phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-12-10 11:39:47 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

cleanup hardcoded dataset registry

Browse source
This commit is contained in:
Xi Yan 2024-10-14 14:19:15 -07:00
parent a9210cd416
commit 9c501d042b
3 changed files with 31 additions and 29 deletions
Download patch file Download diff file Expand all files Collapse all files

23
llama_stack/apis/evals/client.py
View file

@ -12,6 +12,7 @@ import httpx
from termcolor import cprint
from .evals import * # noqa: F403
from ..datasets.client import DatasetsClient
class EvaluationClient(Evals):
@ -54,13 +55,31 @@ class EvaluationClient(Evals):
async def run_main(host: str, port: int):
client = EvaluationClient(f"http://{host}:{port}")
dataset_client = DatasetsClient(f"http://{host}:{port}")
# Custom Eval Task
# 1. register custom dataset
response = await dataset_client.create_dataset(
dataset_def=CustomDatasetDef(
identifier="mmlu-simple-eval-en",
url="https://openaipublic.blob.core.windows.net/simple-evals/mmlu.csv",
),
)
cprint(f"datasets/create: {response}", "cyan")
# 2. run evals on the registered dataset
response = await client.run_evals(
model="Llama3.1-8B-Instruct",
dataset="mmlu-simple-eval-en",
task="mmlu",
)
if response.formatted_report:
cprint(response.formatted_report, "green")
else:
cprint(f"Response: {response}", "green")
# Eleuther Eval Task
# response = await client.run_evals(
# model="Llama3.1-8B-Instruct",
@ -70,10 +89,6 @@ async def run_main(host: str, port: int):
# n_samples=2,
# ),
# )
if response.formatted_report:
cprint(response.formatted_report, "green")
else:
cprint(f"Response: {response}", "green")
def main(host: str, port: int):