src/display/utils.py

from dataclasses import dataclass, make_dataclass
from enum import Enum

import pandas as pd

from src.about import QATasks, CodeGenTasks


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


@dataclass
class ColumnContent:
    name: str
    type: str
    displayed_by_default: bool
    hidden: bool = False
    never_hidden: bool = False


# ---- QA Leaderboard columns ----
qa_column_dict = []
qa_column_dict.append(["rank", ColumnContent, ColumnContent("#", "number", True, never_hidden=True)])
qa_column_dict.append(["model", ColumnContent, ColumnContent("Model", "markdown", True, never_hidden=True)])
qa_column_dict.append(["size_access", ColumnContent, ColumnContent("Size / Access", "str", True)])
for task in QATasks:
    qa_column_dict.append([task.name, ColumnContent, ColumnContent(task.value.col_name, "number", True)])
qa_column_dict.append(["delta_overall", ColumnContent, ColumnContent("Delta vs 8B", "number", True)])

QALeaderboardColumn = make_dataclass("QALeaderboardColumn", qa_column_dict, frozen=True)

QA_COLS = [c.name for c in fields(QALeaderboardColumn) if not c.hidden]
QA_BENCHMARK_COLS = [t.value.col_name for t in QATasks]


# ---- Code Generation Leaderboard columns ----
codegen_column_dict = []
codegen_column_dict.append(["rank", ColumnContent, ColumnContent("#", "number", True, never_hidden=True)])
codegen_column_dict.append(["model", ColumnContent, ColumnContent("Model", "markdown", True, never_hidden=True)])
codegen_column_dict.append(["method", ColumnContent, ColumnContent("Method", "str", True)])
for task in CodeGenTasks:
    codegen_column_dict.append([task.name, ColumnContent, ColumnContent(task.value.col_name, "number", True)])

CodeGenLeaderboardColumn = make_dataclass("CodeGenLeaderboardColumn", codegen_column_dict, frozen=True)

CODEGEN_COLS = [c.name for c in fields(CodeGenLeaderboardColumn) if not c.hidden]
CODEGEN_BENCHMARK_COLS = [t.value.col_name for t in CodeGenTasks]


# ---- Model Types (kept for submission compatibility) ----
@dataclass
class ModelDetails:
    name: str
    display_name: str = ""
    symbol: str = ""


class ModelType(Enum):
    PT = ModelDetails(name="pretrained", symbol="🟢")
    FT = ModelDetails(name="fine-tuned", symbol="🔶")
    IFT = ModelDetails(name="instruction-tuned", symbol="⭕")
    RL = ModelDetails(name="RL-tuned", symbol="🟦")
    Unknown = ModelDetails(name="", symbol="?")

    def to_str(self, separator=" "):
        return f"{self.value.symbol}{separator}{self.value.name}"

    @staticmethod
    def from_str(type):
        if "fine-tuned" in type or "🔶" in type:
            return ModelType.FT
        if "pretrained" in type or "🟢" in type:
            return ModelType.PT
        if "RL-tuned" in type or "🟦" in type:
            return ModelType.RL
        if "instruction-tuned" in type or "⭕" in type:
            return ModelType.IFT
        return ModelType.Unknown


class WeightType(Enum):
    Adapter = ModelDetails("Adapter")
    Original = ModelDetails("Original")
    Delta = ModelDetails("Delta")


class Precision(Enum):
    float16 = ModelDetails("float16")
    bfloat16 = ModelDetails("bfloat16")
    Unknown = ModelDetails("?")

    def from_str(precision):
        if precision in ["torch.float16", "float16"]:
            return Precision.float16
        if precision in ["torch.bfloat16", "bfloat16"]:
            return Precision.bfloat16
        return Precision.Unknown