-
Threw a pretty [easy PR](https://github.com/pixray/pixray/pull/52) up which implements the feature described by post title:
I know it's not perfect as I think maybe other non-pixray libs may still …
-
from https://github.com/nagadomi/nunif/discussions/28#discussioncomment-7247459
>i have 2 gpus.
>Haw can i use all gpu in iw3?
-
I did look up how the h264 code (freerdp/src/FreeRDP-2.5.0/libfreerdp/codec/h264_ffmpeg.c) uses Vaapi currently and it will always select the DRI device 128:
```
#ifdef WITH_VAAPI
#define VAAPI_DEV…
-
I hit upon an error in HuggingFace for which there are strangely zero google search results
"ValueError: Calculated loss must be on the original device" I can see this error source code in huggingf…
-
When running on HPG, the print output that gives us "Training batch is on device _" is only reading "device 0". Is this missing computations on the other GPU (i.e. "device 1"), which the program state…
-
### Feature request
Pass in `torch_dtype` in model_kwargs, as supported by sentence_transformers when specifying dtype in the infinity_emb v2 cli when InferenceEngine type is torch.
This would all…
-
1. multi nodes and multi gpus examples
2. system information log(GPU/CPU memory,utilization, temperature, Disk io/utilization and so on
3. different devices log synchronize?
4. support yaml(hydr…
-
# Update
It seems the have something to do with `--machine.num-devices 8`. Without that argument the training works as expected, at least for `nerfacto`. I will test out `splatfacto` later since the …
-
### System Info
```Shell
- `Accelerate` version: 0.33.0
- Platform: Linux-5.15.133+-x86_64-with-glibc2.35
- `accelerate` bash location: /opt/conda/bin/accelerate
- Python version: 3.10.14
- Nu…
-
### Code of Conduct & Contributing Guidelines
- [X] I agree to follow the code of conduct and the contributing guidelines.
### Have you checked that no other similar feature request(s) already e…