python3 demo_mnist.py --mnist_data_path ./ --num_epoch 5 --split_shape 2 10 --train_epsilon 2

Round 1/10 modelling:
 epoch 1/5 - curr/avg acc: 0.937500/0.764717                - curr/avg loss: 0.065108/0.434133, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.918600                    - curr/avg unique acc: 0.583333/0.793312, [   79/   79]

 epoch 2/5 - curr/avg acc: 0.843750/0.939783                - curr/avg loss: 0.149960/0.105769, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.937500                    - curr/avg unique acc: 0.904762/0.855651, [   79/   79]

 epoch 3/5 - curr/avg acc: 0.875000/0.955400                - curr/avg loss: 0.157828/0.077489, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.947200                    - curr/avg unique acc: 0.904762/0.854722, [   79/   79]

 epoch 4/5 - curr/avg acc: 0.937500/0.962433                - curr/avg loss: 0.057682/0.064954, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.951400                    - curr/avg unique acc: 0.904762/0.852575, [   79/   79]

 epoch 5/5 - curr/avg acc: 1.000000/0.967100                - curr/avg loss: 0.010077/0.056886, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.952500                    - curr/avg unique acc: 0.904762/0.873066, [   79/   79]

unsupervised cluster results of random variable 0 is:
     0     1     2    3    4    5    6     7    8    9
0    4  1127    19   32  974   16   13  1017   67  983
1  976     8  1013  978    8  876  945    11  907   26
accs: [0.9525], mean: 0.9525, std: 0.0.
Round 2/10 modelling:
 epoch 1/5 - curr/avg acc: 0.875000/0.759767                - curr/avg loss: 0.204678/0.462099, [  938/  938]

 prediction - curr/avg acc: 0.875000/0.902900                    - curr/avg unique acc: 0.696970/0.770836, [   79/   79]

 epoch 2/5 - curr/avg acc: 0.968750/0.925133                - curr/avg loss: 0.078313/0.140312, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.923400                    - curr/avg unique acc: 0.904762/0.810233, [   79/   79]

 epoch 3/5 - curr/avg acc: 0.968750/0.944017                - curr/avg loss: 0.043446/0.102822, [  938/  938]

 prediction - curr/avg acc: 0.875000/0.934000                    - curr/avg unique acc: 0.696970/0.838496, [   79/   79]

 epoch 4/5 - curr/avg acc: 0.968750/0.953833                - curr/avg loss: 0.073088/0.083013, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.940400                    - curr/avg unique acc: 1.000000/0.843789, [   79/   79]

 epoch 5/5 - curr/avg acc: 1.000000/0.960800                - curr/avg loss: 0.048570/0.071524, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.948500                    - curr/avg unique acc: 1.000000/0.876297, [   79/   79]

unsupervised cluster results of random variable 0 is:
     0     1    2    3    4    5    6     7    8    9
0   16     2   40   45  970   34    8  1007   23  973
1  964  1133  992  965   12  858  950    21  951   36
accs: [0.9525, 0.9485], mean: 0.9505, std: 0.0020000000000000018.
Round 3/10 modelling:
 epoch 1/5 - curr/avg acc: 0.875000/0.759850                - curr/avg loss: 0.191088/0.464887, [  938/  938]

 prediction - curr/avg acc: 0.875000/0.900300                    - curr/avg unique acc: 0.523810/0.701187, [   79/   79]

 epoch 2/5 - curr/avg acc: 0.968750/0.921367                - curr/avg loss: 0.113671/0.132884, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.923000                    - curr/avg unique acc: 1.000000/0.770069, [   79/   79]

 epoch 3/5 - curr/avg acc: 0.937500/0.940233                - curr/avg loss: 0.069024/0.093764, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.938300                    - curr/avg unique acc: 1.000000/0.838258, [   79/   79]

 epoch 4/5 - curr/avg acc: 0.937500/0.950333                - curr/avg loss: 0.051732/0.078558, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.946900                    - curr/avg unique acc: 0.904762/0.849690, [   79/   79]

 epoch 5/5 - curr/avg acc: 0.906250/0.958767                - curr/avg loss: 0.103760/0.067187, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.948700                    - curr/avg unique acc: 1.000000/0.868269, [   79/   79]

unsupervised cluster results of random variable 0 is:
     0     1     2    3    4    5    6     7    8    9
0    2  1119    28   15  964   11   14  1008   74  978
1  978    16  1004  995   18  881  944    20  900   31
accs: [0.9525, 0.9485, 0.9487], mean: 0.9499, std: 0.001840289832245641.
Round 4/10 modelling:
 epoch 1/5 - curr/avg acc: 0.937500/0.743833                - curr/avg loss: 0.088722/0.477379, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.913800                    - curr/avg unique acc: 0.904762/0.792173, [   79/   79]

 epoch 2/5 - curr/avg acc: 1.000000/0.935867                - curr/avg loss: 0.038728/0.110775, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.936600                    - curr/avg unique acc: 0.904762/0.830871, [   79/   79]

 epoch 3/5 - curr/avg acc: 0.937500/0.951150                - curr/avg loss: 0.052349/0.082316, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.944500                    - curr/avg unique acc: 0.714286/0.849540, [   79/   79]

 epoch 4/5 - curr/avg acc: 0.968750/0.959200                - curr/avg loss: 0.060864/0.067421, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.945900                    - curr/avg unique acc: 0.904762/0.845595, [   79/   79]

 epoch 5/5 - curr/avg acc: 0.906250/0.965750                - curr/avg loss: 0.133120/0.058295, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.951600                    - curr/avg unique acc: 1.000000/0.872535, [   79/   79]

unsupervised cluster results of random variable 0 is:
     0     1     2    3    4    5    6     7    8    9
0    1  1118    17   17  962   16    7  1007   21  974
1  979    17  1015  993   20  876  951    21  953   35
accs: [0.9525, 0.9485, 0.9487, 0.9516], mean: 0.950325, std: 0.0017555269864060808.
Round 5/10 modelling:
 epoch 1/5 - curr/avg acc: 0.875000/0.737683                - curr/avg loss: 0.223500/0.488425, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.890000                    - curr/avg unique acc: 0.791667/0.732396, [   79/   79]

 epoch 2/5 - curr/avg acc: 0.968750/0.919850                - curr/avg loss: 0.066743/0.142979, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.921400                    - curr/avg unique acc: 0.714286/0.792969, [   79/   79]

 epoch 3/5 - curr/avg acc: 0.843750/0.939933                - curr/avg loss: 0.135009/0.107406, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.930300                    - curr/avg unique acc: 0.904762/0.823504, [   79/   79]

 epoch 4/5 - curr/avg acc: 0.906250/0.950433                - curr/avg loss: 0.235836/0.088654, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.938900                    - curr/avg unique acc: 1.000000/0.838093, [   79/   79]

 epoch 5/5 - curr/avg acc: 1.000000/0.956767                - curr/avg loss: 0.008683/0.076135, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.943100                    - curr/avg unique acc: 0.824561/0.842861, [   79/   79]

unsupervised cluster results of random variable 0 is:
     0     1     2     3    4    5    6    7    8    9
0    7  1133  1003  1000   23   36   18  176  926   40
1  973     2    29    10  959  856  940  852   48  969
accs: [0.9525, 0.9485, 0.9487, 0.9516, 0.9431], mean: 0.94888, std: 0.00328901200970746.
Round 6/10 modelling:
 epoch 1/5 - curr/avg acc: 0.906250/0.726550                - curr/avg loss: 0.199102/0.538939, [  938/  938]

 prediction - curr/avg acc: 0.875000/0.895100                    - curr/avg unique acc: 0.473684/0.746510, [   79/   79]

 epoch 2/5 - curr/avg acc: 0.875000/0.924717                - curr/avg loss: 0.264351/0.136592, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.934100                    - curr/avg unique acc: 0.791667/0.800556, [   79/   79]

 epoch 3/5 - curr/avg acc: 0.968750/0.945733                - curr/avg loss: 0.052807/0.093945, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.944400                    - curr/avg unique acc: 0.756098/0.835326, [   79/   79]

 epoch 4/5 - curr/avg acc: 0.937500/0.955767                - curr/avg loss: 0.090666/0.077681, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.943800                    - curr/avg unique acc: 0.848485/0.862631, [   79/   79]

 epoch 5/5 - curr/avg acc: 0.937500/0.961767                - curr/avg loss: 0.080417/0.067327, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.946300                    - curr/avg unique acc: 0.756098/0.846233, [   79/   79]

unsupervised cluster results of random variable 0 is:
     0     1     2    3    4    5    6    7    8    9
0  972    27  1009   19  974   61  956   55  197   71
1    8  1108    23  991    8  831    2  973  777  938
accs: [0.9525, 0.9485, 0.9487, 0.9516, 0.9431, 0.9463], mean: 0.9484499999999999, std: 0.003152644392675231.
Round 7/10 modelling:
 epoch 1/5 - curr/avg acc: 0.875000/0.683033                - curr/avg loss: 0.252353/0.580256, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.899700                    - curr/avg unique acc: 0.655172/0.697894, [   79/   79]

 epoch 2/5 - curr/avg acc: 0.968750/0.924633                - curr/avg loss: 0.109435/0.132388, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.925900                    - curr/avg unique acc: 0.756098/0.810338, [   79/   79]

 epoch 3/5 - curr/avg acc: 0.937500/0.945800                - curr/avg loss: 0.118879/0.092939, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.939700                    - curr/avg unique acc: 1.000000/0.816670, [   79/   79]

 epoch 4/5 - curr/avg acc: 0.968750/0.956167                - curr/avg loss: 0.065038/0.077464, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.940300                    - curr/avg unique acc: 0.791667/0.829864, [   79/   79]

 epoch 5/5 - curr/avg acc: 0.937500/0.961500                - curr/avg loss: 0.072438/0.068292, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.947800                    - curr/avg unique acc: 0.904762/0.860490, [   79/   79]

unsupervised cluster results of random variable 0 is:
     0     1     2    3    4    5    6    7    8    9
0  974    18  1021  993   19  884  948   31  955   31
1    6  1117    11   17  963    8   10  997   19  978
accs: [0.9525, 0.9485, 0.9487, 0.9516, 0.9431, 0.9463, 0.9478], mean: 0.9483571428571428, std: 0.0029276305108676525.
Round 8/10 modelling:
 epoch 1/5 - curr/avg acc: 0.906250/0.734450                - curr/avg loss: 0.172349/0.507811, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.902100                    - curr/avg unique acc: 0.649123/0.721199, [   79/   79]

 epoch 2/5 - curr/avg acc: 0.968750/0.926317                - curr/avg loss: 0.053999/0.128218, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.926000                    - curr/avg unique acc: 0.824561/0.827848, [   79/   79]

 epoch 3/5 - curr/avg acc: 0.968750/0.946383                - curr/avg loss: 0.114458/0.093271, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.932200                    - curr/avg unique acc: 0.824561/0.846955, [   79/   79]

 epoch 4/5 - curr/avg acc: 0.937500/0.955650                - curr/avg loss: 0.087148/0.077452, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.940400                    - curr/avg unique acc: 1.000000/0.869075, [   79/   79]

 epoch 5/5 - curr/avg acc: 0.906250/0.962200                - curr/avg loss: 0.071269/0.066439, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.946800                    - curr/avg unique acc: 1.000000/0.871889, [   79/   79]

unsupervised cluster results of random variable 0 is:
     0     1    2    3    4    5    6     7    8    9
0    8  1112   45   22  968   15   13  1015   24  974
1  972    23  987  988   14  877  945    13  950   35
accs: [0.9525, 0.9485, 0.9487, 0.9516, 0.9431, 0.9463, 0.9478, 0.9468], mean: 0.9481625, std: 0.0027865469222677643.
Round 9/10 modelling:
 epoch 1/5 - curr/avg acc: 0.812500/0.746700                - curr/avg loss: 0.202877/0.467570, [  938/  938]

 prediction - curr/avg acc: 0.875000/0.893500                    - curr/avg unique acc: 0.791667/0.767033, [   79/   79]

 epoch 2/5 - curr/avg acc: 0.968750/0.925700                - curr/avg loss: 0.051613/0.130480, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.929200                    - curr/avg unique acc: 0.791667/0.831789, [   79/   79]

 epoch 3/5 - curr/avg acc: 1.000000/0.946967                - curr/avg loss: 0.033066/0.094040, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.937600                    - curr/avg unique acc: 0.791667/0.868467, [   79/   79]

 epoch 4/5 - curr/avg acc: 0.968750/0.958017                - curr/avg loss: 0.073920/0.074112, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.944300                    - curr/avg unique acc: 0.824561/0.882899, [   79/   79]

 epoch 5/5 - curr/avg acc: 1.000000/0.965317                - curr/avg loss: 0.024808/0.061461, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.954300                    - curr/avg unique acc: 0.904762/0.895687, [   79/   79]

unsupervised cluster results of random variable 0 is:
     0     1     2    3    4    5    6    7    8    9
0  975     5    26   20  975   54  952  999   57  970
1    5  1130  1006  990    7  838    6   29  917   39
accs: [0.9525, 0.9485, 0.9487, 0.9516, 0.9431, 0.9463, 0.9478, 0.9468, 0.9543], mean: 0.9488444444444445, std: 0.0032592129626341482.
Round 10/10 modelling:
 epoch 1/5 - curr/avg acc: 0.937500/0.716217                - curr/avg loss: 0.106335/0.531104, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.905100                    - curr/avg unique acc: 0.904762/0.734539, [   79/   79]

 epoch 2/5 - curr/avg acc: 0.937500/0.927633                - curr/avg loss: 0.100228/0.121998, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.930300                    - curr/avg unique acc: 0.791667/0.792071, [   79/   79]

 epoch 3/5 - curr/avg acc: 0.875000/0.945967                - curr/avg loss: 0.182634/0.091346, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.937400                    - curr/avg unique acc: 0.904762/0.820408, [   79/   79]

 epoch 4/5 - curr/avg acc: 1.000000/0.955083                - curr/avg loss: 0.019311/0.076379, [  938/  938]

 prediction - curr/avg acc: 1.000000/0.942800                    - curr/avg unique acc: 0.904762/0.834914, [   79/   79]

 epoch 5/5 - curr/avg acc: 0.968750/0.960667                - curr/avg loss: 0.051632/0.066998, [  938/  938]

 prediction - curr/avg acc: 0.937500/0.947600                    - curr/avg unique acc: 0.714286/0.843407, [   79/   79]

unsupervised cluster results of random variable 0 is:
     0     1     2    3    4    5    6     7    8    9
0    5  1120    20   28  957    8   10  1005   16  966
1  975    15  1012  982   25  884  948    23  958   43
accs: [0.9525, 0.9485, 0.9487, 0.9516, 0.9431, 0.9463, 0.9478, 0.9468, 0.9543, 0.9476], mean: 0.94872, std: 0.0031144180836875446.
hello world~