[Evals API][2/n] datasets / datasetio meta-reference implementation (#288)

* skeleton dataset / datasetio

* dataset datasetio

* config

* address comments

* delete dataset_utils

* address comments

* naming fix
This commit is contained in:
Xi Yan 2024-10-22 16:12:16 -07:00 committed by GitHub
parent 8a01b9e40c
commit 821810657f
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
16 changed files with 452 additions and 8 deletions

View file

@ -11,6 +11,7 @@ from llama_models.llama3.api.datatypes import * # noqa: F403
from llama_stack.apis.models import * # noqa: F403
from llama_stack.apis.shields import * # noqa: F403
from llama_stack.apis.memory_banks import * # noqa: F403
from llama_stack.apis.datasets import * # noqa: F403
from llama_stack.distribution.datatypes import * # noqa: F403
@ -27,6 +28,10 @@ async def register_object_with_provider(obj: RoutableObject, p: Any) -> None:
await p.register_shield(obj)
elif api == Api.memory:
await p.register_memory_bank(obj)
elif api == Api.datasetio:
await p.register_dataset(obj)
else:
raise ValueError(f"Unknown API {api} for registering object with provider")
Registry = Dict[str, List[RoutableObjectWithProvider]]
@ -80,6 +85,16 @@ class CommonRoutingTableImpl(RoutingTable):
add_objects(memory_banks)
elif api == Api.datasetio:
p.dataset_store = self
datasets = await p.list_datasets()
# do in-memory updates due to pesky Annotated unions
for d in datasets:
d.provider_id = pid
add_objects(datasets)
async def shutdown(self) -> None:
for p in self.impls_by_provider_id.values():
await p.shutdown()
@ -137,6 +152,7 @@ class CommonRoutingTableImpl(RoutingTable):
raise ValueError(f"Provider `{obj.provider_id}` not found")
p = self.impls_by_provider_id[obj.provider_id]
await register_object_with_provider(obj, p)
if obj.identifier not in self.registry:
@ -190,3 +206,19 @@ class MemoryBanksRoutingTable(CommonRoutingTableImpl, MemoryBanks):
self, memory_bank: MemoryBankDefWithProvider
) -> None:
await self.register_object(memory_bank)
class DatasetsRoutingTable(CommonRoutingTableImpl, Datasets):
async def list_datasets(self) -> List[DatasetDefWithProvider]:
objects = []
for objs in self.registry.values():
objects.extend(objs)
return objects
async def get_dataset(
self, dataset_identifier: str
) -> Optional[DatasetDefWithProvider]:
return self.get_object_by_identifier(identifier)
async def register_dataset(self, dataset_def: DatasetDefWithProvider) -> None:
await self.register_object(dataset_def)