Open
Description
Describe the bug
num samples metric_option is not respected
To Reproduce
regardless of options set on task definition and metric settings, it's not generating multiple samples
Expected behavior
setting
metric_options={
"extractive_match": {
"num_samples": 16,
},
or
# Define tasks
task = LightevalTaskConfig(
num_samples=NUM_SAMPLES, ...
)
should respect the setting
Version info
install
conda create -n eval python=3.11
pip install vllm==0.7.2
pip install git+https://github.com/huggingface/lighteval.git#egg=lighteval math-verify==0.5.2