Klaus Greff

1. But unless you want to modify the global scope in some hacky way, you would have to do `x, y, z = ex.post_capture([x, y, z])` no? And that you...

Observers are [sorted by priority](https://github.com/IDSIA/sacred/blob/master/sacred/run.py#L212) and the observer with the highest priority gets to choose the `_id`. The problem in your case is that you are adding a `MongoObserver` later...

@JarnoRFB Documentation on this could be a bit better I guess. Questions about the behavior of `_id` are rather common. Maybe we should add a section detailing the `_id` system...

Overwriting the `info` dict is intended behavior, but it is obviously really bad for your usecase. Resuming experiments seems like a common usecase, and I agree that we need a...

> I understand that overwrite should lead to everything being overwritten, but why does the info dict end up being empty in the second run? Hmm, strange. For me it...

Hey @JarnoRFB. I am very sorry about my lack of support recently. It is bad timing for you, but I am afraid, that I have too much going on at...

My workflow for doing exactly this is the following: ``` python from my_experiment import ex, create_model from pymongo import MongoClient client = MongoClient() run_entry = client.sacred.find_one({ 'experiment.name': 'train_area_models_4x4', 'config.use_conditional': False,...

This is definitely a common usecase, which is not adequately supported. Here's how I do it: ``` python from my_experiment import ex, my_captured_function r = ex._create_run(named_configs=['path/to/your/config.json']) # at this point...

Good point. I agree that the stacktrace is useless and maybe even harmful in these cases. But I am not sure how I feel about printing usage information. At least...

@thequilo These suggestions all sound very useful to me, and I wholeheartedly support each one of them. Some of them seem quite tricky to implement. But if you decide to...