| 6 Mar 2024 |
|  Shelby Lynn joined the room. | 14:50:47 |
| 26 Mar 2024 |
|  Martin Charles joined the room. | 08:02:33 |
| 7 Apr 2024 |
|  mehmetoguzderin changed their profile picture. | 10:12:05 |
|  raggi joined the room. | 19:49:29 |
| raggi left the room. | 19:50:18 |
| 10 Apr 2024 |
| Shelby Lynn changed their display name from shelby_nya to Shelby Lynn. | 17:41:28 |
| 11 Apr 2024 |
| mehmetoguzderin changed their profile picture. | 11:09:55 |
| 25 Apr 2024 |
|  @repi:matrix.org left the room. | 11:15:49 |
| 6 May 2024 |
|  feynon joined the room. | 20:08:50 |
| 11 May 2024 |
|  @ashley_:matrix.org left the room. | 05:22:09 |
| 25 May 2024 |
|  eschryn changed their profile picture. | 10:08:24 |
| 28 May 2024 |
|  @unixawesome:tchncs.de joined the room. | 18:02:17 |
| @unixawesome:tchncs.de left the room. | 18:03:09 |
| 5 Jun 2024 |
|  Jelly Terra λ joined the room. | 02:30:18 |
| 8 Jun 2024 |
|  yangyin1996520 joined the room. | 08:41:02 |
| 20 Jun 2024 |
|  voxel changed their profile picture. | 17:55:13 |
| 14 Jul 2024 |
|  joshprk joined the room. | 07:58:54 |
| 16 Jul 2024 |
|  Paddi ⚡️ changed their profile picture. | 23:49:46 |
| 22 Jul 2024 |
|  @gschian0:matrix.org left the room. | 14:22:58 |
| 27 Jul 2024 |
|  @katiew:matrix.org joined the room. | 08:13:30 |
| 30 Jul 2024 |
| @katiew:matrix.org left the room. | 03:52:07 |
| 1 Aug 2024 |
|  xorio42 joined the room. | 15:41:32 |
| 10 Aug 2024 |
|  ErichDonGubler changed their profile picture. | 17:43:54 |
| 13 Aug 2024 |
|  @theemeraldbee:matrix.org joined the room. | 02:54:54 |
| @theemeraldbee:matrix.org left the room. | 02:55:15 |
| 20 Aug 2024 |
| xorio42 | Is there any OSS Nvidia drivers written in Rust? Or some C bindings but the driver to be OSS.
I'm interested in creating a vGPU a virtual GPU where you can run your ML apps like you would on a real GPU. I want then to send the GPU instructions over the wire and execute them in the cloud on real GPUs.
I've seen wgpu as an abstraction layer and then on remote I can use ash to execute them on real GPU.
But how can I create the Virtual interface (I imagined it needs to be (like) a driver) so the local app communicates with it. Can I do it just with wgpu somehow exposing the interface? I would need also support for CUDA.
| 16:05:34 |
| xorio42 | * Do you know of any OSS Nvidia drivers written in Rust? Or some C bindings but the driver to be OSS.
I'm interested in creating a vGPU a virtual GPU where you can run your ML apps like you would on a real GPU. I want then to send the GPU instructions over the wire and execute them in the cloud on real GPUs.
I've seen wgpu as an abstraction layer and then on remote I can use ash to execute them on real GPU.
But how can I create the Virtual interface (I imagined it needs to be (like) a driver) so the local app communicates with it. Can I do it just with wgpu somehow exposing the interface? I would need also support for CUDA.
I've read about nova but not sure in what state it is, also hear Redhat writing a new driver for Vulkan in Rust, maybe it's nova?
| 16:07:16 |
| xorio42 | * Do you know of any OSS Nvidia drivers written in Rust? Or some C bindings but the driver to be OSS.
I'm interested in creating a vGPU a virtual GPU where you can run your ML apps like you would on a real GPU. I want then to send the GPU instructions over the wire and execute them in the cloud on real GPUs.
I've seen wgpu as an abstraction layer and then on remote I can use ash to execute them on real GPU.
But how can I create the Virtual interface (I imagined it needs to be (like) a driver) so the local app communicates with it. Can I do it just with wgpu somehow exposing the interface? I would need also support for CUDA.
I've read about nova but not sure in what state it is, also hear Redhat writing a new driver for Vulkan in Rust, maybe it's nova?
| 16:08:06 |
| 21 Aug 2024 |
 tanriol | In reply to @xorio42:matrix.org
Do you know of any OSS Nvidia drivers written in Rust? Or some C bindings but the driver to be OSS.
I'm interested in creating a vGPU a virtual GPU where you can run your ML apps like you would on a real GPU. I want then to send the GPU instructions over the wire and execute them in the cloud on real GPUs.
I've seen wgpu as an abstraction layer and then on remote I can use ash to execute them on real GPU.
But how can I create the Virtual interface (I imagined it needs to be (like) a driver) so the local app communicates with it. Can I do it just with wgpu somehow exposing the interface? I would need also support for CUDA.
I've read about nova but not sure in what state it is, also hear Redhat writing a new driver for Vulkan in Rust, maybe it's nova?
What do you mean by "virtual GPU"? At what level do you want it to exist? | 13:14:05 |
| xorio42 | I want the python or some apps that run cuda instructions to send them to the vgpu | 13:15:14 |
