migrate evals to resource (#421)

* migrate evals to resource

* remove listing of providers's evals

* change the order of params in register

* fix after rebase

* linter fix

---------

Co-authored-by: Dinesh Yeduguru <dineshyv@fb.com>

llama_stack/providers/datatypes.py

@@ -12,7 +12,7 @@ from llama_models.schema_utils import json_schema_type
 from pydantic import BaseModel, Field
 
 from llama_stack.apis.datasets import Dataset
-from llama_stack.apis.eval_tasks import EvalTaskDef
+from llama_stack.apis.eval_tasks import EvalTask
 from llama_stack.apis.memory_banks.memory_banks import MemoryBank
 from llama_stack.apis.models import Model
 from llama_stack.apis.scoring_functions import ScoringFnDef
@@ -67,9 +67,7 @@ class ScoringFunctionsProtocolPrivate(Protocol):
 
 
 class EvalTasksProtocolPrivate(Protocol):
-    async def list_eval_tasks(self) -> List[EvalTaskDef]: ...
-
-    async def register_eval_task(self, eval_task_def: EvalTaskDef) -> None: ...
+    async def register_eval_task(self, eval_task: EvalTask) -> None: ...
 
 
 @json_schema_type

llama_stack/providers/inline/eval/meta_reference/eval.py

@@ -11,7 +11,7 @@ from .....apis.eval.eval import Eval, EvalTaskConfig, EvaluateResponse, JobStatu
 from llama_stack.apis.common.type_system import *  # noqa: F403
 from llama_stack.apis.datasetio import DatasetIO
 from llama_stack.apis.datasets import Datasets
-from llama_stack.apis.eval_tasks import EvalTaskDef
+from llama_stack.apis.eval_tasks import EvalTask
 from llama_stack.apis.inference import Inference
 from llama_stack.apis.scoring import Scoring
 from llama_stack.providers.datatypes import EvalTasksProtocolPrivate
@@ -53,15 +53,12 @@ class MetaReferenceEvalImpl(Eval, EvalTasksProtocolPrivate):
 
     async def shutdown(self) -> None: ...
 
-    async def register_eval_task(self, task_def: EvalTaskDef) -> None:
+    async def register_eval_task(self, task_def: EvalTask) -> None:
         self.eval_tasks[task_def.identifier] = task_def
 
-    async def list_eval_tasks(self) -> List[EvalTaskDef]:
-        return list(self.eval_tasks.values())
-
     async def validate_eval_input_dataset_schema(self, dataset_id: str) -> None:
-        dataset_def = await self.datasets_api.get_dataset(dataset_identifier=dataset_id)
-        if not dataset_def.dataset_schema or len(dataset_def.dataset_schema) == 0:
+        dataset_def = await self.datasets_api.get_dataset(dataset_id=dataset_id)
+        if not dataset_def.schema or len(dataset_def.schema) == 0:
             raise ValueError(f"Dataset {dataset_id} does not have a schema defined.")
 
         expected_schemas = [
@@ -77,7 +74,7 @@ class MetaReferenceEvalImpl(Eval, EvalTasksProtocolPrivate):
             },
         ]
 
-        if dataset_def.dataset_schema not in expected_schemas:
+        if dataset_def.schema not in expected_schemas:
             raise ValueError(
                 f"Dataset {dataset_id} does not have a correct input schema in {expected_schemas}"
             )

llama_stack/providers/tests/eval/test_eval.py

@@ -11,12 +11,9 @@ from llama_models.llama3.api import SamplingParams, URL
 
 from llama_stack.apis.common.type_system import ChatCompletionInputType, StringType
 
-from llama_stack.apis.datasetio.datasetio import DatasetDefWithProvider
-
 from llama_stack.apis.eval.eval import (
     AppEvalTaskConfig,
     BenchmarkEvalTaskConfig,
-    EvalTaskDefWithProvider,
     ModelCandidate,
 )
 from llama_stack.apis.scoring_functions import LLMAsJudgeScoringFnParams
@@ -70,13 +67,11 @@ class Testeval:
             "meta-reference::equality",
         ]
         task_id = "meta-reference::app_eval"
-        task_def = EvalTaskDefWithProvider(
-            identifier=task_id,
+        await eval_tasks_impl.register_eval_task(
+            eval_task_id=task_id,
             dataset_id="test_dataset_for_eval",
             scoring_functions=scoring_functions,
-            provider_id="meta-reference",
         )
-        await eval_tasks_impl.register_eval_task(task_def)
         response = await eval_impl.evaluate_rows(
             task_id=task_id,
             input_rows=rows.rows,
@@ -125,13 +120,11 @@ class Testeval:
         ]
 
         task_id = "meta-reference::app_eval-2"
-        task_def = EvalTaskDefWithProvider(
-            identifier=task_id,
+        await eval_tasks_impl.register_eval_task(
+            eval_task_id=task_id,
             dataset_id="test_dataset_for_eval",
             scoring_functions=scoring_functions,
-            provider_id="meta-reference",
         )
-        await eval_tasks_impl.register_eval_task(task_def)
         response = await eval_impl.run_eval(
             task_id=task_id,
             task_config=AppEvalTaskConfig(
@@ -169,35 +162,29 @@ class Testeval:
             pytest.skip(
                 "Only huggingface provider supports pre-registered remote datasets"
             )
-        # register dataset
-        mmlu = DatasetDefWithProvider(
-            identifier="mmlu",
-            url=URL(uri="https://huggingface.co/datasets/llamastack/evals"),
-            dataset_schema={
+
+        await datasets_impl.register_dataset(
+            dataset_id="mmlu",
+            schema={
                 "input_query": StringType(),
                 "expected_answer": StringType(),
                 "chat_completion_input": ChatCompletionInputType(),
             },
+            url=URL(uri="https://huggingface.co/datasets/llamastack/evals"),
             metadata={
                 "path": "llamastack/evals",
                 "name": "evals__mmlu__details",
                 "split": "train",
             },
-            provider_id="",
         )
 
-        await datasets_impl.register_dataset(mmlu)
-
         # register eval task
-        meta_reference_mmlu = EvalTaskDefWithProvider(
-            identifier="meta-reference-mmlu",
+        await eval_tasks_impl.register_eval_task(
+            eval_task_id="meta-reference-mmlu",
             dataset_id="mmlu",
             scoring_functions=["meta-reference::regex_parser_multiple_choice_answer"],
-            provider_id="",
         )
 
-        await eval_tasks_impl.register_eval_task(meta_reference_mmlu)
-
         # list benchmarks
         response = await eval_tasks_impl.list_eval_tasks()
         assert len(response) > 0