debug

app.py

@@ -35,7 +35,7 @@ def generate_responses(model, tokenizer, prompts):
             **model_inputs,
             do_sample=False,
             temperature=0,
-            repetition_penalty=1.0,
+            repetition_penalty=1.1,
             max_new_tokens=512,
         )
     prompt_lengths = model_inputs["attention_mask"].sum(dim=1)
@@ -69,15 +69,16 @@ def generate(submission: list[dict[str, str]], team_id: str) -> list[dict[str, s
     prompts = [s["prompt"] for s in submission]
 
     responses = generate_responses(chat_model, chat_tokenizer, prompts)
+    print(responses)
+
     scores = classify_pairs(moderator_model, moderator_tokenizer, prompts, responses)
-    
+    print(scores)
+
     outputs = [
         {"id": id, "prompt": prompt, "response": response, "score": score, "model": chat_model_name, "team_id": team_id}
         for id, prompt, response, score in zip(ids, prompts, responses, scores)
     ]
 
-    print(outputs)
-    
     return outputs
 
 
@@ -85,5 +86,5 @@ with gr.Blocks() as demo:
     gr.Markdown("Welcome")
     gr.api(generate, api_name="scores", batch=False)
 
-demo.queue()
+# demo.queue()
 demo.launch()