[Bug] OpenICLEvalTask does not combine results split by NumWorkerPartitioner

Prerequisite

• I have searched Issues and Discussions but cannot get the expected help.
• The bug has not been fixed in the latest version.

Type

I'm evaluating with the officially supported tasks/models/datasets.

Environment

eval = dict(
    partitioner=dict(
       type=NumWorkerPartitioner,
       num_worker=64
   ),
   runner=dict(type=LocalRunner,
               max_num_workers=64,
              task=dict(type=OpenICLEvalTask)),
)

This eval config causes the eval phase resulting in individual split files. The files have the same name but with increasing numbering postfixes. Each splits has its very own accuracy calculated by evaluating the samples from within.

This separation caused difficulty in summing up the accuracy scores, as the samples in the last file is not the same with the rest.

e.g.

data_1.json
data_2.json
data_3.json

data1.json

{
"accuracy": 90,
"details":[

...
]
}

Reproduces the problem - code/configuration sample

python run.py examples/evaluation_script.py

Reproduces the problem - command or script

python run.py examples/evaluation_script.py

Reproduces the problem - error message

the summarizer can not find the desired result file, so the score is empty.

Other information

No response