UniformIntDistribution
is exclusive in its upper bound
#56675
Labels
comp:xla
XLA
stat:awaiting tensorflower
Status - Awaiting response from tensorflower
TF 2.8
type:support
Support issues
Click to expand!
Issue Type
Bug
Source
source (I actually use this binary but that isn't compiled by Google)
Tensorflow Version
2.8
Custom Code
Yes
OS Platform and Distribution
Ubuntu 20.04
Mobile device
n/a
Python version
n/a
Bazel version
2.4.1
GCC/Compiler version
Unknown, between 7.5 and 9.3
CUDA/cuDNN version
CUDA Version: 11.6
GPU model and memory
NVIDIA GeForce GTX 1070
Current Behaviour?
UniformIntDistribution
appears to be inclusive in its lower bound, but exclusive in its upper bound. I think this is wrong for a number of reasons:UniformFloatingPointDistribution
appears to be inclusive in its upper bound. I came to this conclusion as sampling that function with equal bounds returns the common bound, rather than, say NaN. The two functions thus have incongruous behaviour.*Example samples for equal bounds of 0 and 0
Standalone code to reproduce the issue
Relevant log output
pred[] false
The text was updated successfully, but these errors were encountered: