Update app.py

app.py

@@ -8,6 +8,10 @@ tokenizer = AutoTokenizer.from_pretrained("gpt2")
 model = AutoModelForCausalLM.from_pretrained("gpt2")
 
 print("Loading finished.")
+n_steps=12
+n_beams=1
+length_penalty=1
+num_return_sequences=3
 
 print(f"Is CUDA available: {torch.cuda.is_available()}")
 # True
@@ -436,9 +440,6 @@ def get_beam_search_html(
         do_sample=False,
     )
     markdown = "The conclusive sequences are the ones that end in an `<|endoftext|>` token or at the end of generation."
-    markdown += "\n\nThey are ranked by their scores, as given by the formula `score = cumulative_score / (output_length ** length_penalty)`.\n\n"
-    markdown += "Only the top `num_beams` scoring sequences are returned: in the tree they are highlighted in **<span style='color:var(--secondary-500)!important'>blue</span>**."
-    markdown += " The non-selected sequences are also shown in the tree, highlighted in **<span style='color:var(--primary-500)!important'>yellow</span>**."
     markdown += "\n#### <span style='color:var(--secondary-500)!important'>Output sequences:</span>"
     # Sequences are padded anyway so you can batch decode them
     decoded_sequences = tokenizer.batch_decode(outputs.sequences)
@@ -484,16 +485,13 @@ with gr.Blocks(
         value="Hugging Face is",
     )
 
-    n_steps=12
-    n_beams=1
-    length_penalty=1
-    num_return_sequences=3
+   
     button = gr.Button()
     out_html = gr.Markdown()
     out_markdown = gr.Markdown()
     button.click(
         get_beam_search_html,
-        inputs=[text, n_steps, n_beams, length_penalty, num_return_sequences],
+        inputs=[text],
         outputs=[out_html, out_markdown],
     )