Update app.py

app.py

@@ -1,64 +1,105 @@
 import gradio as gr
-from huggingface_hub import InferenceClient
+from llama_cpp import Llama
+from transformers import AutoTokenizer
+from huggingface_hub import hf_hub_download
+import os
 
-"""
-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
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+# Model paths
+MODEL_REPO_ID = "LGAI-EXAONE/EXAONE-4.0-1.2B"
+GGUF_REPO_ID = "lmstudio-community/LGAI-EXAONE-4.0-1.2B-GGUF"
+GGUF_FILENAME = "A.X-4.0-Light-imatrix-IQ1_S.gguf"
+TOKENIZER_DIR = "exaone-tokenizer"
+TEMPLATE_PATH = os.path.join(TOKENIZER_DIR, "chat_template.jinja")
 
+# Download GGUF model 
+print("🔄 Downloading GGUF model...")
+model_path = hf_hub_download(repo_id=GGUF_REPO_ID, filename=GGUF_FILENAME)
+print(f"✅ Model downloaded to: {model_path}")
 
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
+# Load tokenizer
+print("🔄 Loading tokenizer...")
+tokenizer = AutoTokenizer.from_pretrained(TOKENIZER_DIR, trust_remote_code=True)
+print("✅ Tokenizer loaded.")
 
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
+try:
+    with open(TEMPLATE_PATH, "r", encoding="utf-8") as f:
+        tokenizer.chat_template = f.read()
+    print("✅ Chat template loaded.")
+except Exception as e:
+    print(f"Could not load chat_template.jinja: {e}")
+    tokenizer.chat_template = None
 
-    messages.append({"role": "user", "content": message})
+# Load model 
+llm = Llama(
+    model_path=model_path,
+    n_ctx=2048,
+    n_threads=os.cpu_count(),
+    n_gpu_layers=-1,
+    use_mlock=True,
+    verbose=False
+)
+
+# Streaming chat function 
+def format_prompt(messages):
+    formatted = ""
+    for m in messages:
+        role = m["role"].upper()
+        formatted += f"{role}: {m['content']}\n"
+    return formatted + "ASSISTANT:"
+
+def user_input_handler(user_message, history):
+    return "", history + [[user_message, None]]
+
+def bot_stream(history):
+    user_message = history[-1][0]
+    history[-1][1] = ""
+
+    # Convert chat history to OpenAI format
+    messages = [{"role": "system", "content": "You are a helpful assistant."}]
+    for human, assistant in history[:-1]:
+        messages.append({"role": "user", "content": human})
+        if assistant:
+            messages.append({"role": "assistant", "content": assistant})
+    messages.append({"role": "user", "content": user_message})
 
-    response = ""
+    try:
+        prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+    except Exception:
+        prompt = format_prompt(messages)
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
+    # Generate streaming output
+    generator = llm.create_completion(
+        prompt=prompt,
+        max_tokens=512,
+        temperature=0.7,
+        top_p=0.9,
         stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
-
-        response += token
-        yield response
-
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-demo = gr.ChatInterface(
-    respond,
-    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.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)",
-        ),
-    ],
-)
+        stop=["</s>", "<|endoftext|>", "USER:", "ASSISTANT:"]
+    )
+
+    for chunk in generator:
+        token = chunk["choices"][0]["text"]
+        history[-1][1] += token
+        yield history
+
+# Gradio UI 
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    gr.Markdown("## 🤖 EXAONE-4.0-1.2B Streaming")
+    chatbot = gr.Chatbot(label="Chat History", height=600)
+    with gr.Row():
+        msg = gr.Textbox(placeholder="Type a message...", label="Your Message", scale=8)
+        send_btn = gr.Button("Send", scale=1)
+        clear_btn = gr.Button("Clear Chat", scale=1, variant="secondary")
+
+    # Chat events
+    msg.submit(user_input_handler, [msg, chatbot], [msg, chatbot], queue=False).then(
+        bot_stream, chatbot, chatbot
+    )
+    send_btn.click(user_input_handler, [msg, chatbot], [msg, chatbot], queue=False).then(
+        bot_stream, chatbot, chatbot
+    )
 
+    clear_btn.click(lambda: [], None, chatbot)
 
-if __name__ == "__main__":
-    demo.launch()
+demo.queue()
+demo.launch()