Should sequence nature of input data within a mini-batch be maintained? - deep-learning

Question

Suppose a multivariate time series prediction problem for the following data
rows = 2000
cols = 6
data = np.arange(int(rows*cols)).reshape(-1,rows).transpose()
print(data[0:20])
print('\n {} \n'.format(data.shape))
print(data[-20:])
which prints out following data
[[ 0 2000 4000 6000 8000 10000]
[ 1 2001 4001 6001 8001 10001]
[ 2 2002 4002 6002 8002 10002]
[ 3 2003 4003 6003 8003 10003]
[ 4 2004 4004 6004 8004 10004]
[ 5 2005 4005 6005 8005 10005]
[ 6 2006 4006 6006 8006 10006]
[ 7 2007 4007 6007 8007 10007]
[ 8 2008 4008 6008 8008 10008]
[ 9 2009 4009 6009 8009 10009]
[ 10 2010 4010 6010 8010 10010]
[ 11 2011 4011 6011 8011 10011]
[ 12 2012 4012 6012 8012 10012]
[ 13 2013 4013 6013 8013 10013]
[ 14 2014 4014 6014 8014 10014]
[ 15 2015 4015 6015 8015 10015]
[ 16 2016 4016 6016 8016 10016]
[ 17 2017 4017 6017 8017 10017]
[ 18 2018 4018 6018 8018 10018]
[ 19 2019 4019 6019 8019 10019]]
(2000, 6)
[[ 1980 3980 5980 7980 9980 11980]
[ 1981 3981 5981 7981 9981 11981]
[ 1982 3982 5982 7982 9982 11982]
[ 1983 3983 5983 7983 9983 11983]
[ 1984 3984 5984 7984 9984 11984]
[ 1985 3985 5985 7985 9985 11985]
[ 1986 3986 5986 7986 9986 11986]
[ 1987 3987 5987 7987 9987 11987]
[ 1988 3988 5988 7988 9988 11988]
[ 1989 3989 5989 7989 9989 11989]
[ 1990 3990 5990 7990 9990 11990]
[ 1991 3991 5991 7991 9991 11991]
[ 1992 3992 5992 7992 9992 11992]
[ 1993 3993 5993 7993 9993 11993]
[ 1994 3994 5994 7994 9994 11994]
[ 1995 3995 5995 7995 9995 11995]
[ 1996 3996 5996 7996 9996 11996]
[ 1997 3997 5997 7997 9997 11997]
[ 1998 3998 5998 7998 9998 11998]
[ 1999 3999 5999 7999 9999 11999]]
Consider the last column is the target time series to be predicted. If the batch-size is 20, Can I randomly skip some points e.g. 1009, 1012, 1015 from the first batch during training?
If yes, this mean we can randomly choose points within a time series to be used as training and test?