File size: 4,680 Bytes
68ed9a2 f8e2509 1613f96 f8e2509 1613f96 827199d 1613f96 68ed9a2 1613f96 f8e2509 1613f96 f8e2509 68ed9a2 f8e2509 68ed9a2 f8e2509 1613f96 f8e2509 1613f96 f8e2509 68ed9a2 1613f96 827199d 68ed9a2 1613f96 827199d 1613f96 827199d 1613f96 827199d 1613f96 68ed9a2 1613f96 68ed9a2 1613f96 |
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 |
#!/usr/bin/env python
# -*- coding: utf-8 -*-
# flake8: noqa E501
from dataclasses import dataclass, field, make_dataclass
from enum import Enum
from functools import partial
import pandas as pd
from src.about import Tasks
def fields(raw_class):
return [v for k, v in raw_class.__dict__.items() if not k.startswith("__") and not k.endswith("__")]
# 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 = field(default='str')
displayed_by_default: bool = field(default=False)
hidden: bool = field(default=False)
never_hidden: bool = field(default=False)
# Helper function to create a ColumnContent with default_factory
def create_column_content(name, type='str', displayed_by_default=False, hidden=False, never_hidden=False):
return field(default_factory=lambda: ColumnContent(name, type, displayed_by_default, hidden, never_hidden))
auto_eval_column_dict = [
("model_type_symbol", ColumnContent, create_column_content("", "str", True, never_hidden=True)),
("model", ColumnContent, create_column_content("Model", "markdown", True, never_hidden=True)),
("average", ColumnContent, create_column_content("Average", "number", True)),
]
# Add task-specific columns
for task in Tasks:
auto_eval_column_dict.append((task.name, ColumnContent, create_column_content(task.value.col_name, "number", True)))
# Add model information columns
model_info_columns = [
("model_type", "Type", "str", False),
("architecture", "Architecture", "str", False),
("weight_type", "Weight type", "str", False, True),
("precision", "Precision", "str", False),
("license", "License", "str", False),
("params", "Parameters ⚙️", "number", False),
("likes", "Likes ❤️", "number", False),
("still_on_hub", "Available on the hub", "bool", False),
("revision", "Model sha", "str", False, False),
]
for col_name, display_name, col_type, displayed_by_default, *args in model_info_columns:
hidden = args[0] if args else False
auto_eval_column_dict.append((col_name, ColumnContent, create_column_content(display_name, col_type, displayed_by_default, hidden)))
# Create the AutoEvalColumn dataclass
AutoEvalColumn = make_dataclass("AutoEvalColumn", 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):
PT = ModelDetails(name="pretrained", symbol="🟢")
FT = ModelDetails(name="fine-tuned", symbol="🔶")
IFT = ModelDetails(name="instruction-tuned", symbol="⭕")
RL = ModelDetails(name="RL-tuned", symbol="🟦")
CS = ModelDetails(name="closed-source", symbol="🔒")
BrainDAO = ModelDetails(name="braindao", 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
# 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]
|