Update app.py

app.py

@@ -1,47 +1,59 @@
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from peft import PeftModel
 import gradio as gr
-from huggingface_hub import InferenceClient
 
-# Initialize the InferenceClient
-client = InferenceClient(model="Grandediw/lora_model")
+# Load the base model
+base_model_name = "unsloth/llama-3.2-3b-instruct-bnb-4bit"
+tokenizer = AutoTokenizer.from_pretrained(base_model_name, use_fast=False)
+base_model = AutoModelForCausalLM.from_pretrained(
+    base_model_name,
+    device_map="auto",  # Automatically map layers to available devices
+    torch_dtype=torch.float16  # Ensure compatibility with 4-bit quantization
+)
+
+# Load the LoRA adapter
+adapter_path = "Grandediw/lora_model"  # Replace with your model path
+model = PeftModel.from_pretrained(base_model, adapter_path)
+model.eval()  # Set the model to evaluation mode
 
+# Define the inference function
 def respond(
     message,
     history: list[tuple[str, str]],
-    system_message,
     max_tokens,
     temperature,
     top_p,
 ):
-    context = system_message + "\n"
+    # Build context from history
+    context = ""
     for user_message, assistant_message in history:
         context += f"User: {user_message}\nAssistant: {assistant_message}\n"
     context += f"User: {message}\nAssistant:"
 
-    try:
-        response = client.text_generation(
-            context,
-            max_new_tokens=max_tokens,
-            temperature=temperature,
-            top_p=top_p,
-        )
-        yield response
-    except Exception as e:
-        yield f"Error: {e}"
+    # Tokenize the input
+    inputs = tokenizer(context, return_tensors="pt").to("cuda")
+
+    # Generate a response
+    outputs = model.generate(
+        input_ids=inputs.input_ids,
+        max_new_tokens=max_tokens,
+        temperature=temperature,
+        top_p=top_p,
+        do_sample=True
+    )
+
+    # Decode and return the response
+    response = tokenizer.decode(outputs[:, inputs.input_ids.shape[-1]:][0], skip_special_tokens=True)
+    return response
 
 # Build the Gradio ChatInterface
 demo = gr.ChatInterface(
-    respond,
+    fn=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)",
-        ),
+        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max Tokens"),
+        gr.Slider(minimum=0.1, maximum=4.0, value=1.5, step=0.1, label="Temperature"),
+        gr.Slider(minimum=0.1, maximum=1.0, value=0.9, step=0.05, label="Top-p"),
     ],
 )