DickJC123 opened a new issue #20898:
URL: https://github.com/apache/incubator-mxnet/issues/20898
## Description
I discovered this issue when troubleshooting a unittest failure of
./tests/python/unittest/test_numpy_op.py::test_np_standard_binary_funcs. The
command-line repro is given in the `steps to reproduce` section below. The
test involves adding 2 operands, one with dtype float16, the other of dtype
float64.
The standard numpy (`onp` below) has a somewhat strange behavior when one of
the inputs is broadcast: if the broadcast input is a 0-dim array comprising a
single scalar, then the output will be the dtype of the other input, and the
scalar is cast to that other input's precision. However, if the broadcast
input is a 1-element array of shape (1,), then the output has the dtype of the
most precise input, and the least precise input is cast-up to that precision.
MXNet's numpy does not match this behavior: in both cases the output is
cast-up to the highest precision.
See the sequence below. The lines `a_onp + b_onp` and `a_np + b_np` match,
showing the same behavior when broadcasting a 1-element array. However, the
lines `a2_onp + b_onp` and `a2_np + b_np` do not match, showing a different
behavior when broadcasting a 0-dim array:
```
$ python
Python 3.8.10 (default, Nov 26 2021, 20:14:08)
[GCC 9.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import numpy as onp
>>> a_onp = onp.array([-0.5456035234182584], dtype=onp.float64)
# dim-1 array
>>> b_onp = onp.array([0.54833984, 0], dtype=onp.float16)
>>> a_onp + b_onp
array([ 0.00273632, -0.54560352])
>>> type(a_onp + b_onp)
<class 'numpy.ndarray'>
>>> (a_onp + b_onp).dtype
dtype('float64')
>>> a2_onp = onp.array(-0.5456035234182584, dtype=onp.float64)
# dim-0 array
>>> a2_onp + b_onp
array([ 0.00293, -0.5454 ], dtype=float16)
# numpy 'ground truth' for issue
>>> import mxnet.numpy as np
>>> a_np = np.array([-0.5456035234182584], dtype=np.float64)
# dim-1 array
>>> b_np = np.array([0.54833984, 0], dtype=np.float16)
>>> a_np + b_np
array([ 0.00273632, -0.54560352], dtype=float64)
# mxnet.numpy matches numpy
>>> type(a_np + b_np)
<class 'mxnet.numpy.ndarray'>
>>> (a_np + b_np).dtype
dtype('float64')
>>> a2_np = np.array(-0.5456035234182584, dtype=np.float64)
# dim-0 array
>>> a2_np + b_np
array([ 0.00273632, -0.54560352], dtype=float64)
# mxnet.numpy differs from numpy 'ground truth'
```
### Steps to reproduce
```
MXNET_TEST_SEED=1117974739 pytest --verbose -s
tests/python/unittest/test_numpy_op.py::test_np_standard_binary_funcs[lshape6-rshape6-add-add-True-numeric-\<lambda\>-None--1.0-1.0]
```
### Error Message
Partial output:
```
E AssertionError:
E Items are not equal:
E Error 1.495529 exceeds tolerance rtol=1.000000e-02,
atol=1.000000e-04 (mismatch 16.666667%).
E Location of maximum error: (0, 0), a=0.00273632, b=0.00292969
E ACTUAL: array([[ 0.00273632, 0.04570507, -0.57712818],
E [-0.14863087, -1.35322071, -0.93403126]])
E DESIRED: array([[ 0.00292969, 0.04589844, -0.57714844],
E [-0.1484375 , -1.35351562, -0.93359375]])
```
## Environment
Ubuntu 20.04. Recent master build. Failure of cpu test, so GPU config not
relevant.
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
