app.py

## app.py ##
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline,GPTNeoXForCausalLM, AutoConfig
from gradio import Interface
import gradio as gr
from accelerate import init_empty_weights
import json 
# Create a dictionary of models
MODELS = {
    "T5": "lmsys/fastchat-t5-3b-v1.0",
    "LSpanishGPT2": "PlanTL-GOB-ES/gpt2-large-bne",
    "GPT2": "datificate/gpt2-small-spanish",
    "OpenAssistant": "OpenAssistant"
}

# Define your function
def generate_and_analyze(model_name, input_text):
    model= MODELS[model_name]
    tokenizer = AutoTokenizer.from_pretrained(model)
    TOKENIZERS[model_name] = tokenizer

    text_generator = pipeline('text-generation', model=model, tokenizer=tokenizer, device=0) # Use GPU if available
    result = text_generator(input_text, max_length=250, do_sample=True)[0]
    return result['generated_text']

# Define your interface
iface = gr.Interface(
    fn=generate_and_analyze, 
    inputs=[
        gr.inputs.Dropdown(choices=list(MODELS.keys()) + ["bloom"], label="Model"),
        gr.inputs.Textbox(lines=2, label="Input Text")
    ], 
    outputs="text"
)
iface.launch()