-
It looks like cupy supports record arrays, but fails to accept a list dtype that was not properly packed as a dtype object. My guess is that this could
```python
In [1]: import numpy as np
I…
-
One way for huge optimizations would be to support natively Cupy or Cuspatial. I tried replacing the numpy import on the sys level with cupy however a few small inconsistencies arise. Although cupy re…
-
Distributed training support has been a long-term goal for our NLP library [spaCy](https://spacy.io). I think this could be relevant to Uber as well, as I think it's in use in some places.
spaCy's …
-
# Description
following the discussion in #231 it should be quite easy to get a CuPy backend into pyhf
https://cupy.chainer.org/
it's advertised as a drop-in replacement for numpy. This will …
-
when i try to train on voc, i come across this error:
from collections import namedtuple
faster_rcnn = FasterRCNNVGG16()
File "/home/yangshao/workspace/faster_rcnn/model/faster_rcnn_vgg16.py"…
-
### Description
The dtype of returned indices in `cupy.unique(..., return_index=True, return_inverse=True)` appear to be always `int64`, even if the input array is `int32` and its size is within the …
-
### Description
cupy.argsort uses int64 dtype for its results. When arrays of length less than 2 billion entries are used (which is nearly all of the use cases, see later), it would more efficient …
-
### Description
help
### Additional Information
_No response_
-
Given the successful speedups using CUDA for parts of the Sync3N algorithm, we should implement a similar GPU implementation for building the CL matrix.
For unit test sized problems our current imp…
-
Got some errors during install, but the installation in the end said it installed successfully, so not sure what that really means?
A restart didn't so any issues in the restart process, just the i…