Spaces:
Runtime error
Runtime error
File size: 5,943 Bytes
3d8dbe8 8b1f7a0 3b3db42 370d5a0 29546b4 3d8dbe8 91e8a06 8b1f7a0 29546b4 3d8dbe8 3b86dfc 3d8dbe8 3b86dfc 3d8dbe8 3b86dfc 3d8dbe8 8b1f7a0 5f7fcf4 370d5a0 29546b4 8b1f7a0 3b86dfc 8b1f7a0 29546b4 8b1f7a0 3d8dbe8 8b1f7a0 2a860f6 8b1f7a0 3b86dfc 3d8dbe8 8b1f7a0 3b86dfc 8b1f7a0 3b86dfc 8b1f7a0 ceb2102 8b1f7a0 |
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 |
from dataclasses import dataclass, make_dataclass
from enum import Enum
import pandas as pd
from src.about import Tasks, SingleTableTasks, SingleColumnTasks
def fields(raw_class):
return [v for k, v in raw_class.__dict__.items() if k[:2] != "__" and k[-2:] != "__"]
# 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
never_hidden: bool = False
## Leaderboard columns
auto_eval_column_dict = []
# Init
# auto_eval_column_dict.append(["model_type_symbol", ColumnContent, ColumnContent("T", "str", True, never_hidden=True)])
auto_eval_column_dict.append(["dataset", ColumnContent, ColumnContent("Dataset", "str", True, never_hidden=True)])
auto_eval_column_dict.append(["model", ColumnContent, ColumnContent("Model", "markdown", True, never_hidden=True)])
#Scores
# auto_eval_column_dict.append(["average", ColumnContent, ColumnContent("Average ⬆️", "number", True)])
for task in Tasks:
auto_eval_column_dict.append([task.name, ColumnContent, ColumnContent(task.value.col_name, "number", True)])
# Model information
# auto_eval_column_dict.append(["model_type", ColumnContent, ColumnContent("Type", "str", False)])
# auto_eval_column_dict.append(["architecture", ColumnContent, ColumnContent("Architecture", "str", False)])
# auto_eval_column_dict.append(["weight_type", ColumnContent, ColumnContent("Weight type", "str", False, True)])
# auto_eval_column_dict.append(["precision", ColumnContent, ColumnContent("Precision", "str", False)])
# auto_eval_column_dict.append(["license", ColumnContent, ColumnContent("Hub License", "str", False)])
# auto_eval_column_dict.append(["params", ColumnContent, ColumnContent("#Params (B)", "number", False)])
# auto_eval_column_dict.append(["likes", ColumnContent, ColumnContent("Hub ❤️", "number", False)])
# auto_eval_column_dict.append(["still_on_hub", ColumnContent, ColumnContent("Available on the hub", "bool", False)])
# auto_eval_column_dict.append(["revision", ColumnContent, ColumnContent("Model sha", "str", False, False)])
# We use make dataclass to dynamically fill the scores from Tasks
AutoEvalColumn = make_dataclass("AutoEvalColumn", auto_eval_column_dict, frozen=True)
singletable_auto_eval_column_dict = []
# Init
singletable_auto_eval_column_dict.append(["dataset", ColumnContent, ColumnContent("Dataset", "str", True, never_hidden=True)])
# singletable_auto_eval_column_dict.append(["table", ColumnContent, ColumnContent("Table", "str", True, never_hidden=True)])
singletable_auto_eval_column_dict.append(["model", ColumnContent, ColumnContent("Model", "markdown", True, never_hidden=True)])
#Scores
for task in SingleTableTasks:
singletable_auto_eval_column_dict.append([task.name, ColumnContent, ColumnContent(task.value.col_name, "number", True)])
singletable_AutoEvalColumn = make_dataclass("AutoEvalColumn", singletable_auto_eval_column_dict, frozen=True)
# SINGLE COLUMN
singlecolumn_auto_eval_column_dict = []
# Init
singlecolumn_auto_eval_column_dict.append(["dataset", ColumnContent, ColumnContent("Dataset", "str", True, never_hidden=True)])
singlecolumn_auto_eval_column_dict.append(["model", ColumnContent, ColumnContent("Model", "markdown", True, never_hidden=True)])
singlecolumn_auto_eval_column_dict.append(["table", ColumnContent, ColumnContent("Table", "str", True, never_hidden=True)])
#Scores
for task in SingleColumnTasks:
singlecolumn_auto_eval_column_dict.append([task.name, ColumnContent, ColumnContent(task.value.col_name, "number", True)])
singlecolumn_AutoEvalColumn = make_dataclass("AutoEvalColumn", singlecolumn_auto_eval_column_dict, frozen=True)
## For the queue columns in the submission tab
@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)
## All the model information that we might need
@dataclass
class ModelDetails:
name: str
display_name: str = ""
symbol: str = "" # emoji
class ModelType(Enum):
OS = ModelDetails(name="open-source", symbol="🆓")
CS = ModelDetails(name="closed-source", symbol="🔒")
# 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 "open-source" in type or "🔶" in type:
return ModelType.OS
if "closed-source" in type or "🟢" in type:
return ModelType.CS
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
# Column selection
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]
|