Spaces:
Runtime error
Runtime error
File size: 4,695 Bytes
2171b06 72c2877 2171b06 72c2877 2171b06 72c2877 2171b06 72c2877 2171b06 72c2877 2171b06 72c2877 2171b06 72c2877 2171b06 72c2877 2171b06 72c2877 2171b06 72c2877 2171b06 72c2877 2171b06 72c2877 2171b06 72c2877 2171b06 72c2877 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 |
import json
import requests
import streamlit as st
st.set_page_config(layout="wide")
with open("utils/table_contents.md", "r") as f:
contents = f.read()
st.sidebar.markdown(contents)
st.title("The Stack Bot 🤖")
intro = """
The Stack Bot is a tool to help you get started with tools developed in [BigCode](https://huggingface.co/bigcode),
such as [The Stack](https://huggingface.co/bigcode/the-stack) dataset and [SantaCoder](https://huggingface.co/bigcode/santacoder) model.
"""
st.markdown(intro, unsafe_allow_html=True)
@st.cache()
def load_languages():
with open("utils/languages.json", "r") as f:
languages = json.load(f)
return languages
def how_to_load(language):
text = f"""
```python
from datasets import load_dataset
dataset = load_dataset("bigcode/the-stack", data_dir="data/{language}", split="train")
# print first element
print(dataset[0])
```
"""
st.markdown(text)
def load_model(values, language):
model = values["model"]
if not model:
text = f"""No model available for {language.capitalize()}. If you trained a model on this language, let us know at contact@bigcode.com to feature your model!\n\
You can also train your own model on The Stack using the instructions below 🚀"""
st.write(text)
if st.button("Fine-tune your own model", key=4):
st.write("Code available at [GitHub link] + add preview")
else:
text = f"""{model} is a model that was trained on the {language.capitalize()} subset of The Stack. Here's how to use it:"""
code = f"""
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained({model})
model = AutoModelForCausalLM.from_pretrained({model}, trust_remote_code=True)
inputs = tokenizer.encode("def print_hello_world():", return_tensors="pt")
outputs = model.generate(inputs)
print(tokenizer.decode(outputs[0]))
```
"""
st.write(text)
st.markdown(code)
st.write(f"The scores of this model are the following: {values['scores']}")
def generate_code(
demo, gen_prompt, max_new_tokens=40, temperature=0.2, seed=0
):
# call space using its API endpoint
#try:
url = (
f"{demo}/run/predict/"
)
r = requests.post(
url=url, json={"data": [gen_prompt, max_new_tokens, temperature, seed]}
)
generated_text = r.json()["data"][0]
return generated_text
def init_nested_buttons():
if "Models trained on dataset" not in st.session_state:
st.session_state["Models trained on dataset"] = False
if "Generate code" not in st.session_state:
st.session_state["Generate code"] = False
if st.button("Models trained on dataset"):
st.session_state["Models trained on dataset"] = not st.session_state["Models trained on dataset"]
languages = load_languages()
col1, col2 = st.columns([1, 1.5])
with col1:
selected_language = st.selectbox("Select one of 358 languages in The Stack", list(languages.keys()), key=1)
st.write(f"Here's how you can load the {selected_language.capitalize()} subset of The Stack:")
code = how_to_load(selected_language)
if st.button("More info about the dataset", key=2):
st.write(f"The dataset contains {languages[selected_language]['num_examples']} examples.")
# we can add some stats about files
init_nested_buttons()
if st.session_state["Models trained on dataset"]:
load_model(languages[selected_language], selected_language)
if languages[selected_language]["model"] and languages[selected_language]["gradio_demo"]:
st.write(f"Here's a demo to try the model, for more flexibilty you can use the [Gradio demo]({languages[selected_language]['gradio_demo']}).")
gen_prompt = st.text_area(
"Generate code with prompt:",
value="# Implement a function to print hello world",
height=100,
).strip()
if st.button("Generate code"):
st.session_state["Generate code"] = not st.session_state["Generate code"]
if st.session_state["Generate code"]:
with st.spinner("Generating code..."):
generated_text = generate_code(
demo=languages[selected_language]["gradio_demo"],
gen_prompt=gen_prompt,
)
if not generated_text:
st.markdown(f"Error: could not generate code. Make sure the Gradio demo at [{languages[selected_language]['gradio_demo']}]({languages[selected_language]['gradio_demo']}) works.")
else:
st.code(generated_text) |