joddiy edited a comment on issue #744:
URL: https://github.com/apache/singa/issues/744#issuecomment-646732001


   @chrishkchris I have fixed the average pool and max pool issues above in  
https://github.com/apache/singa/pull/746
   
   However, for the equal, I found this is because there is some error when we 
do the `sub` operator for int32 input:
   
   for example for input:
   
   ```
   A = np.array([[[17, 4, 9, 22, 18], [-9, 9, -1, -1, 4], [1, 14, 7, 1, 4],
                   [3, 14, -2, 3, -8]],
                   [[-25, 6, 8, -7, 22], [-14, 0, -1, 15, 14],
                   [1, 3, -8, -19, -3], [1, 12, 12, -3, -3]],
                   [[-10, -14, -17, 19, -5], [-4, -12, 7, -16, -2],
                   [-8, 3, -5, -11, 0], [4, 0, 3, -6, -3]]],
                       dtype=np.int32)
   B = np.array([[[-6, -3, -8, -17, 1], [-4, -16, 4, -9, 0],
                   [7, 1, 11, -12, 4], [-6, -8, -5, -3, 0]],
                   [[-11, 9, 4, -15, 14], [18, 11, -1, -10, 10],
                   [-4, 12, 2, 9, 3], [7, 0, 17, 1, 4]],
                   [[18, -13, -12, 9, -11], [19, -4, -7, 19, 14],
                   [18, 9, -8, 19, -2], [8, 9, -1, 6, 9]]],
                       dtype=np.int32)
   ```
   The result of `singa.__sub__(a, b)` is:
   
   ```
   [[[         -6          -3          -8         -17          17]
     [         -9          -3          -1          -1           3]
     [         -6          -3          -8         -17           3]
     [         -6          -3          -2         -17          -8]]
   
    [[        -25          -3          -8          -7          21]
     [        -14          -3          -1         -17          13]
     [         -6          -3          -8         -19          -3]
     [         -6          -3          -8          -3          -3]]
   
    [[        -10         -14         -17         -17          -5]
     [         -4         -12          -8         -16          -2]
     [         -8          -3          -5         -11 -2147483647]
     [         -6          -3          -8          -6          -3]]]
   Fs[[[         -6          -3          -8         -17          17]
     [         -9         -16          -1          -1           4]
     [-2147483642          13 -2147483644         -12           0]
     [         -6          -8          -2          -3          -8]]
   
    [[        -25 -2147483645           4          -7           8]
     [        -14 -2147483637          -1         -10           4]
     [         -4 -2147483639          -8         -19          -3]
     [-2147483642          12 -2147483643          -3          -3]]
   
    [[        -10         -14         -17          10          -5]
     [         -4         -12          -7         -16          -2]
     [         -8 -2147483642          -5         -11          -2]
     [-2147483644 -2147483639          -1          -6          -3]]]
   ```
   
   The correct result should be:
   ```
   [[[ 23,   7,  17,  39,  17],
           [ -5,  25,  -5,   8,   4],
           [ -6,  13,  -4,  13,   0],
           [  9,  22,   3,   6,  -8]],
   
          [[-14,  -3,   4,   8,   8],
           [-32, -11,   0,  25,   4],
           [  5,  -9, -10, -28,  -6],
           [ -6,  12,  -5,  -4,  -7]],
   
          [[-28,  -1,  -5,  10,   6],
           [-23,  -8,  14, -35, -16],
           [-26,  -6,   3, -30,   2],
           [ -4,  -9,   4, -12, -12]]]
   ```
   
   @dcslin  can we fix it?


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
[email protected]


Reply via email to