run_batch2.py

from __future__ import print_function
import threading
from joblib import Parallel, delayed
import Queue
import os

# Fix print 
_print = print
_rlock = threading.RLock()
def print(*args, **kwargs):
    with _rlock:
        _print(*args, **kwargs)
        
# Define number of GPUs available
N_GPU = 4

# Put indices in queue
q = Queue.Queue(maxsize=N_GPU)
for i in range(N_GPU):
  q.put(i)

def runner(x):
    gpu = q.get()
    print (x, gpu)

    # Put here your job cmd
    cmd =  "python main.py %s" % x
    os.system("CUDA_VISIBLE_DEVICES=%d %s" % (gpu, cmd))

    # return gpu id to queue
    q.put(gpu)

# Change loop    
Parallel(n_jobs=N_GPU, backend="threading")(delayed(runner)(i) for i in range(100))