Update app.py

app.py

@@ -1,69 +1,99 @@
+import threading
+
 import gradio as gr
-from huggingface_hub import InferenceClient
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
 
+MODEL_ID = "HuggingFaceTB/nanowhale-100m"
 
-def respond(
-    message,
-    history: list[dict[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-    hf_token: gr.OAuthToken,
-):
+print(f"Loading model {MODEL_ID} ...")
+tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
+model = AutoModelForCausalLM.from_pretrained(
+    MODEL_ID,
+    torch_dtype=torch.bfloat16 if torch.cuda.is_available() else torch.float32,
+    device_map="auto",
+)
+model.eval()
+print("Model loaded.")
+
+DEVICE = next(model.parameters()).device
+
+
+def build_prompt(system_message: str, history: list[dict[str, str]], user_message: str) -> str:
     """
-    For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
+    Try to use the tokenizer's built-in chat template.
+    Fall back to a simple newline-delimited format if none exists.
     """
-    client = InferenceClient(token=hf_token.token, model="openai/gpt-oss-20b")
-
     messages = [{"role": "system", "content": system_message}]
-
     messages.extend(history)
+    messages.append({"role": "user", "content": user_message})
 
-    messages.append({"role": "user", "content": message})
+    if tokenizer.chat_template is not None:
+        return tokenizer.apply_chat_template(
+            messages,
+            tokenize=False,
+            add_generation_prompt=True,
+        )
+
+    # Fallback format
+    parts = [f"System: {system_message}\n"]
+    for msg in history:
+        role = "User" if msg["role"] == "user" else "Assistant"
+        parts.append(f"{role}: {msg['content']}\n")
+    parts.append(f"User: {user_message}\nAssistant:")
+    return "".join(parts)
 
-    response = ""
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
+def respond(
+    message: str,
+    history: list[dict[str, str]],
+    system_message: str,
+    max_new_tokens: int,
+    temperature: float,
+    top_p: float,
+):
+    prompt = build_prompt(system_message, history, message)
+
+    inputs = tokenizer(prompt, return_tensors="pt").to(DEVICE)
+    input_len = inputs["input_ids"].shape[-1]
+
+    streamer = TextIteratorStreamer(
+        tokenizer,
+        skip_prompt=True,
+        skip_special_tokens=True,
+    )
+
+    generation_kwargs = dict(
+        **inputs,
+        max_new_tokens=max_new_tokens,
         temperature=temperature,
         top_p=top_p,
-    ):
-        choices = message.choices
-        token = ""
-        if len(choices) and choices[0].delta.content:
-            token = choices[0].delta.content
+        do_sample=temperature > 0.0,
+        streamer=streamer,
+    )
 
+    thread = threading.Thread(target=model.generate, kwargs=generation_kwargs)
+    thread.start()
+
+    response = ""
+    for token in streamer:
         response += token
         yield response
 
+    thread.join()
+
 
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
 chatbot = gr.ChatInterface(
     respond,
+    type="messages",
+    title="Timmy — powered by nanowhale-100m",
     additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
+        gr.Textbox(value="You are a friendly chatbot.", label="System message"),
+        gr.Slider(minimum=1, maximum=512, value=128, step=1, label="Max new tokens"),
         gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
+        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"),
     ],
 )
 
-with gr.Blocks() as demo:
-    with gr.Sidebar():
-        gr.LoginButton()
-    chatbot.render()
-
-
 if __name__ == "__main__":
-    demo.launch()
+    chatbot.launch()