hopefully fixed

app.py

@@ -1,5 +1,6 @@
 import gradio as gr
 from transformers import GPT2LMHeadModel, GPT2Tokenizer
+import torch
 
 # Load model and tokenizer from Hugging Face Hub
 model_name = "Electricarchmage/cookbookgpt"
@@ -16,26 +17,30 @@ def respond(
     top_p,
 ):
     # Preparing the messages for context (the history and the new message)
-    input_text = system_message + "\n"
+    messages = [{"role": "system", "content": system_message}]
     
+    # Convert history to the required format with 'role' and 'content'
     for val in history:
         if val[0]:
-            input_text += f"User: {val[0]}\n"
+            messages.append({"role": "user", "content": val[0]})
         if val[1]:
-            input_text += f"Assistant: {val[1]}\n"
-
-    input_text += f"User: {message}\nAssistant:"
+            messages.append({"role": "assistant", "content": val[1]})
+    
+    messages.append({"role": "user", "content": message})
 
     # Tokenize the input and generate a response
-    inputs = tokenizer(input_text, return_tensors="pt", max_length=1024, truncation=True)
+    inputs = tokenizer([msg["content"] for msg in messages], return_tensors="pt", padding=True, truncation=True)
+    attention_mask = inputs.get('attention_mask', torch.ones_like(inputs['input_ids']))  # Default to ones if not provided
 
     # Generate output tokens
     output = model.generate(
         inputs["input_ids"],
+        attention_mask=attention_mask,
         max_length=max_tokens + len(inputs["input_ids"][0]),
         temperature=temperature,
         top_p=top_p,
         num_return_sequences=1,
+        do_sample=True,  # Enable sampling for more dynamic responses
         no_repeat_ngram_size=2,
     )