app.py

import streamlit as st
from transformers import AutoTokenizer, AutoModelForCausalLM, set_seed
from transformers import pipeline
import json

@st.cache(allow_output_mutation=True)
def load_tokenizer(model_ckpt):
    return AutoTokenizer.from_pretrained(model_ckpt)

@st.cache(allow_output_mutation=True)
def load_model(model_ckpt):
    model = AutoModelForCausalLM.from_pretrained(model_ckpt)
    return model

@st.cache()
def load_examples():
    with open("examples.json", "r") as f:
        examples = json.load(f)
    return dict([(x["name"], x["value"]) for x in examples])

st.set_page_config(page_icon=':parrot:', layout="wide")

default_code = '''\
def print_hello_world():\
'''

model_ckpt = "lvwerra/codeparrot"
tokenizer = load_tokenizer(model_ckpt)
model = load_model(model_ckpt)
examples = load_examples()
set_seed(42)
gen_kwargs = {}

st.title("CodeParrot 🦜")
st.markdown('##')

pipe = pipeline('text-generation', model=model, tokenizer=tokenizer)
st.sidebar.header("Examples:")
selected_example = st.sidebar.selectbox("Select one of the following examples:", examples.keys())
example_text = examples[selected_example]
st.sidebar.header("Generation settings:")
gen_kwargs["do_sample"] = st.sidebar.radio("Decoding strategy",  ["Greedy", "Sample"]) == "Sample"
gen_kwargs["max_new_tokens"] = st.sidebar.slider("Number of tokens to generate", value=32, min_value=8, step=8, max_value=256)
if gen_kwargs["do_sample"]:
    gen_kwargs["temperature"] = st.sidebar.slider("Temperature", value = 0.2, min_value = 0.0, max_value=2.0, step=0.05)
    gen_kwargs["top_k"] = st.sidebar.slider("Top-k", min_value = 0, max_value=100, value = 0)
    gen_kwargs["top_p"] = st.sidebar.slider("Top-p", min_value = 0.0, max_value=1.0, step = 0.01, value = 0.95)

gen_prompt = st.text_area("Generate code with prompt:", value=example_text, height=220,).strip()
if st.button("Generate code!"):
    with st.spinner("Generating code..."):
        generated_text = pipe(gen_prompt, **gen_kwargs)[0]['generated_text']
    st.code(generated_text)