File size: 2,134 Bytes
12efa10 f07d235 12efa10 eec2226 f07d235 6b21330 ca48878 12efa10 6b21330 f07d235 ca48878 12efa10 f07d235 12efa10 f07d235 12efa10 |
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 |
from dataclasses import dataclass, make_dataclass
from src.about import EvalDimensions
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(["rank", ColumnContent, ColumnContent("Rank", "str", True, False)])
auto_eval_column_dict.append(["model_source", ColumnContent, ColumnContent("Source", "str", True, False)])
auto_eval_column_dict.append(["model_category", ColumnContent, ColumnContent("Size", "str", True, False)])
auto_eval_column_dict.append(["model", ColumnContent, ColumnContent("Model Name", "markdown", True, never_hidden=True)])
#Scores
auto_eval_column_dict.append(["average_score", ColumnContent, ColumnContent("Benchmark Score (0-10)", "number", True)])
for eval_dim in EvalDimensions:
if eval_dim.value.metric in ["speed", "contamination_score"]:
auto_eval_column_dict.append([eval_dim.name, ColumnContent, ColumnContent(eval_dim.value.col_name, "number", True)])
else:
auto_eval_column_dict.append([eval_dim.name, ColumnContent, ColumnContent(eval_dim.value.col_name, "number", 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)
status = ColumnContent("status", "str", True)
# 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 EvalDimensions]
|