app.py

import gradio as gr
import transformers
from transformers import pipeline
from transformers import BloomTokenizerFast




def generate(checkpoint, input_prompt):

    tokenizer = BloomTokenizerFast.from_pretrained("bigscience/bloom")
    generator = pipeline("text-generation", model='simonosgoode/bloom-560m-finetuned-cdn_law', tokenizer=tokenizer)
    
    generated_judgement = generator(input_prompt
                              , max_length = 100
                              , num_return_sequences = 1
                              , return_full_text = True
                              , verbose = 0
                              #, num_beams = 1
                              #, early_stopping = True    
                              , temperature = 0.7
                              #, top_k = 50     # Default 50
                              , top_p = 1    # Default 1.0
                              , no_repeat_ngram_size = 3   # Default = 0
                              , repetition_penalty = 1.0   # Default = 1.0
                              #, do_sample = True     # Default = False
                              )[0]["generated_text"]
    return generated_judgement

with gr.Blocks() as judgements:
    inputs = gr.Textbox(lines=10, label="Input paragraph")
    output = gr.Textbox(lines=10, label="Output paragraph")
    btn = gr.Button("Generate the next paragraph of a judgement")
    btn.click(fn=generate, inputs=inputs, outputs=output)