| from dataclasses import dataclass |
| from enum import Enum |
|
|
| import pandas as pd |
|
|
| from src.about import Tasks |
|
|
|
|
| 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 |
|
|
|
|
| BENCHMARK_DISPLAY_NAME_OVERRIDES = { |
| "Scientific Figure": "Sci. Fig", |
| } |
|
|
|
|
| def benchmark_display_name(name: str) -> str: |
| return BENCHMARK_DISPLAY_NAME_OVERRIDES.get(name, name) |
|
|
|
|
| def benchmark_internal_name(name: str) -> str: |
| for internal_name, display_name in BENCHMARK_DISPLAY_NAME_OVERRIDES.items(): |
| if name == display_name: |
| return internal_name |
| return name |
|
|
|
|
| |
| auto_eval_column_dict = [] |
| |
| auto_eval_column_dict.append(["model", ColumnContent, ColumnContent("Model", "str", True, never_hidden=True)]) |
| |
| auto_eval_column_dict.append(["average", ColumnContent, ColumnContent("Average", "str", True)]) |
| auto_eval_column_dict.append(["dom_webpage", ColumnContent, ColumnContent("Webpage", "str", True)]) |
| auto_eval_column_dict.append(["dom_poster", ColumnContent, ColumnContent("Poster", "str", True)]) |
| auto_eval_column_dict.append(["dom_chart", ColumnContent, ColumnContent("Chart", "str", True)]) |
| auto_eval_column_dict.append( |
| ["dom_scientific_figure", ColumnContent, ColumnContent("Sci. Fig", "str", True)] |
| ) |
| auto_eval_column_dict.append(["dim_layout", ColumnContent, ColumnContent("Layout", "str", True)]) |
| auto_eval_column_dict.append(["dim_attribute", ColumnContent, ColumnContent("Attribute", "str", True)]) |
| auto_eval_column_dict.append(["dim_text", ColumnContent, ColumnContent("Text", "str", True)]) |
| auto_eval_column_dict.append(["dim_knowledge", ColumnContent, ColumnContent("Knowledge", "str", True)]) |
| auto_eval_column_dict.append(["dom_slides", ColumnContent, ColumnContent("Slides", "str", True)]) |
|
|
| |
| |
| AutoEvalColumn = type("AutoEvalColumn", (), {name: value for name, _, value in auto_eval_column_dict}) |
|
|
|
|
| |
| @dataclass(frozen=True) |
| class EvalQueueColumn: |
| 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) |
|
|
|
|
| |
| @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 |
|
|
|
|
| |
| COLS = [c.name for c in fields(AutoEvalColumn) if not c.hidden] |
|
|
| EVAL_COLS = [c.name for c in fields(EvalQueueColumn)] |
| EVAL_TYPES = [c.type for c in fields(EvalQueueColumn)] |
|
|
| BENCHMARK_COLS = [t.value.col_name for t in Tasks] |
|
|
