decoding_visualizer

Running on Zero

multimodalart HF Staff commited on May 23

Commit

b7439a7

verified ·

1 Parent(s): c05aa51

Update app.py

Files changed (1) hide show

app.py CHANGED Viewed

@@ -430,7 +430,7 @@ def get_beam_search_html(
 ):
     input_with_template=f"<|im_start|>system /n You are a helpful chatbot.<|im_end|><|im_start|>{input_text}"
-    inputs = tokenizer([input_with_template], return_tensors="pt")
     outputs = model.generate(
         **inputs,

 ):
     input_with_template=f"<|im_start|>system /n You are a helpful chatbot.<|im_end|><|im_start|>{input_text}"
+    inputs = tokenizer([input_with_template], return_tensors="pt").to("cuda")
     outputs = model.generate(
         **inputs,