Update app.py

app.py

@@ -22,15 +22,15 @@ rm_model = AutoModelForSequenceClassification.from_pretrained('OpenAssistant/rew
 
 @spaces.GPU
 def generate_text(usertitle, content, temperature, max_length, N=3):
-    msg ={
-    'title': usertitle,
-    'content': content
-    }
-    # input_text = f"title: {usertitle}\ncontent: {content}"
+    # msg ={
+    # 'title': usertitle,
+    # 'content': content
+    # }
+    input_text = f"[[[title:]]] {usertitle}\n[[[content:]]]{content}\n\n"
     inputs = tokenizer.apply_chat_template(msg, return_tensors='pt').cuda()
     attention_mask = torch.ones(inputs['input_ids'].shape, dtype=torch.long, device='cuda')
     generated_sequences = model.generate(inputs['input_ids'], attention_mask=attention_mask, temperature=temperature, max_length=max_length, pad_token_id=tokenizer.eos_token_id, num_return_sequences=N, do_sample=True)
-    decoded_sequences = [tokenizer.decode(g, skip_special_tokens=True) for g in generated_sequences]
+    decoded_sequences = [tokenizer.decode(g) for g in generated_sequences]#.strip().split(tokenizer.eos_token)[0]
 
     def score(sequence):
         inputs = rm_tokenizer(sequence, return_tensors='pt', padding=True, truncation=True, max_length=512).to('cuda')