Merlin PyTorch Support Matrix

Merlin PyTorch Support Matrix#

This container enables you to train and deploy NVTabular workflows and PyTorch models to the Triton Inference Server for production.

23.xx Container Images#

Container Release

Release 23.02

DGX

DGX System

  • DGX-1

  • DGX-2

  • DGX A100

  • DGX Station

Operating System

Ubuntu 20.04.5 LTS

NVIDIA Certified Systems

NVIDIA Driver

NVIDIA Driver version 465.19.01

or later is required. However,

if you’re running on Data Center

GPUs (formerly Tesla) such as T4,

you can use any of the following

NVIDIA Driver versions:

  • 418.40 (or later R418)

  • 440.33 (or later R440)

  • 450.51 (or later R450)

  • 460.27 (or later R460)

Note: The CUDA Driver

Compatibility Package does not

support all drivers.

GPU Model

Base Container Image

Container Operating System

Ubuntu 20.04.5 LTS

Base Container

Triton version 22.12

CUDA

11.8.0.065

RMM

22.08.00a+62.gf6bf047.dirty

cuDF

22.08.00a+304.g6ca81bbc78.dirty

cuDNN

8.7.0.84

Merlin Core

23.2.0

Merlin Dataloader

23.2.0

Merlin Models

23.2.0

Merlin Systems

23.2.0

Distributed Embeddings

Not applicable

NVTabular

23.2.0

Transformers4Rec

23.2.0

HugeCTR

Not applicable

PyTorch

1.13.1

Triton Inference Server

2.29.0

Compressed Size

6.7 GB

22.xx Container Images#

Container Release

Release 22.12

Release 22.11

Release 22.10

Release 22.09

Release 22.07

Release 22.06

DGX

DGX System

  • DGX-1

  • DGX-2

  • DGX A100

  • DGX Station

  • DGX-1

  • DGX-2

  • DGX A100

  • DGX Station

  • DGX-1

  • DGX-2

  • DGX A100

  • DGX Station

  • DGX-1

  • DGX-2

  • DGX A100

  • DGX Station

  • DGX-1

  • DGX-2

  • DGX A100

  • DGX Station

  • DGX-1

  • DGX-2

  • DGX A100

  • DGX Station

Operating System

Ubuntu 20.04.5 LTS

Ubuntu 20.04.4 LTS

Ubuntu 20.04.4 LTS

Ubuntu 20.04.4 LTS

Ubuntu 20.04.4 LTS

Ubuntu 20.04.4 LTS

NVIDIA Certified Systems

NVIDIA Driver

NVIDIA Driver version 465.19.01

or later is required. However,

if you’re running on Data Center

GPUs (formerly Tesla) such as T4,

you can use any of the following

NVIDIA Driver versions:

  • 418.40 (or later R418)

  • 440.33 (or later R440)

  • 450.51 (or later R450)

  • 460.27 (or later R460)

Note: The CUDA Driver

Compatibility Package does not

support all drivers.

NVIDIA Driver version 465.19.01

or later is required. However,

if you’re running on Data Center

GPUs (formerly Tesla) such as T4,

you can use any of the following

NVIDIA Driver versions:

  • 418.40 (or later R418)

  • 440.33 (or later R440)

  • 450.51 (or later R450)

  • 460.27 (or later R460)

Note: The CUDA Driver

Compatibility Package does not

support all drivers.

NVIDIA Driver version 465.19.01

or later is required. However,

if you’re running on Data Center

GPUs (formerly Tesla) such as T4,

you can use any of the following

NVIDIA Driver versions:

  • 418.40 (or later R418)

  • 440.33 (or later R440)

  • 450.51 (or later R450)

  • 460.27 (or later R460)

Note: The CUDA Driver

Compatibility Package does not

support all drivers.

NVIDIA Driver version 465.19.01

or later is required. However,

if you’re running on Data Center

GPUs (formerly Tesla) such as T4,

you can use any of the following

NVIDIA Driver versions:

  • 418.40 (or later R418)

  • 440.33 (or later R440)

  • 450.51 (or later R450)

  • 460.27 (or later R460)

Note: The CUDA Driver

Compatibility Package does not

support all drivers.

NVIDIA Driver version 465.19.01

or later is required. However,

if you’re running on Data Center

GPUs (formerly Tesla) such as T4,

you can use any of the following

NVIDIA Driver versions:

  • 418.40 (or later R418)

  • 440.33 (or later R440)

  • 450.51 (or later R450)

  • 460.27 (or later R460)

Note: The CUDA Driver

Compatibility Package does not

support all drivers.

NVIDIA Driver version 465.19.01

or later is required. However,

if you’re running on Data Center

GPUs (formerly Tesla) such as T4,

you can use any of the following

NVIDIA Driver versions:

  • 418.40 (or later R418)

  • 440.33 (or later R440)

  • 450.51 (or later R450)

  • 460.27 (or later R460)

Note: The CUDA Driver

Compatibility Package does not

support all drivers.

GPU Model

Base Container Image

Container Operating System

Ubuntu 20.04.5 LTS

Ubuntu 20.04.4 LTS

Ubuntu 20.04.4 LTS

Ubuntu 20.04.4 LTS

Ubuntu 20.04.4 LTS

Ubuntu 20.04.4 LTS

Base Container

Triton version 22.10

Triton version 22.08

Triton version 22.08

Triton version 22.08

Triton version 22.05

Triton version 22.05

CUDA

11.8.0.065

11.7.1.017

11.7.1.017

11.7.1.017

11.7.1.014

11.7.0.022

RMM

22.08.00a+62.gf6bf047.dirty

22.06.00a+76.g185c18e6

22.06.00a+76.g185c18e6

22.06.00a+76.g185c18e6

21.12.0

21.12.0

cuDF

22.08.00a+304.g6ca81bbc78.dirty

22.06.00a+319.g97422602b8

22.06.00a+319.g97422602b8

22.06.00a+319.g97422602b8

22.4.0

22.4.0

cuDNN

8.7.0.80

8.5.0.96

8.5.0.96

8.5.0.96

8.4.1.50+cuda11.6

8.4.0.27+cuda11.6

Merlin Core

0.10.0

0.9.0

0.8.0

0.7.0

0.5.0+1.g1354dcf

0.4.0

Merlin Dataloader

0.0.4

0.0.3

Not applicable

Not applicable

Not applicable

Not applicable

Merlin Models

0.11.0

0.10.0

0.9.0

0.8.0

0.6.0+4.g046077b8

0.5.0+7.g886cf6de

Merlin Systems

0.9.0

0.8.0

0.7.0

0.6.0

0.4.0+1.g07bf4ab

0.3.0+1.g712b04d

Distributed Embeddings

Not applicable

Not applicable

Not applicable

Not applicable

Not applicable

Not applicable

NVTabular

1.8.0

1.7.0

1.6.0

1.5.0

1.3.3+1.g32ed5fa12

1.2.2

Transformers4Rec

0.1.16

0.1.15

0.1.14

0.1.13

0.1.11

0.1.10

HugeCTR

Not applicable

Not applicable

Not applicable

Not applicable

Not applicable

Not applicable

PyTorch

1.13.1

1.13.0

1.12.1

1.12.1

1.12.0

1.11.0

Triton Inference Server

2.28.0

2.25.0

2.25.0

2.25.0

2.23.0

2.22.0

Compressed Size

11.13 GB

9.76 GB

10.96 GB

10.35 GB

7.18 GB

6.68 GB