-
# Issue Description
Intel-xpu-backend-for-triton `>` operator does not support `i1` but triton for CUDA can.
# Code Example
tmp83 = tmp77 != tmp77
tmp84 = tmp80 != tmp80
tmp85 = tmp…
-
There is large variance in the softmax with NN
-
### Describe the bug
The default latent preview method of ComfyUI is causing crashes when running. Doing a git bisect, https://github.com/comfyanonymous/ComfyUI/commit/4ae1515f14d338a347fdf8f647e06…
-
Issue found running some examples and tests from Huggingface transformers on the system with only XPU device and pytorch built w/o CUDA. As of:
* pytorch at https://github.com/pytorch/pytorch/commit/…
-
### 🐛 Describe the bug
PYTORCH_ENABLE_XPU_FALLBACK=1 PYTORCH_TEST_WITH_SLOW=1 pytest -v test_ops_xpu.py -k 'test_backward_diagonal_xpu_float32'
![image](https://github.com/intel/torch-xpu-ops/asse…
-
### Issue Description
Hunyuan model generates garbled images on Arc A770.
For example, for this sample prompt: An astronaut riding a horse
Size: 1024x1024
It generated the attached image.
![00296…
-
Here are two calls to `is_integer_type`:
1. https://github.com/intel/intel-xpu-backend-for-triton/blob/llvm-target/python/triton/language/standard.py#L196
2. https://github.com/intel/intel-xpu-backe…
-
### 🐛 Describe the bug
Please refer to https://github.com/intel/torch-xpu-ops/actions/runs/8995661712/job/24710982088, there are 6 models got failed.
Error info
Run failed with return code: -11
O…
-
### Describe the issue
If I tried the following lines of code:
```
import torch
embedding = torch.nn.Embedding(10, 3)
embedding.to('xpu')
```
And I got the following error:
```
PyTo…
-
### Describe the issue
When trying to upload a simple model to 'xpu' device, the script is taking a lot of time, and also doesn't finish.
Firstly tried with a resnet50, but after that with the simpl…