improve interface and ZeroGPU logic

app.py

@@ -7,6 +7,8 @@ import spaces
 model_path = "nvidia/Mistral-NeMo-Minitron-8B-Instruct"
 tokenizer = AutoTokenizer.from_pretrained(model_path)
 model = AutoModelForCausalLM.from_pretrained(model_path)
+if tokenizer.pad_token_id is None:
+    tokenizer.pad_token_id = tokenizer.eos_token_id
 
 pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
 
@@ -36,15 +38,16 @@ def generate_response(message, history, system_message, max_tokens, temperature,
     if use_pipeline:
         response = pipe(full_prompt, max_new_tokens=max_tokens, temperature=temperature, top_p=top_p, do_sample=True)[0]['generated_text']
     else:
-        inputs = tokenizer(full_prompt, return_tensors="pt").to(model.device)
-        
+        inputs = tokenizer(full_prompt, return_tensors="pt", padding=True, truncation=True)
+        inputs = {k: v.to(model.device) for k, v in inputs.items()}        
         with torch.no_grad():
             output_ids = model.generate(
                 inputs.input_ids,
                 max_new_tokens=max_tokens,
                 temperature=temperature,
                 top_p=top_p,
-                do_sample=True
+                do_sample=True,
+                attention_mask=inputs['attention_mask']
             )
         
         response = tokenizer.decode(output_ids[0], skip_special_tokens=True)