Create app.py

app.py

@@ -0,0 +1,36 @@
+import gradio as gr
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from peft import PeftModel
+import torch
+
+# Replace with your model repository ID
+model_repo_id = "ubiodee/Plutuslearn-Llama-3.2-3B-Instruct"
+
+# Load the tokenizer
+tokenizer = AutoTokenizer.from_pretrained(model_repo_id)
+
+# Load the base model and apply the PEFT adapter
+base_model = AutoModelForCausalLM.from_pretrained(
+    "meta-llama/Llama-3.2-3B-Instruct",
+    torch_dtype=torch.float16,
+    device_map="auto"
+)
+model = PeftModel.from_pretrained(base_model, model_repo_id)
+
+# Define the prediction function
+def predict(text):
+    inputs = tokenizer(text, return_tensors="pt").to("cuda")
+    outputs = model.generate(**inputs, max_length=100)  # Adjust parameters as needed
+    return tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+# Create Gradio interface
+demo = gr.Interface(
+    fn=predict,
+    inputs=gr.Textbox(label="Input Text"),
+    outputs=gr.Textbox(label="Model Output"),
+    title="My Model Demo",
+    description="Test the fine-tuned model hosted on Hugging Face."
+)
+
+# Launch the app
+demo.launch()