app.py

from transformers import AutoTokenizer, AutoModelForCausalLM
import gradio as gr

tokenizer = AutoTokenizer.from_pretrained("microsoft/phi-2", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("microsoft/phi-2", torch_dtype="auto", flash_attn=True, flash_rotary=True, fused_dense=True, device_map="cuda", trust_remote_code=True)

def generate(prompt, length):
    inputs = tokenizer(prompt, return_tensors="pt", return_attention_mask=False)
    outputs = model.generate(**inputs, max_length=length)
    return tokenizer.batch_decode(outputs)[0]

demo = gr.Interface(fn=generate, inputs=["text", "number"], outputs="text")

if __name__ == "__main__":
   demo.launch(show_api=False)