shenzhi-wang commited on
Commit
733ec1d
·
1 Parent(s): df32955

[fix] load in 16 bit

Browse files
Files changed (1) hide show
  1. app.py +2 -2
app.py CHANGED
@@ -45,7 +45,7 @@ CSS = """
45
 
46
 
47
  tokenizer = AutoTokenizer.from_pretrained("shenzhi-wang/Gemma-2-27B-Chinese-Chat")
48
- model = AutoModelForCausalLM.from_pretrained("shenzhi-wang/Gemma-2-27B-Chinese-Chat", device_map="auto")
49
 
50
 
51
  @spaces.GPU
@@ -127,4 +127,4 @@ with gr.Blocks(css=CSS) as demo:
127
 
128
 
129
  if __name__ == "__main__":
130
- demo.launch()
 
45
 
46
 
47
  tokenizer = AutoTokenizer.from_pretrained("shenzhi-wang/Gemma-2-27B-Chinese-Chat")
48
+ model = AutoModelForCausalLM.from_pretrained("shenzhi-wang/Gemma-2-27B-Chinese-Chat", device_map="auto", torch_dtype="auto")
49
 
50
 
51
  @spaces.GPU
 
127
 
128
 
129
  if __name__ == "__main__":
130
+ demo.launch()