-
The `transformers` backend does not appear to use the GPU.
I could not get any of the settings such as `cuda: true` or `device: cuda` to work. In the code, I could not see any of these options being …
kno10 updated
3 months ago
-
[Windows doesn't support cudaMemPrefetchAsync()](https://github.com/TimDettmers/bitsandbytes/issues/453) which is used in the call to `prefetch`. Would be a need to investigate alternatives for Window…
-
When I execute main.exe which I have built with github actions and cx_Freeze i am getting the following error:
`Could not locate cudnn_ops_infer64_8.dll. Please make sure it is in your library pat…
-
[/usr/local/lib/python3.10/dist-packages/triton/compiler/backends/cuda.py](https://localhost:8080/#) in make_llir(src, metadata, options, capability)
171 add_external_libs(src, names,…
-
Hey,
when I try to use 3 GPUs but only 2 are available the library behaves as expected and the third process crashes.
```
python gpu-acquisitor.py --backend pytorch --id 1 --nb-gpus 1 & python3 g…
-
### 🐛 Describe the bug
As of `torch==2.4`, we're unable to register an out-of-tree backend using the `PrivateUse1` mechanism described in [this documentation](https://pytorch.org/tutorials/advanced/e…
-
## 🐛 Bug
User of dgl may not need graphbolt, so import dgl should not automatically import graphbolt. While in the latest dgl, import dgl will automatically import grapbolt
## To Reproduce
St…
-
I found and isolated an error in `ggml_get_rows` with the cuda backend where for tensors of where the first dimension is greater than `65535`, the program fails with the following output:
````
loa…
-
### Issue Type
Bug
### Source
binary
### Keras Version
3.5.0
### Custom Code
No
### OS Platform and Distribution
Linux Ubuntu v22.04
### Python version
3.10.12
### …
-
Tried to run VirtualGL using EGL backend with the official [CUDA OpenGL interop sample code](https://github.com/NVIDIA/cuda-samples/tree/master/Samples/0_Introduction/simpleCUDA2GL), but it keeps fail…