-
## 🚀 Feature
The method converts a PIL Image or numpy.ndarray to a tensor with scaling pixel values to the range [0,1]. The new parameter will provide an option to the user whether the pixel values w…
-
### 🐛 Describe the bug
1. torch.random.fork_rng(devices=rng_device) does not support XPU backend
TestRandomTensorCreationXPU.test_randperm_xpu
![image](https://github.com/intel/torch-xpu-ops/asse…
-
# Feature request
We've made some strides improving the quadrature handling in GPflow. It would be amazing if we could also make it possible to use alternatives to the default Gauss-Hermite quadratur…
-
See a [relevant discord discussion here](https://discord.com/channels/211786369951989762/573613501164159016/783413636852613221).
This could be usefull for blockart stuff or similar.
I will probably …
-
The following does not work right now:
CTensor psi = CTensor::random(5, 4, 3);
CTensor slice = psi(0, range(), 2);
because this calls the function that takes only ranges, and mixing of ranges and sl…
-
Is there anyway large-scale inference can be sped up?
I tried the removing the conversion in #4 , which did speed up by a factor of ~3.
But it is still substantially slower, than say using text-gene…
wj210 updated
9 months ago
-
The Lua version of Torch implements a [Locally Connected Layer](https://github.com/torch/nn/blob/master/doc/convolution.md#spatialconvolutio) (Convolution without shared weights). It would be good to …
-
Hi, In **neural-style**, inappropriate dependency versioning constraints can cause risks.
Below are the dependencies and version constraints that the project is using
```
numpy
Pillow
scipy
…
-
Hi,
Thank you and the matching results are very good. :)
Earlier i used **Sift** for feature extraction and **cv.detail_BestOf2NearestMatcher()** for matching. Then i use **cv.detail_HomographyB…
-
## 🐛 Bug
Identified by @albanD, from offline discussion with @ezyang: "You cannot assume that there isn't an error already set when THPVariable_dealloc is called.
If you need to call some function…