File size: 5,273 Bytes
6254b87
460d762
8c49cb6
6254b87
 
 
 
460d762
8c49cb6
 
460d762
 
 
8c49cb6
 
d52179b
460d762
8c49cb6
460d762
 
 
8c49cb6
460d762
8c49cb6
b323764
460d762
d16cee2
217b585
 
 
ecef2dc
b323764
8c49cb6
460d762
 
 
8c49cb6
d16cee2
8c49cb6
 
 
 
460d762
 
8c49cb6
460d762
 
 
 
 
 
 
 
8c49cb6
460d762
 
 
6eaad72
12cea14
460d762
 
8c49cb6
 
 
 
 
 
 
460d762
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
699e8ff
8c49cb6
6254b87
699e8ff
 
 
 
 
8c49cb6
 
 
 
 
 
 
699e8ff
 
 
 
 
6254b87
8c49cb6
 
460d762
 
 
 
8c49cb6
460d762
 
 
8c49cb6
460d762
8c49cb6
 
 
 
 
 
 
 
 
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
import os
from dataclasses import dataclass

from huggingface_hub import HfApi

API = HfApi()


# These classes are for user facing column names, to avoid having to change them
# all around the code when a modif is needed
@dataclass
class ColumnContent:
    name: str
    type: str
    displayed_by_default: bool
    hidden: bool = False


def fields(raw_class):
    return [v for k, v in raw_class.__dict__.items() if k[:2] != "__" and k[-2:] != "__"]


@dataclass(frozen=True)
class AutoEvalColumn:  # Auto evals column
    model_type_symbol = ColumnContent("T", "str", True)
    model = ColumnContent("Model", "markdown", True)
    average = ColumnContent("Average ⬆️", "number", True)
    arc = ColumnContent("ARC", "number", True)
    hellaswag = ColumnContent("HellaSwag", "number", True)
    mmlu = ColumnContent("MMLU", "number", True)
    truthfulqa = ColumnContent("TruthfulQA", "number", True)
    model_type = ColumnContent("Type", "str", False)
    precision = ColumnContent("Precision", "str", False)  # , True)
    license = ColumnContent("Hub License", "str", False)
    params = ColumnContent("#Params (B)", "number", False)
    likes = ColumnContent("Hub ❤️", "number", False)
    still_on_hub = ColumnContent("Available on the hub", "bool", False)
    revision = ColumnContent("Model sha", "str", False, False)
    dummy = ColumnContent(
        "model_name_for_query", "str", True
    )  # dummy col to implement search bar (hidden by custom CSS)


@dataclass(frozen=True)
class EloEvalColumn:  # Elo evals column
    model = ColumnContent("Model", "markdown", True)
    gpt4 = ColumnContent("GPT-4 (all)", "number", True)
    human_all = ColumnContent("Human (all)", "number", True)
    human_instruct = ColumnContent("Human (instruct)", "number", True)
    human_code_instruct = ColumnContent("Human (code-instruct)", "number", True)


@dataclass(frozen=True)
class EvalQueueColumn:  # Queue column
    model = ColumnContent("model", "markdown", True)
    revision = ColumnContent("revision", "str", True)
    private = ColumnContent("private", "bool", True)
    precision = ColumnContent("precision", "str", True)
    weight_type = ColumnContent("weight_type", "str", "Original")
    status = ColumnContent("status", "str", True)


LLAMAS = [
    "huggingface/llama-7b",
    "huggingface/llama-13b",
    "huggingface/llama-30b",
    "huggingface/llama-65b",
]


KOALA_LINK = "https://huggingface.co./TheBloke/koala-13B-HF"
VICUNA_LINK = "https://huggingface.co./lmsys/vicuna-13b-delta-v1.1"
OASST_LINK = "https://huggingface.co./OpenAssistant/oasst-sft-4-pythia-12b-epoch-3.5"
DOLLY_LINK = "https://huggingface.co./databricks/dolly-v2-12b"
MODEL_PAGE = "https://huggingface.co./models"
LLAMA_LINK = "https://ai.facebook.com/blog/large-language-model-llama-meta-ai/"
VICUNA_LINK = "https://huggingface.co./CarperAI/stable-vicuna-13b-delta"
ALPACA_LINK = "https://crfm.stanford.edu/2023/03/13/alpaca.html"


def model_hyperlink(link, model_name):
    return f'<a target="_blank" href="{link}" style="color: var(--link-text-color); text-decoration: underline;text-decoration-style: dotted;">{model_name}</a>'


def make_clickable_model(model_name):
    link = f"https://huggingface.co./{model_name}"

    if model_name in LLAMAS:
        link = LLAMA_LINK
        model_name = model_name.split("/")[1]
    elif model_name == "HuggingFaceH4/stable-vicuna-13b-2904":
        link = VICUNA_LINK
        model_name = "stable-vicuna-13b"
    elif model_name == "HuggingFaceH4/llama-7b-ift-alpaca":
        link = ALPACA_LINK
        model_name = "alpaca-13b"
    if model_name == "dolly-12b":
        link = DOLLY_LINK
    elif model_name == "vicuna-13b":
        link = VICUNA_LINK
    elif model_name == "koala-13b":
        link = KOALA_LINK
    elif model_name == "oasst-12b":
        link = OASST_LINK

    details_model_name = model_name.replace("/", "__")
    details_link = f"https://huggingface.co./datasets/open-llm-leaderboard/details_{details_model_name}"

    if not bool(os.getenv("DEBUG", "False")):
        # We only add these checks when not debugging, as they are extremely slow
        print(f"details_link: {details_link}")
        try:
            check_path = list(
                API.list_files_info(
                    repo_id=f"open-llm-leaderboard/details_{details_model_name}",
                    paths="README.md",
                    repo_type="dataset",
                )
            )
            print(f"check_path: {check_path}")
        except Exception as err:
            # No details repo for this model
            print(f"No details repo for this model: {err}")
            return model_hyperlink(link, model_name)

    return model_hyperlink(link, model_name) + "  " + model_hyperlink(details_link, "📑")


def styled_error(error):
    return f"<p style='color: red; font-size: 20px; text-align: center;'>{error}</p>"


def styled_warning(warn):
    return f"<p style='color: orange; font-size: 20px; text-align: center;'>{warn}</p>"


def styled_message(message):
    return f"<p style='color: green; font-size: 20px; text-align: center;'>{message}</p>"


def has_no_nan_values(df, columns):
    return df[columns].notna().all(axis=1)


def has_nan_values(df, columns):
    return df[columns].isna().any(axis=1)