Update app.py

app.py

@@ -1,76 +1,66 @@
 import gradio as gr
-from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig
-from peft import PeftModel
-import torch
+from huggingface_hub import InferenceClient
 
-# -----------------------------
-# CPU-friendly model
-# -----------------------------
-MODEL_NAME = "tiiuae/falcon-7b-instruct"  # smaller CPU-friendly model
+def respond(
+    message,
+    history: list[dict[str, str]],
+    system_message,
+    max_tokens,
+    temperature,
+    top_p,
+    hf_token: gr.OAuthToken,
+):
+    """
+    Simple CPU-compatible chat using Hugging Face InferenceClient.
+    Works without GPU and 4-bit models.
+    """
+    client = InferenceClient(token=hf_token.token, model="openai/gpt-oss-20b")  # keep your model
 
-print("Loading tokenizer...")
-tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, use_fast=False)
+    # Build messages for chat
+    messages = [{"role": "system", "content": system_message}]
+    messages.extend(history)
+    messages.append({"role": "user", "content": message})
 
-print("Loading model...")
-base_model = AutoModelForCausalLM.from_pretrained(
-    MODEL_NAME,
-    device_map=None,  # CPU only
-    torch_dtype=torch.float32
-)
-
-# Load LoRA if exists (optional)
-try:
-    model = PeftModel.from_pretrained(base_model, MODEL_NAME, device_map=None)
-except:
-    model = base_model
-
-model.eval()
-
-# -----------------------------
-# Response function
-# -----------------------------
-def respond(message, history, system_message, max_tokens, temperature, top_p):
-    # Limit max tokens for CPU safety
-    if max_tokens > 128:
-        max_tokens = 128
+    response = ""
+    try:
+        # Streaming response
+        for chunk in client.chat_completion(
+            messages,
+            max_tokens=max_tokens,
+            stream=True,
+            temperature=temperature,
+            top_p=top_p,
+        ):
+            choices = getattr(chunk, "choices", None) or chunk.get("choices", [])
+            token = ""
+            if len(choices) and getattr(choices[0].delta, "content", None):
+                token = choices[0].delta.content
+            elif len(choices) and choices[0].get("delta", {}).get("content"):
+                token = choices[0]["delta"]["content"]
+            response += token
+            yield response
 
-    # Build prompt
-    prompt = system_message + "\n"
-    for h in history:
-        prompt += f"User: {h['content']}\n"
-    prompt += f"User: {message}\nBot:"
+    except Exception as e:
+        yield f"[Error streaming response] {str(e)}"
 
-    inputs = tokenizer(prompt, return_tensors="pt")
-
-    gen_config = GenerationConfig(
-        max_new_tokens=max_tokens,
-        temperature=temperature,
-        top_p=top_p,
-        do_sample=True
-    )
-
-    with torch.no_grad():
-        output_ids = model.generate(**inputs, **gen_config.to_dict())
-    output = tokenizer.decode(output_ids[0][inputs['input_ids'].shape[1]:], skip_special_tokens=True)
-    return output
-
-# -----------------------------
-# Gradio Chat Interface
-# -----------------------------
+# Gradio ChatInterface
 chatbot = gr.ChatInterface(
     respond,
     type="messages",
     additional_inputs=[
         gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=128, value=64, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=2.0, value=0.7, step=0.1, label="Temperature"),
+        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)"),
     ],
 )
 
+# Build Gradio UI
 with gr.Blocks() as demo:
+    with gr.Sidebar():
+        gr.LoginButton()  # OAuth login
     chatbot.render()
 
 if __name__ == "__main__":
-    demo.launch(server_name="0.0.0.0", server_port=7860)
-
+    # Launch on CPU, with a public link
+    demo.launch(share=True)