Support Matrix v1.13.0

The tables below detail the configurations and versions supported for Gaudi2 and first-gen Gaudi platforms:

SynapseAI

1.13.0

Operating Systems

Ubuntu

Debian

RHEL8

Version

20.04

10.10

8.6

Kernel

5.4.0 and above

5.4.0 and above

4.18.0

Python

  • PyTorch: 3.8

  • TensorFlow: 3.10

  • PyTorch: 3.8

  • TensorFlow: 3.10

  • PyTorch: 3.8

  • TensorFlow: 3.10

Kubernetes

1.25, 1.24

Docker

18.09.0

PyTorch

2.1.0

PyTorch Lightning

or

Lightning

2.1.0

lightning-habana

1.2.0

DeepSpeed

Forked from 0.10.3 of the official DeepSpeed.

TensorFlow

2.13.1

Habana Horovod

Forked from 0.27.0 of the official Horovod

Open MPI

4.1.5

Libfabric

  • General support: 1.16.1 and above

  • Gaudi Direct with Verbs: 1.20.0

Optimum Habana

1.9.0

Transformers

4.34

Components

Version

SynapseAI

Build Number

1.13.0-463

SVN

SVN Version

sec-6

HL-225

SPI Firmware

1.13.0-fw-46.2.0

FIT

1.13.0-fw-46.2.0

CPLD

0x0F

eROM

1.12.1-fw-46.0.5

HLBA-225

PCIe Retimer Firmware

2.2

SerDes Retimer Firmware

0xD00A

CPLD

  • A01 (HLS users)

  • A02SA (non-HLS users)

HLS-Gaudi2

JC BMC

4.0

HIB BMC

3.03

PCIe Switch

002104

SynapseAI

1.13.0

Gaudi Firmware

1.2.3

Gaudi SPI Firmware

1.1.0

Operating Systems

Ubuntu

Ubuntu

Amazon Linux 2

RHEL8

Version

20.04

22.04

8.6

Kernel

5.4.0 and above

5.15 and above

5.4.0 and above

4.18.0

Python

  • PyTorch: 3.8

  • TensorFlow: 3.10

  • PyTorch: 3.10

  • TensorFlow: 3.10

  • PyTorch: 3.8

  • TensorFlow: 3.10

  • PyTorch: 3.8

  • TensorFlow: 3.10

Kubernetes

1.25, 1.24, 1.23

OpenShift

4.12

Docker

18.09.0

PyTorch

2.1.0

PyTorch Lightning

or

Lightning

2.1.0

lightning-habana

1.2.0

DeepSpeed

Forked from 0.10.3 of the official DeepSpeed.

TensorFlow

2.13.1

Habana Horovod

Forked from 0.27.0 of the official Horovod

Open MPI

4.1.5

Libfabric

1.16.1 and above

Optimum Habana

1.9.0

Transformers

4.34