LatinGPT_alpha-01 / app_old.py
fede97's picture
Rename app.py to app_old.py
0c8d65a verified
# Run the script and open the link in the browser.
import os
import gradio as gr
import streamlit as st
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
# scratch with latbert tokenizer
CHECKPOINT_PATH= 'scratch_2-nodes_tokenizer_latbert-original_packing_fcocchi/model.safetensors'
CHECKPOINT_PATH= 'itserr/latin_llm_alpha'
print(f"Loading model from: {CHECKPOINT_PATH}")
tokenizer = AutoTokenizer.from_pretrained(CHECKPOINT_PATH, token=os.environ['HF_TOKEN'])
model = AutoModelForCausalLM.from_pretrained(CHECKPOINT_PATH, token=os.environ['HF_TOKEN'])
description="""
This is a Latin Language Model (LLM) based on GPT-2 and it was trained on a large corpus of Latin texts and can generate text in Latin.
Please enter a prompt in Latin to generate text.
"""
title= "(L<sup>3</sup>) - Latin Large Language Model"
article= "hello world ..."
examples= ['Accidere ex una scintilla', 'Audacter calumniare,', 'Consolatium misero comites']
logo_image= 'ITSERR_row_logo.png'
def generate_text(prompt):
if torch.cuda.is_available(): device = torch.device("cuda")
else:
device = torch.device("cpu")
print("No GPU available")
print("***** Generate *****")
text_generator = pipeline("text-generation", model=model, tokenizer=tokenizer, device=device)
generated_text = text_generator(prompt, max_length=50, do_sample=True, temperature=1.0, repetition_penalty=2.0, truncation=True)
return generated_text[0]['generated_text']
custom_css = """
#logo {
display: block;
margin-left: auto;
margin-right: auto;
width: 512px;
height: 256px;
}
"""
with gr.Blocks(css=custom_css) as demo:
gr.Markdown(f"<h1 style='text-align: center;'>{title}</h1>")
gr.Image(logo_image, elem_id="logo")
with gr.Row():
with gr.Column():
input_text = gr.Textbox(lines=5, placeholder="Enter latin text here...", label="Input Text")
with gr.Column():
output_text = gr.Textbox(lines=5, placeholder="Output text will appear here...", label="Output Text")
clean_button = gr.Button("Generate Text")
clean_button.click(fn=generate_text, inputs=input_text, outputs=output_text)
demo.launch(share=True)