app.py

import gradio as gr
from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig,BitsAndBytesConfig

import torch

model_id = "truongghieu/deci-finetuned_Prj2"

# Check if a GPU is available
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# Just for GPU 
bnb_config = BitsAndBytesConfig(
        load_in_4bit=True, bnb_4bit_quant_type="nf4", bnb_4bit_compute_dtype="float16", bnb_4bit_use_double_quant=True
    )


tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
# Load model in this way if use GPU
if (device == "cuda"):
    model = AutoModelForCausalLM.from_pretrained(model_id, trust_remote_code=True, quantization_config=bnb_config)
elif(device == "cpu"):
    model = AutoModelForCausalLM.from_pretrained("truongghieu/deci-finetuned", trust_remote_code=True)
# Move the model to the GPU if available

generation_config = GenerationConfig(
    penalty_alpha=0.6,
    do_sample=True,
    top_k=3,
    temperature=0.5,
    repetition_penalty=1.2,
    max_new_tokens=50,
    pad_token_id=tokenizer.eos_token_id
)



# Define a function that takes a text input and generates a text output
def generate_text(text):
    input_text = f'###Human: \"{text}\"'
    input_ids = tokenizer.encode(input_text, return_tensors="pt").to(device)
    output_ids = model.generate(input_ids, generation_config=generation_config)
    output_text = tokenizer.decode(output_ids[0], skip_special_tokens=True)
    return output_text

iface = gr.Interface(fn=generate_text, inputs="text", outputs="text")
iface.launch()