Spaces:
Running
Running
File size: 1,855 Bytes
af2acd4 c671908 af2acd4 1edd506 af2acd4 326ac2a 1edd506 326ac2a af2acd4 c671908 1edd506 c671908 1edd506 c671908 1edd506 c671908 |
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 |
from dataclasses import dataclass
import streamlit as st
from huggingface_hub import DatasetFilter, HfApi
from huggingface_hub.hf_api import DatasetInfo
@dataclass(frozen=True, eq=True)
class EvaluationInfo:
task: str
model: str
dataset_name: str
dataset_config: str
dataset_split: str
metrics: set
def compute_evaluation_id(dataset_info: DatasetInfo) -> int:
if dataset_info.cardData is not None:
metadata = dataset_info.cardData["eval_info"]
metadata.pop("col_mapping", None)
# TODO(lewtun): populate dataset cards with metric info
if "metrics" not in metadata:
metadata["metrics"] = frozenset()
metadata["metrics"] = frozenset(metadata["metrics"])
evaluation_info = EvaluationInfo(**metadata)
return hash(evaluation_info)
else:
return None
def get_evaluation_ids():
filt = DatasetFilter(author="autoevaluate")
evaluation_datasets = HfApi().list_datasets(filter=filt, full=True)
return [compute_evaluation_id(dset) for dset in evaluation_datasets]
def filter_evaluated_models(models, task, dataset_name, dataset_config, dataset_split, metrics):
evaluation_ids = get_evaluation_ids()
for idx, model in enumerate(models):
evaluation_info = EvaluationInfo(
task=task,
model=model,
dataset_name=dataset_name,
dataset_config=dataset_config,
dataset_split=dataset_split,
metrics=frozenset(metrics),
)
candidate_id = hash(evaluation_info)
if candidate_id in evaluation_ids:
st.info(
f"Model `{model}` has already been evaluated on this configuration. \
This model will be excluded from the evaluation job..."
)
models.pop(idx)
return models
|