YummyWhale40 you mean like continue training?
YummyWhale40 no idea what the pytorch-lighting guys did there. let me check a the actual code.
maybe the arguments is simply passed to
self._trains = Task.init( project_name=project_name, task_name=task_name, task_type=task_type, reuse_last_task_id=reuse_last_task_id, output_uri=output_uri, auto_connect_arg_parser=auto_connect_arg_parser, auto_connect_frameworks=auto_connect_frameworks, auto_resource_monitoring=auto_resource_monitoring )
I would like to confirm just in case.
In the desired behavior,
reuse_last_task_id=True forces it for any intervals?
oh I got it. my codes output models and the task catch it automatically.
In my case, I write codes and run single batch train-val, which contains model saving, in developing phase. I want TRAINS to overwrite the dev runs for keeping dashboard clean.
I don't mean continuous training but I want to know about your plans for it 😋
if you have any idea to reuse id even if models are outputted, please tell me thx
YummyWhale40 from the code snippet, it seems like the argument is passed.
"reuse_last_task_id=True" is the default, and it means that if the previous run of the task did not create any artifacts/models and was executed 72 hours ago (configurable), The Task will be reset (i.e. all logs cleared) and will be reused in the current run.