How to restore the output files of each step in the remote machine when using the DispatcherExecutor?

[gdbhcxhmjk-z]

The overall workflow can be executed successfully but the output files of each step can only be downloaded from the website "workflow.deepmodeling.com". How can I retore the output files of each step in the remote machine?
All the computing results will be cleaned when the calculation is finished. I'm wondering if there's a way to save the files in the appointed file instead of the temporary files with the hash name.
Here is my DispatcherExecutor:
DispatcherExecutor(
host=host,
port=port,
username=username,
password=password,
remote_root="/data/home/tmp",
image_pull_policy="IfNotPresent",
machine_dict={
"batch_type": "slurm",
"local_root": "/data/home/usr",
"context_type": "SSHContext",
"clean_asynchronously":false},
resources_dict={
"number_node": 1,
"cpu_per_node": 10,
"gpu_per_node": 1,
"queue_name": "gpu",
"group_size": 1,
"custom_flags": [
"#SBATCH --time=0-1000:00:00"
],
"source_list": ["activate deepmd-kit"],
"batch_type": "Slurm"},
merge_sliced_step=False
)
The parameter "local_root" doesn't help. There is nothing in the local_root.

[zjgemi]

There is an argument clean in DispatcherExecutor. Set it to False to preventing cleaning remote working directories.