random.sample() how to control reproducibility

Question

Is there a way to control random.sample()? I fix seed that standard way:

def seed_everything(seed=42):
    random.seed(seed)
    os.environ['PYTHONHASHSEED'] = str(seed)
    np.random.seed(seed)
    torch.manual_seed(seed)
    torch.cuda.manual_seed(seed)
    torch.backends.cudnn.deterministic = True
 
seed_everything(42)

Nevertheless the result of code below is different every time:

idxT=[0,1,2,3,4,5,6]
idxT = [
        idxT[j] for j in sorted(random.sample(range(len(idxT)), 3))
    ]
idxT

Answer 1

I think Ry is on the right track: if you want the return value of random.sample to be the same everytime it is called you will have to set random.seed to the same value prior to every invocation of random.sample .

Here are three simplified examples to illustrate:

random.seed(42)
idxT=[0,1,2,3,4,5,6]
for _ in range(2):
    for _ in range(3):
        print(random.sample(idxT, 3))
    print()

[5, 0, 6]
[5, 2, 1]
[1, 6, 0]

[5, 6, 4]
[0, 4, 3]
[0, 6, 5]

idxT=[0,1,2,3,4,5,6]
for _ in range(2):
    random.seed(42)
    for _ in range(3):
        print(random.sample(idxT, 3))
    print()

[5, 0, 6]
[5, 2, 1]
[1, 6, 0]

[5, 0, 6]
[5, 2, 1]
[1, 6, 0]

idxT=[0,1,2,3,4,5,6]
for _ in range(2):
    for _ in range(3):
        random.seed(42)
        print(random.sample(idxT, 3))
    print()

[5, 0, 6]
[5, 0, 6]
[5, 0, 6]

[5, 0, 6]
[5, 0, 6]
[5, 0, 6]