Multi-XPU support?

[aza]

Hello, 

Is there any guide online to guide how to utilize Multiple XPU of model training/fine-tuneing?

I tried to use accelerate, Pytorch dataparallel but it not worked.

[Wan_Intel]

Hi Aza,

Thanks for reaching out to us.

You may refer to PyTorch 2.4 on Intel® GPUs to learn the basics on how to use PyTorch 2.4 with Intel GPUs. PyTorch 2.4 on Intel® GPUs is designed for developers, data scientists, and AI enthusiasts who want to leverage the capabilities of PyTorch on Intel GPUs (XPUs).

Regards,

Wan

[aza]

That guide is only for single XPU usage

I want one for multiple XPU 

as you can see in the environment there are 8 XPU available. 

@idc-training-gpu-compute-22:~$ xpu-smi discovery
+-----------+--------------------------------------------------------------------------------------+
| Device ID | Device Information |
+-----------+--------------------------------------------------------------------------------------+
| 0 | Device Name: Intel(R) Data Center GPU Max 1100 |
| | Vendor Name: Intel(R) Corporation |
| | SOC UUID: 00000000-0000-000f-0000-002f0bda8086 |
| | PCI BDF Address: 0000:0f:00.0 |
| | DRM Device: /dev/dri/card1 |
| | Function Type: physical |
+-----------+--------------------------------------------------------------------------------------+
| 1 | Device Name: Intel(R) Data Center GPU Max 1100 |
| | Vendor Name: Intel(R) Corporation |
| | SOC UUID: 00000000-0000-0016-0000-002f0bda8086 |
| | PCI BDF Address: 0000:16:00.0 |
| | DRM Device: /dev/dri/card2 |
| | Function Type: physical |
+-----------+--------------------------------------------------------------------------------------+
| 2 | Device Name: Intel(R) Data Center GPU Max 1100 |
| | Vendor Name: Intel(R) Corporation |
| | SOC UUID: 00000000-0000-001a-0000-002f0bda8086 |
| | PCI BDF Address: 0000:1a:00.0 |
| | DRM Device: /dev/dri/card3 |
| | Function Type: physical |
+-----------+--------------------------------------------------------------------------------------+
| 3 | Device Name: Intel(R) Data Center GPU Max 1100 |
| | Vendor Name: Intel(R) Corporation |
| | SOC UUID: 00000000-0000-001e-0000-002f0bda8086 |
| | PCI BDF Address: 0000:1e:00.0 |
| | DRM Device: /dev/dri/card4 |
| | Function Type: physical |
+-----------+--------------------------------------------------------------------------------------+
| 4 | Device Name: Intel(R) Data Center GPU Max 1100 |
| | Vendor Name: Intel(R) Corporation |
| | SOC UUID: 00000000-0000-008a-0000-002f0bda8086 |
| | PCI BDF Address: 0000:8a:00.0 |
| | DRM Device: /dev/dri/card5 |
| | Function Type: physical |
+-----------+--------------------------------------------------------------------------------------+
| 5 | Device Name: Intel(R) Data Center GPU Max 1100 |
| | Vendor Name: Intel(R) Corporation |
| | SOC UUID: 00000000-0000-008e-0000-002f0bda8086 |
| | PCI BDF Address: 0000:8e:00.0 |
| | DRM Device: /dev/dri/card6 |
| | Function Type: physical |
+-----------+--------------------------------------------------------------------------------------+
| 6 | Device Name: Intel(R) Data Center GPU Max 1100 |
| | Vendor Name: Intel(R) Corporation |
| | SOC UUID: 00000000-0000-00c0-0000-002f0bda8086 |
| | PCI BDF Address: 0000:c0:00.0 |
| | DRM Device: /dev/dri/card7 |
| | Function Type: physical |
+-----------+--------------------------------------------------------------------------------------+
| 7 | Device Name: Intel(R) Data Center GPU Max 1100 |
| | Vendor Name: Intel(R) Corporation |
| | SOC UUID: 00000000-0000-00c4-0000-002f0bda8086 |
| | PCI BDF Address: 0000:c4:00.0 |
| | DRM Device: /dev/dri/card8 |
| | Function Type: physical |
+-----------+--------------------------------------------------------------------------------------+

[Wan_Intel]

Hi Aza,

Thanks for the information.

Let me check with relevant team and I'll update you as soon as possible.

Regards,

Wan

[Wan_Intel · Accepted solution]

Hi Aza,

Thanks for your patience. We've received feedback from the relevant team.

For your information, we are sorry to tell you that utilization of all XPUs in the Jupyter Lab is not available because users are only given access to one shared GPU.

Sorry for the inconvenience and thank you for your support.

Regards,

Wan

[aza]

Between is there any guide to use multiple xpu for knowledge purposes 

[Wan_Intel]

Hi Aza,

You may use DataParallel with Intel® Extension for PyTorch.

For more information, please refer to Multi-GPU AI Training (Data-Parallel) with Intel® Extension for PyTorch* | Intel Software.

Regards,

Wan

[Wan_Intel]

Hi Aza,

Thanks for your question.

If you need additional information from Intel, please submit a new question as this thread will no longer be monitored.

Regards,

Wan