diff --git a/unsloth/models/llama.py b/unsloth/models/llama.py index 9746bcb885..3340b89eae 100644 --- a/unsloth/models/llama.py +++ b/unsloth/models/llama.py @@ -1890,9 +1890,9 @@ 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" - debug_info = debug_info.replace("multiplier__", "(" + 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)" + debug_info = debug_info.replace("multiplier__", multiplier) debug_info = debug_info.replace("P__", "get_model_param_count") debug_info = debug_info.split('\n')