phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-12-09 19:29:18 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

generator + scorer Api for MMLU

Browse source
This commit is contained in:
Xi Yan 2024-10-13 23:27:02 -07:00
parent fb565dfb06
commit a25aff290e
14 changed files with 618 additions and 131 deletions
Download patch file Download diff file Expand all files Collapse all files

25
llama_stack/apis/evals/client.py
View file

@ -33,7 +33,7 @@ class EvaluationClient(Evals):
) -> EvaluateResponse:
async with httpx.AsyncClient() as client:
response = await client.post(
f"{self.base_url}/evals/run",
f"{self.base_url}/evals/run_eval_task",
json={
"model": model,
"task": task,
@ -55,28 +55,25 @@ async def run_main(host: str, port: int):
client = EvaluationClient(f"http://{host}:{port}")
# Custom Eval Task
response = await client.run_evals(
model="Llama3.1-8B-Instruct",
dataset="mmlu-simple-eval-en",
task="mmlu",
)
# Eleuther Eval Task
# response = await client.run_evals(
# model="Llama3.1-8B-Instruct",
# dataset="mmlu-simple-eval-en",
# task="mmlu",
# # task="meta_mmlu_pro_instruct",
# task="meta_ifeval",
# eval_task_config=EvaluateTaskConfig(
# n_samples=2,
# ),
# )
# Eleuther Eval Task
response = await client.run_evals(
model="Llama3.1-8B-Instruct",
# task="meta_mmlu_pro_instruct",
task="meta_ifeval",
eval_task_config=EvaluateTaskConfig(
n_samples=2,
),
)
if response.formatted_report:
cprint(response.formatted_report, "green")
else:
cprint(f"evaluate response={response}", "green")
cprint(f"Response: {response}", "green")
def main(host: str, port: int):