File size: 5,105 Bytes
e2e6875
25557b5
 
 
 
 
 
 
 
05c90f4
 
 
 
 
6679087
 
 
 
 
 
25557b5
38f4369
ddc25db
 
 
 
 
 
 
 
 
38f4369
25557b5
 
 
c660995
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25557b5
 
05c90f4
c660995
25557b5
05c90f4
 
 
6679087
05c90f4
 
6679087
e29ab28
 
6679087
 
 
 
 
81f1dd1
0d4c659
6679087
 
0d4c659
6679087
 
 
 
 
 
0d4c659
 
6679087
0d4c659
6679087
 
 
a0691fa
0d4c659
81f1dd1
6679087
 
 
5b4c5f8
 
 
 
 
 
 
 
25557b5
 
e611814
 
 
 
 
 
 
 
 
 
 
6679087
e611814
 
 
6679087
ddc25db
 
 
 
 
 
5b4c5f8
ddc25db
e611814
 
2436603
6679087
05c90f4
6679087
fb9885c
6679087
25557b5
e611814
6679087
 
 
 
 
 
 
5b4c5f8
 
 
 
e611814
 
6679087
 
 
 
 
 
 
5b4c5f8
 
 
 
ddc25db
 
6679087
 
 
e611814
 
 
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
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
import io
import json

import gradio as gr
import pandas as pd
from huggingface_hub import HfFileSystem


RESULTS_DATASET_ID = "datasets/open-llm-leaderboard/results"
EXCLUDED_KEYS =  {
    "pretty_env_info",
    "chat_template",
    "group_subtasks",
}
# EXCLUDED_RESULTS_KEYS = {
#     "leaderboard",
# }
# EXCLUDED_RESULTS_LEADERBOARDS_KEYS = {
#     "alias",
# }


TASKS = {
    "leaderboard_arc_challenge": ("ARC", "leaderboard_arc_challenge"),
    "leaderboard_bbh": ("BBH", "leaderboard_bbh"),
    "leaderboard_gpqa": ("GPQA", "leaderboard_gpqa"),
    "leaderboard_ifeval": ("IFEval", "leaderboard_ifeval"),
    "leaderboard_math_hard": ("MATH", "leaderboard_math"),
    "leaderboard_mmlu_pro": ("MMLU-Pro", "leaderboard_mmlu_pro"),
    "leaderboard_musr": ("MuSR", "leaderboard_musr"),
}

fs = HfFileSystem()


def fetch_result_paths():
    paths = fs.glob(f"{RESULTS_DATASET_ID}/**/**/*.json")
    return paths


def filter_latest_result_path_per_model(paths):
    from collections import defaultdict

    d = defaultdict(list)
    for path in paths:
        model_id, _ = path[len(RESULTS_DATASET_ID) +1:].rsplit("/", 1)
        d[model_id].append(path)
    return {model_id: max(paths) for model_id, paths in d.items()}


def get_result_path_from_model(model_id, result_path_per_model):
    return result_path_per_model[model_id]


def load_data(result_path) -> pd.DataFrame:
    with fs.open(result_path, "r") as f:
        data = json.load(f)
    return data


def load_result_dataframe(model_id):
    result_path = get_result_path_from_model(model_id, latest_result_path_per_model)
    data = load_data(result_path)
    model_name = data.get("model_name", "Model")
    df = pd.json_normalize([{key: value for key, value in data.items() if key not in EXCLUDED_KEYS}])
    # df.columns = df.columns.str.split(".")  # .split return a list instead of a tuple
    return df.set_index(pd.Index([model_name])).reset_index()


def display_results(df_1, df_2, task):
    df = pd.concat([df.set_index("index") for df in [df_1, df_2] if "index" in df.columns])
    df = df.T.rename_axis(columns=None)
    return display_tab("results", df, task), display_tab("configs", df, task)


def display_tab(tab, df, task):
    df = df.style.format(na_rep="")
    df.hide(
        [
            row
            for row in df.index
            if (
                not row.startswith(f"{tab}.")
                or row.startswith(f"{tab}.leaderboard.")
                or row.endswith(".alias")
                or (not row.startswith(f"{tab}.{task}") if task != "All" else False)
            )
        ],
        axis="index",
    )
    start = len(f"{tab}.leaderboard_") if task == "All" else len(f"{tab}.{task} ")
    df.format_index(lambda idx: idx[start:].removesuffix(",none"), axis="index")
    return df.to_html()


def update_tasks(task):
    return gr.Radio(
            ["All"] + list(TASKS.values()),
            label="Tasks",
            info="Evaluation tasks to be displayed",
            value="All",
            interactive=True,
        )


# if __name__ == "__main__":
latest_result_path_per_model = filter_latest_result_path_per_model(fetch_result_paths())

with gr.Blocks(fill_height=True) as demo:
    gr.HTML("<h1 style='text-align: center;'>Compare Results of the 🤗 Open LLM Leaderboard</h1>")
    gr.HTML("<h3 style='text-align: center;'>Select 2 results to load and compare</h3>")

    with gr.Row():
        with gr.Column():
            model_id_1 = gr.Dropdown(choices=list(latest_result_path_per_model.keys()), label="Results")
            load_btn_1 = gr.Button("Load")
            dataframe_1 = gr.Dataframe(visible=False)
        with gr.Column():
            model_id_2 = gr.Dropdown(choices=list(latest_result_path_per_model.keys()), label="Results")
            load_btn_2 = gr.Button("Load")
            dataframe_2 = gr.Dataframe(visible=False)
    with gr.Row():
        task = gr.Radio(
            ["All"] + list(TASKS.values()),
            label="Tasks",
            info="Evaluation tasks to be displayed",
            value="All",
            interactive=False,
        )

    with gr.Row():
        # with gr.Tab("All"):
        #     pass
        with gr.Tab("Results"):
            results = gr.HTML()
        with gr.Tab("Configs"):
            configs = gr.HTML()

    load_btn_1.click(
        fn=load_result_dataframe,
        inputs=model_id_1,
        outputs=dataframe_1,
    ).then(
        fn=display_results,
        inputs=[dataframe_1, dataframe_2, task],
        outputs=[results, configs],
    ).then(
        fn=update_tasks,
        inputs=task,
        outputs=task,
    )
    load_btn_2.click(
        fn=load_result_dataframe,
        inputs=model_id_2,
        outputs=dataframe_2,
    ).then(
        fn=display_results,
        inputs=[dataframe_1, dataframe_2, task],
        outputs=[results, configs],
    ).then(
        fn=update_tasks,
        inputs=task,
        outputs=task,
    )
    task.change(
        fn=display_results,
        inputs=[dataframe_1, dataframe_2, task],
        outputs=[results, configs],
    )

demo.launch()