Merlin PyTorch Inference Support Matrix

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

22.xx Container Images

Container Release

Release 22.04

Release 22.03

DGX

DGX System

  • DGX-1

  • DGX-2

  • DGX A100

  • DGX Station

  • DGX-1

  • DGX-2

  • DGX A100

  • DGX Station

Operating System

Ubuntu 20.04.4 LTS

Ubuntu 20.04.3 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.

GPU Model

Base Container Image

Container Operating System

Ubuntu 20.04.4 LTS

Ubuntu 20.04.3 LTS

Base Container

Triton version 22.03

Triton version 22.02

CUDA

11.6.1.005

11.6.0.021

RMM

21.12.0

21.12.00

cuDF

22.2.0

21.12.02

cuDNN

8.3.3.40+cuda11.5

8.3.2.44+cuda11.5

Merlin Core

0.2.0

v0.1.1+3.gee1d59d

Merlin Models

0.3.0

Not applicable

Merlin Systems

0.1.0

Not applicable

NVTabular

1.0.0

0.11.0

Transformers4Rec

0.1.7

0.1.6

HugeCTR

Not applicable

Not applicable

SM

Not applicable

Not applicable

PyTorch

Not applicable

Not applicable

Triton Inference Server

2.20.0

2.19.0

Size

9.82 GB

9.72 GB