Upload app.py with huggingface_hub

app.py

@@ -31,16 +31,16 @@ def respond(message, history):
             messages.append({"role": "assistant", "content": bot_msg})
     messages.append({"role": "user", "content": message})
     
-    input_ids = tokenizer.apply_chat_template(
-        messages, return_tensors="pt", add_generation_prompt=True
+    tokenized = tokenizer.apply_chat_template(
+        messages, tokenize=True, add_generation_prompt=True, return_tensors="pt", return_dict=True
     )
-    if isinstance(input_ids, dict):
-        input_ids = input_ids["input_ids"]
-    input_ids = input_ids.to(model.device)
+    input_ids = tokenized["input_ids"].to(model.device)
+    attention_mask = tokenized["attention_mask"].to(model.device)
     
     with torch.no_grad():
         outputs = model.generate(
-            input_ids,
+            input_ids=input_ids,
+            attention_mask=attention_mask,
             max_new_tokens=300,
             temperature=0.7,
             top_p=0.9,