()
             torch.cuda.empty_cache()"""
         multiplier = \
-        "(4.5 if getattr(model.config, 'quantization_config', \\{'load_in_4bit' : False\\})['load_in_4bit'] else "\
-        "(8.0 if getattr(model.config, 'quantization_config', \\{'load_in_8bit' : False\\})['load_in_8bit'] else 1.0)"
+            "(4.5 if getattr(model.config, 'quantization_config', \\{'load_in_4bit' : False\\})['load_in_4bit'] else "\
+            "(8.0 if getattr(model.config, 'quantization_config', \\{'load_in_8bit' : False\\})['load_in_8bit'] else 1.0)"
         debug_info = debug_info.replace("multiplier__", multiplier)
         debug_info = debug_info.replace("P__", "get_model_param_count")