unexpected interaction between numpy's random number generator and initiating a ClearML Task

[d13g0]

Describe the bug

A clear and concise description of what the bug is.

Starting a task interferes with numpy's RGN

To reproduce

Exact steps to reproduce the bug. Provide example code if possible.

import numpy as np
import uuid
from clearml import Task, Logger

def get_experiment_id():
    return str(uuid.uuid1())[:8]

def run():
    indices = np.arange(100)

    exp_id = get_experiment_id()

    task = Task.init(project_name='auto-dummy',
                     task_name=f'exp-{exp_id}',
                     reuse_last_task_id=False, 
                     continue_last_task=False,
                     tags=['demo']
                     )
    task.connect({
        'a':1
    })

    np.random.shuffle(indices)
    print(indices)

if __name__=='__main__':
    run()

Expected behaviour

What is the expected behaviour? What should've happened but didn't?

the numbers in the array should appear in a different order on consecutive runs. However the array is always shuffled the exact same way.

If I do the shuffling before doing Task.init then it works.

Environment

• Server type (self hosted \ app.clear.ml)
• ClearML SDK Version = 1.12.0
• ClearML Server Version (Only for self hosted). Can be found on the bottom right corner of the settings screen. = N/A
• Python Version = 3.10.12
• OS (Windows \ Linux \ Macos) = Mac

Related Discussion

If this continues a slack thread, please provide a link to the original slack thread.

[jkhenning]

Hi @d13g0,

This behavior is by design, see here: https://clear.ml/docs/latest/docs/clearml_sdk/task_sdk/#setting-random-seed

[d13g0]

Thank you so much Jake! 🙏