wehub-resource-sync
e768098d0e
Flake8 Lint / flake8 (push) Waiting to run
Spell check CI / Spell_Check (push) Waiting to run
tools_continuous_delivery / Private PyPI non-main branch release (push) Has been skipped
tools_continuous_delivery / Private PyPI main branch release (push) Failing after 2m42s
Publish Promptflow Doc / Build (push) Has been cancelled
Publish Promptflow Doc / Deploy (push) Has been cancelled
55 lines
2.1 KiB
Python
55 lines
2.1 KiB
Python
import asyncio
|
|
from dataclasses import dataclass, field
|
|
from typing import Any, Callable, Dict, List, Optional
|
|
|
|
|
|
@dataclass
|
|
class EvalResult:
|
|
per_row_outputs: List[Any]
|
|
metrics: Dict[str, Any]
|
|
errors: List[tuple] = field(default_factory=list)
|
|
|
|
|
|
class EvalRunner:
|
|
def __init__(self, workflow_factory: Callable[[], Any], aggregate_fn: Callable[..., dict],
|
|
concurrency: int = 5, input_mapping: Optional[Dict[str, str]] = None):
|
|
self._workflow_factory = workflow_factory
|
|
self._aggregate_fn = aggregate_fn
|
|
self._concurrency = concurrency
|
|
self._input_mapping = input_mapping
|
|
|
|
async def run(self, dataset: List[Any]) -> EvalResult:
|
|
semaphore = asyncio.Semaphore(self._concurrency)
|
|
per_row_outputs: List[Any] = [None] * len(dataset)
|
|
errors: List[tuple] = []
|
|
|
|
async def _run_row(index: int, row: Any) -> None:
|
|
async with semaphore:
|
|
wf = self._workflow_factory()
|
|
result = await wf.run(row)
|
|
per_row_outputs[index] = result.get_outputs()[0]
|
|
|
|
tasks = [_run_row(i, row) for i, row in enumerate(dataset)]
|
|
results = await asyncio.gather(*tasks, return_exceptions=True)
|
|
succeeded_outputs: List[Any] = []
|
|
for i, r in enumerate(results):
|
|
if isinstance(r, Exception):
|
|
errors.append((i, r))
|
|
else:
|
|
succeeded_outputs.append(per_row_outputs[i])
|
|
|
|
aggregation_inputs = self._transpose(succeeded_outputs)
|
|
if self._input_mapping:
|
|
aggregation_inputs = {self._input_mapping.get(k, k): v for k, v in aggregation_inputs.items()}
|
|
metrics = self._aggregate_fn(**aggregation_inputs)
|
|
return EvalResult(per_row_outputs=succeeded_outputs, metrics=metrics, errors=errors)
|
|
|
|
@staticmethod
|
|
def _transpose(outputs: List[Any]) -> Dict[str, Any]:
|
|
if not outputs:
|
|
return {"values": []}
|
|
if not isinstance(outputs[0], dict):
|
|
return {"values": outputs}
|
|
keys = outputs[0].keys()
|
|
return {k: [o[k] for o in outputs] for k in keys}
|