Hi everyone. I have an issue with the simple pipeline - it runs two similar nn training steps (tf2.3, windows10, python 3.7) with only difference is a batch size. I'm running first separately each step to have them in ClearML project page. Then I run pipeline controller, which makes a clone of each step and runs smoothly. If I run pipeline from command string again, it works Ok. However, if I clone and enqueue the pipeline, it starts, creates the clone of the fist step pending and then nothing happens. First step remains pending and doesn't start. Can anyone help with the issue? Here's the pipeline controller code:
` from clearml import Task
from clearml.automation.controller import PipelineController

Connecting ClearML with the current process,

from here on everything is logged automatically

task = Task.init(project_name='Tom', task_name='test pipeline',
task_type=Task.TaskTypes.controller, reuse_last_task_id=False)

pipe = PipelineController(default_execution_queue='default', add_pipeline_tags=False)
pipe.add_step(name='train_1st_nn_copy', base_task_project='Tom', base_task_name='train_1st_nn', parameter_override={'batch_size': 8})
pipe.add_step(name='train_2nd_nn_copy', parents=['train_1st_nn_copy', ],
base_task_project='Tom', base_task_name='train_2nd_nn',
parameter_override={'batch_size': 4})

Starting the pipeline (in the background)

pipe.start()

Wait until pipeline terminates

pipe.wait()

cleanup everything

pipe.stop()

print('done') `If I abort pipeline controller task, pending "train_1st_nn" task executes ok.