From 5973d6c75338ee0cf2cc01ecaba6283fc6b4cf05 Mon Sep 17 00:00:00 2001 From: Xin Qiu Date: Wed, 8 May 2024 14:27:51 +0800 Subject: [PATCH] make gemma's output better (#10943) --- .../GPU/HF-Transformers-AutoModels/Model/gemma/generate.py | 1 + 1 file changed, 1 insertion(+) diff --git a/python/llm/example/GPU/HF-Transformers-AutoModels/Model/gemma/generate.py b/python/llm/example/GPU/HF-Transformers-AutoModels/Model/gemma/generate.py index bbe4f68b..2ca0ab90 100644 --- a/python/llm/example/GPU/HF-Transformers-AutoModels/Model/gemma/generate.py +++ b/python/llm/example/GPU/HF-Transformers-AutoModels/Model/gemma/generate.py @@ -48,6 +48,7 @@ if __name__ == '__main__': load_in_4bit=True, optimize_model=True, trust_remote_code=True, + mixed_precision=True, use_cache=True) model = model.to('xpu')