Spaces:
Sleeping
Sleeping
File size: 3,845 Bytes
9d22eee 2a5f9fb df66f6e 1be492d efeee6d 9d22eee 314f91a 2a5f9fb efeee6d 9d22eee b2b2434 9d22eee 1be492d 78df786 1be492d 9d22eee b431323 9d22eee 632297e 5b72ce2 9d22eee 2a5f9fb efeee6d 2a5f9fb 708fbc1 2a5f9fb efeee6d 2a5f9fb 9d22eee 2a5f9fb 9833cdb 2a5f9fb 9d22eee 2a5f9fb 9d22eee 2a5f9fb b1a1395 2a5f9fb |
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 |
from dataclasses import dataclass, make_dataclass
from enum import Enum
import pandas as pd
from src.about import Tasks, N_Tasks, Detail_Tasks
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", ColumnContent, ColumnContent("Model", "markdown", True, never_hidden=True)])
auto_eval_column_dict.append(["output_format", ColumnContent, ColumnContent("Output Format", "str", 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)])
for task in Detail_Tasks:
auto_eval_column_dict.append([task.name, ColumnContent, ColumnContent(task.value.col_name, "number", False)])
# Model information
auto_eval_column_dict.append(["architecture", ColumnContent, ColumnContent("Architecture", "str", False)])
auto_eval_column_dict.append(["revision", ColumnContent, ColumnContent("Model sha", "str", False, False)])
auto_eval_column_dict.append(["dataset_version", ColumnContent, ColumnContent("Task Version", "str", False, False)])
# We use make dataclass to dynamically fill the scores from Tasks
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)
output_format = ColumnContent("output_format", "str", True)
dataset_version = ColumnContent("dataset_version", "str", True)
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="🟦")
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]
|