Random sample without replacement while maintaining natural order of tabular data

I have time series data which is not monotonically increasing, so calling sort/shuffle is out of the question.

I want to randomly pull out n% of the data, while maintaining it relative order, to act as validation or test set, which can be shown as:

my_ndarray = [ 1, 20, 10, 3, 90, 5, 80, 50, 4, 1] # (number of samples = 1645, number of timesteps = 10, number of features = 7)
# custom_train_test_split()
train = [1, 20, 90, 5, 50, 4, 1]
valid = [10, 3, 80]

I would appreciate some guidance on how to do this efficiently. To my understanding Java style iteration is inefficient in Python. I suspect a 3D boolean table mask would be the pythonic and vectorized way.

Solution

Here is what the solution may look like:

Add a temporary additional dimension to the array, in which you add indices to each item in the array.
Shuffle the array.
Take the desired portions of the array, then sort each of them by the dimension.
Remove the temporary dimension from the chosen portions.

Here is the solution using plain Python lists:

my_ndarray = [ 1, 20, 10, 3, 90, 5, 80, 50, 4, 1] 
# Add temporary dimension by converting each item 
# to a sublist, where the index is the first element of each sublist
nda=[[i,my_ndarray[i]] for i in len(my_ndarray)]
np.random.shuffle(nda)
# Training data is the first 7 items
traindata=nda[0:7]
traindata.sort()
traindata=[x[1] for x in traindata]
# Test data is the rest
testdata=nda[7:10]
testdata.sort()
testdata=[x[1] for x in testdata]