Update app.py

app.py

@@ -1,64 +1,46 @@
-import gradio as gr
-from huggingface_hub import InferenceClient
-
-"""
-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")
-
-
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    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]})
+# app.py
 
-    messages.append({"role": "user", "content": message})
-
-    response = ""
-
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
+import gradio as gr
+from transformers import AutoTokenizer, AutoModelForCausalLM
+from peft import PeftModel
+import torch
+
+# Load the model and tokenizer
+adapter_model_name = "acezxn/SOC_Task_Generation_Base"
+base_model_name = "unsloth/Llama-3.2-3B-Instruct-unsloth-bnb-4bit"  # You may want to change this to a standard model name
+
+# Load the tokenizer and model for CPU use (no bitsandbytes)
+tokenizer = AutoTokenizer.from_pretrained(base_model_name)
+
+# Load model for CPU usage without 4-bit quantization
+base_model = AutoModelForCausalLM.from_pretrained(
+    base_model_name,
+    # Do not use bitsandbytes for quantization, just use the normal model
+    load_in_4bit=False,  # Ensure not using 4-bit quantization
+    device_map=None,  # Use CPU (no device mapping needed)
+    trust_remote_code=True  # If necessary for running with remote code
+)
 
-        response += token
-        yield response
+# Move model to CPU (explicit, but optional)
+base_model.to('cpu')
 
+# Load the LoRA adapter
+adapter_model = PeftModel.from_pretrained(base_model, adapter_model_name)
 
-"""
-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)",
-        ),
-    ],
-)
+# Function to generate a response using the model
+def generate_response(input_text):
+    inputs = tokenizer(input_text, return_tensors="pt").to('cpu')  # Ensure inputs are on CPU
+    outputs = adapter_model.generate(**inputs)
+    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    return response
 
+# Create Gradio interface
+iface = gr.Interface(fn=generate_response,
+                     inputs=gr.Textbox(lines=2, placeholder="Enter your input here..."),
+                     outputs=gr.Textbox(),
+                     title="Llama LORA Adapter - SOC Task Generation",
+                     description="This is a Gradio app that uses a Llama LORA adapter (acezxn/SOC_Task_Generation_Base) with the base model Llama-3.2-3B-Instruct-unsloth-bnb-4bit to generate task-related responses.")
 
+# Launch the interface
 if __name__ == "__main__":
-    demo.launch()
+    iface.launch()