Update app.py

app.py

@@ -1,66 +1,66 @@
-import os
-import torch
 import gradio as gr
-from transformers import AutoTokenizer, AutoModelForCausalLM
+from transformers import AutoModelForCausalLM, AutoTokenizer
 from peft import PeftModel
-from huggingface_hub import login
-
-# Login using HF token from secrets
-hf_token = os.environ.get("HF_TOKEN")
-if not hf_token:
-    raise RuntimeError("Missing HF_TOKEN in secrets.")
-login(token=hf_token)
+import torch
 
-# Base and LoRA model paths
+# Define the base and LoRA model IDs
 base_model_id = "unsloth/gemma-2-9b-bnb-4bit"
 lora_model_id = "Futuresony/future_12_10_2024"
 
-# Load tokenizer and base model
-tokenizer = AutoTokenizer.from_pretrained(base_model_id)
+# Load the base model on CPU with float16
 base_model = AutoModelForCausalLM.from_pretrained(
     base_model_id,
     torch_dtype=torch.float16,
-    device_map="auto"
+    device_map="cpu",  # Load the model on CPU, no GPU
 )
 
-# Load LoRA weights
+# Load the PEFT LoRA model
 model = PeftModel.from_pretrained(base_model, lora_model_id)
-model.eval()
 
-# Chat function
-def generate_response(message, history, system_message, max_tokens, temperature, top_p):
-    prompt = system_message + "\n\n"
-    for user_input, bot_response in history:
-        prompt += f"User: {user_input}\nAssistant: {bot_response}\n"
-    prompt += f"User: {message}\nAssistant:"
+# Tokenizer for the model
+tokenizer = AutoTokenizer.from_pretrained(base_model_id)
 
-    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
-    outputs = model.generate(
-        **inputs,
-        max_new_tokens=max_tokens,
+# Function to respond to the user's input
+def respond(message, history: list[tuple[str, str]], system_message, max_tokens, temperature, top_p):
+    # Prepare the message history for chat completion
+    messages = [{"role": "system", "content": system_message}]
+    for val in history:
+        if val[0]:
+            messages.append({"role": "user", "content": val[0]})
+        if val[1]:
+            messages.append({"role": "assistant", "content": val[1]})
+
+    messages.append({"role": "user", "content": message})
+
+    # Generate a response
+    response = ""
+    for message in model.chat_completion(
+        messages,
+        max_tokens=max_tokens,
+        stream=True,
         temperature=temperature,
         top_p=top_p,
-        do_sample=True,
-        pad_token_id=tokenizer.eos_token_id
-    )
+    ):
+        token = message.choices[0].delta.content
+        response += token
+        yield response
 
-    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    final_response = response.split("Assistant:")[-1].strip()
-    return final_response
-
-# Gradio interface
+# Gradio interface setup
 demo = gr.ChatInterface(
-    fn=generate_response,
+    respond,
     additional_inputs=[
-        gr.Textbox(value="You are a helpful assistant.", label="System Message"),
-        gr.Slider(50, 1024, value=256, step=1, label="Max Tokens"),
-        gr.Slider(0.1, 1.5, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(0.1, 1.0, value=0.95, step=0.05, label="Top-p"),
+        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)",
+        ),
     ],
-    title="LoRA Chat Assistant (Gemma-2)",
-    description="Chat with your fine-tuned Gemma-2 LoRA model"
 )
 
 if __name__ == "__main__":
     demo.launch()
-