Rasa test nlu produces wrong partitions on using the percentages argument

[AMR-KELEG]

Rasa version: rasa==1.10.4

Python version: Python 3.6.9

Operating system (windows, osx, ...): Ubuntu 18

Issue:
Currently rasa test nlu allows specifying percentages of the training dataset to be excluded while training using the -P flag. This isn't currently done correctly.
E.g: If the percentages were -P 50 25 then you would expect that the number of training samples would be 500 and 250 if the original training dataset size was 1000 samples.
BUT, the way the function currently work will generate number of training samples of 500 and 125 samples.

rasa/rasa/nlu/test.py

Lines 1848 to 1865 in 4830308

	train, test = data.train_test_split()
	io_utils.write_text_file(test.nlu_as_markdown(), test_path)
	for percentage in exclusion_percentages:
	percent_string = f"{percentage}%_exclusion"
	_, train = train.train_test_split(percentage / 100)
	# only count for the first run and ignore the others
	if run == 0:
	training_examples_per_run.append(len(train.training_examples))
	model_output_path = os.path.join(run_path, percent_string)
	train_split_path = os.path.join(model_output_path, "train")
	train_nlu_split_path = os.path.join(train_split_path, TRAIN_DATA_FILE)
	train_nlg_split_path = os.path.join(train_split_path, NLG_DATA_FILE)
	io_utils.create_path(train_nlu_split_path)
	io_utils.write_text_file(train.nlu_as_markdown(), train_nlu_split_path)
	io_utils.write_text_file(train.nlg_as_markdown(), train_nlg_split_path)

Command or request that led to error:

rasa test nlu -c config.yml config_dup.yml -r 1 -p 50 25

I am willing to submit a PR to solve this so kindly assign the issue to me.

Thanks

[sara-tagger]

Thanks for the issue, @koaning will get back to you about it soon!

You may find help in the docs and the forum, too 🤗

[koaning]

Just to check, are you using tensorboard in your runs?

[AMR-KELEG]

Just to check, are you using tensorboard in your runs?

Nope.

[tabergma]

@AMR-KELEG Thanks for catching this! Seem like you are correct. And thanks for create a PR already. We will give it a review soon.