Raise error when inputs and mask don't have same shape in Softmax #19736
Conversation
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #19736 +/- ##
===========================================
- Coverage 78.53% 56.30% -22.24%
===========================================
Files 498 498
Lines 45753 45755 +2
Branches 8455 8456 +1
===========================================
- Hits 35933 25762 -10171
- Misses 8086 18402 +10316
+ Partials 1734 1591 -143
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. |
| @@ -50,6 +50,12 @@ def __init__(self, axis=-1, **kwargs): | |||
|
|
|||
| def call(self, inputs, mask=None): | |||
| if mask is not None: | |||
| if mask.shape != inputs.shape: | |||
| raise ValueError( | |||
| "`mask` and `inputs` must have same shape. " | |||
There was a problem hiding this comment.
They just need to be broadcastable. For instance, if the mask has shape (b, t) and the output has shape (b, t, c) that's fine.
There was a problem hiding this comment.
Actually documentation states that mask should be of same shape. May be we need to add here that same shape or broadcastable shape.
mask: A boolean mask of the same shape as `inputs`. The mask
specifies 1 to keep and 0 to mask. Defaults to `None`.
Also the first dimension of shape i.e batch_size should not be considered in shape. Because if we pass input_shape as (5, 4, 1) and mask as (5,4) we get error as "Incompatible shapes: [5,4,1] vs. [5,4]". Do we need to mention this explicitly in docs?
Also the compute_output_shape method simply returns input_shape. Suppose if I pass
x = np.random.rand(7, 1, 5)
mask = np.ones((5, 1))
l_1 = keras.layers.Softmax(axis=-1,trainable = True,dtype='float32',autocast = True)
print(l_1(x).shape) #(7, 1, 5)
print(l_1(x,mask).shape) #(7, 5,1)
print(l_1.compute_output_shape(x.shape)) #(7, 1, 5)
Noticed output shape is different from compute_output_shape method as the mask is broadcastable here.
If I change mask shape to (1,5) instead of (5,1) then all shapes are same.
Attached gist for reference in case.
There was a problem hiding this comment.
Hi @fchollet Any update on this PR? Please. Thank you!
There was a problem hiding this comment.
Hi @fchollet Any update on this PR? Please. Thank you!
Currently keras.layers.Softmax not having check to verify inputs and mask shape same or not. As per documnetation the mask should be A boolean mask of the same shape as inputs.
But Its generating output even though mask shape is not same as that of inputs.
Hence proposing a check to raise exception when inputs and mask don't have same shape.
Fixes #19722 .