Upload folder using huggingface_hub

app.py

@@ -22,7 +22,7 @@ tokenizer = AutoTokenizer.from_pretrained(
 print(f"[startup] Loading model {MODEL_ID}...", flush=True)
 model = AutoModelForCausalLM.from_pretrained(
     MODEL_ID,
-    torch_dtype=torch.float32,
+    dtype=torch.float32,
     device_map="cpu",
     low_cpu_mem_usage=True,
     trust_remote_code=True,
@@ -34,7 +34,7 @@ print("[startup] Model loaded successfully!", flush=True)
 
 def chat_response(
     message: str,
-    history: list[dict[str, str]],
+    history: list[list[str]],
     system_prompt: str,
     temperature: float,
     top_p: float,
@@ -45,8 +45,11 @@ def chat_response(
     if system_prompt.strip():
         messages.append({"role": "system", "content": system_prompt})
 
-    for item in history:
-        messages.append(item)
+    for user_msg, assistant_msg in history:
+        if user_msg:
+            messages.append({"role": "user", "content": user_msg})
+        if assistant_msg:
+            messages.append({"role": "assistant", "content": assistant_msg})
 
     messages.append({"role": "user", "content": message})
 
@@ -93,7 +96,6 @@ def chat_response(
 
 demo = gr.ChatInterface(
     fn=chat_response,
-    type="messages",
     title="K2-Horizon-0.9B Chat Demo",
     description="Interactive demo for [IFM/K2-Horizon-0.9B](https://huggingface.co/IFM/K2-Horizon-0.9B) using PyTorch and Transformers on CPU.",
     additional_inputs=[