app.py

import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer
from accelerate import infer_auto_device_map

# Load the model name
model_name = "ai4bharat/Airavata"

# Load the tokenizer
tokenizer = AutoTokenizer.from_pretrained(model_name)

# Load the model first
model = AutoModelForCausalLM.from_pretrained(model_name, load_in_8bit=True)

# Now infer the device map
device_map = infer_auto_device_map(model)

# Move model to the appropriate device based on device_map
model.to(device_map)

# Define the inference function
def generate_text(prompt):
    inputs = tokenizer(prompt, return_tensors="pt")
    outputs = model.generate(**inputs)
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

# Create the Gradio interface
interface = gr.Interface(
    fn=generate_text,
    inputs="text",
    outputs="text",
    title="Airavata Text Generation Model",
    description="This is the AI4Bharat Airavata model for text generation in Indic languages."
)

# Launch the interface
interface.launch()