File size: 3,862 Bytes
c9e8e4a
a16fa71
c9e8e4a
 
 
3bce3fb
a16fa71
c9e8e4a
fa5e188
 
 
c9e8e4a
 
 
 
 
 
 
 
 
 
 
f4313df
c9e8e4a
 
 
5740d40
c5fafcd
7982bc6
a16fa71
5740d40
a16fa71
7982bc6
d297c51
5740d40
c9e8e4a
7cf1a13
33c3beb
e22d1d3
f4313df
7982bc6
 
a16fa71
9be3f4c
f25abd8
fa5e188
f4313df
3bce3fb
c9e8e4a
 
 
5740d40
a16fa71
 
f25abd8
5b9eb09
f25abd8
a19ffbb
f25abd8
5740d40
7cf1a13
90b0361
fa5e188
18d2e11
22fef42
 
 
 
 
a16fa71
7cf1a13
 
 
a16fa71
 
 
f4313df
7cf1a13
 
 
3283b93
a16fa71
 
7cf1a13
 
 
 
a16fa71
 
 
5740d40
a16fa71
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
import streamlit as st
from transformers import AutoTokenizer, AutoModelForCausalLM
from transformers import pipeline
import torch
import json
import pandas as pd
import requests

GITHUB_CODE = "https://huggingface.co/datasets/lvwerra/github-code"
INCODER_IMG = "https://huggingface.co/datasets/loubnabnl/repo-images/raw/main/incoder.png"

@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, low_cpu_mem_usage=True)
    return model

@st.cache()
def load_examples():
    with open("utils/examples.json", "r") as f:
        examples = json.load(f)
    return examples

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

st.sidebar.header("Models")
models = ["CodeParrot", "InCoder"]
selected_models = st.sidebar.multiselect("Select code generation models to compare", models, default=["CodeParrot"])

st.sidebar.header("Tasks")
tasks = [" ", "Pretraining datasets", "Model architecture", "Model evaluation", "Code generation"]
selected_task = st.sidebar.selectbox("Select a task", tasks)


if selected_task == " ":
    st.title("Code Generation Models")
    with open("utils/intro.txt", "r") as f:
        intro = f.read()
    st.markdown(intro)
    
elif selected_task == "Pretraining datasets":
    st.title("Pretraining datasets πŸ“š")
    st.markdown(f"Preview of some code files from Github repositories in [Github-code dataset]({GITHUB_CODE}):")   
    df = pd.read_csv("utils/data_preview.csv")
    st.dataframe(df)
    for model in selected_models:
        with open(f"datasets/{model.lower()}.txt", "r") as f:
            text = f.read()
        st.markdown(f"### {model}")
        st.markdown(text)  
        
elif selected_task == "Model architecture":
    st.title("Model architecture πŸ”¨")
    for model in selected_models:
        with open(f"architectures/{model.lower()}.txt", "r") as f:
            text = f.read()
        st.markdown(f"## {model}")
        st.markdown(text) 
        if model == "InCoder":
            st.image(INCODER_IMG, caption="Figure 1: InCoder training", width=700)

elif selected_task == "Model evaluation":
    st.title("Code models evaluation πŸ“Š")
    with open("evaluation/intro.txt", "r") as f:
        intro = f.read()
    st.markdown(intro)
    
elif selected_task == "Code generation":
    st.title("Code generation πŸ’»")
    st.sidebar.header("Examples")
    examples = load_examples()
    example_names = [example["name"] for example in examples]
    name2id = dict([(name, i) for i, name in enumerate(example_names)])
    selected_example = st.sidebar.selectbox("Select one of the following examples or implement yours", example_names)
    example_text = examples[name2id[selected_example]]["value"]
    default_length = examples[name2id[selected_example]]["length"]
    st.sidebar.header("Generation settings")
    temperature = st.sidebar.slider("Temperature:", value=0.2, min_value=0.0, step=0.1, max_value=2.0)
    max_new_tokens = st.sidebar.slider("Number of tokens to generate:", value=default_length, min_value=8, step=8, max_value=256)
    seed = st.sidebar.slider("Random seed:", value=42, min_value=0, step=1, max_value=1000)
    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..."):
            for model in selected_models:
                url = f'https://hf.space/embed/loubnabnl/{model.lower()}-subspace/+/api/predict/'
                r = requests.post(url=url, json={"data": [gen_prompt, max_new_tokens, temperature, seed]})
                generated_text = r.json()['data'][0]
                st.markdown(f"{model}")
                st.code(generated_text)