phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-06-28 02:53:30 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Folder restructure for evals/datasets/scoring (#419)

Browse source 
* rename evals related stuff

* fix datasetio

* fix scoring test

* localfs -> LocalFS

* refactor scoring

* refactor scoring

* remove 8b_correctness scoring_fn from tests

* tests w/ eval params

* scoring fn braintrust fixture

* import
This commit is contained in:
Xi Yan 2024-11-11 17:35:40 -05:00 • committed by GitHub
parent 2b7d70ba86
commit b4416b72fd
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
37 changed files with 141 additions and 100 deletions
Download patch file Download diff file Expand all files Collapse all files

20
llama_stack/providers/tests/eval/test_eval.py
View file

@ -19,9 +19,10 @@ from llama_stack.apis.eval.eval import (
EvalTaskDefWithProvider,
ModelCandidate,
)
from llama_stack.apis.scoring_functions import LLMAsJudgeScoringFnParams
from llama_stack.distribution.datatypes import Api
from llama_stack.providers.tests.datasetio.test_datasetio import register_dataset
from .constants import JUDGE_PROMPT
# How to run this test:
#
@ -65,7 +66,7 @@ class Testeval:
assert len(rows.rows) == 3
scoring_functions = [
"meta-reference::llm_as_judge_8b_correctness",
"meta-reference::llm_as_judge_base",
"meta-reference::equality",
]
task_id = "meta-reference::app_eval"
@ -85,11 +86,22 @@ class Testeval:
model="Llama3.2-3B-Instruct",
sampling_params=SamplingParams(),
),
scoring_params={
"meta-reference::llm_as_judge_base": LLMAsJudgeScoringFnParams(
judge_model="Llama3.1-8B-Instruct",
prompt_template=JUDGE_PROMPT,
judge_score_regexes=[
r"Total rating: (\d+)",
r"rating: (\d+)",
r"Rating: (\d+)",
],
)
},
),
)
assert len(response.generations) == 3
assert "meta-reference::llm_as_judge_8b_correctness" in response.scores
assert "meta-reference::equality" in response.scores
assert "meta-reference::llm_as_judge_base" in response.scores
@pytest.mark.asyncio
async def test_eval_run_eval(self, eval_stack):
@ -109,7 +121,6 @@ class Testeval:
)
scoring_functions = [
"meta-reference::llm_as_judge_8b_correctness",
"meta-reference::subset_of",
]
@ -138,7 +149,6 @@ class Testeval:
assert eval_response is not None
assert len(eval_response.generations) == 5
assert "meta-reference::subset_of" in eval_response.scores
assert "meta-reference::llm_as_judge_8b_correctness" in eval_response.scores
@pytest.mark.asyncio
async def test_eval_run_benchmark_eval(self, eval_stack):