Third-party support matrix

Applies to Linux

2024-10-01

3 min read time

ROCm™ supports a variety of third-party libraries and frameworks. The supported versions have been tested so you can be assured that they function as expected. Non-supported third-party versions might work but haven’t been tested.

Deep learning

ROCm releases support the most recent and two prior releases of PyTorch and TensorFlow.

ROCm	PyTorch	TensorFlow
6.2.x	1.13, 2.0, 2.1, 2.2, 2.3	2.14.1, 2.15.1, 2.16.1
6.1.x	1.13, 2.0, 2.1	2.13.1, 2.14.0, 2.15.0
6.0.x	1.13, 2.0, 2.1	2.12.1, 2.13.1, 2.14.0
5.7.x	1.12.1, 1.13, 2.0, 2.1	2.12, 2.13, 2.14
5.6.x	1.12.1, 1.13, 2.0, 2.1	2.12, 2.13, 2.14
5.5.x	1.11, 1.12.1, 1.13, 2.0	2.10, 2.11, 2.12
5.4.x	1.10.1, 1.11, 1.12.1, 1.13	2.8, 2.9, 2.10, 2.11
5.3.x	1.10.1, 1.11, 1.12.1, 1.13	2.8, 2.9, 2.10
5.2.x	1.10, 1.11, 1.12	2.7, 2.8, 2.9
5.1.3	1.9, 1.10, 1.11	2.7, 2.8, 2.9
5.0.2	1.8, 1.9, 1.10	2.6, 2.7, 2.8

Note

See the Compatibility matrix for an overview of deep learning framework support across ROCm releases.

Communication libraries

ROCm supports OpenUCX, an open-source, production-grade communication framework for data-centric and high performance applications.

ROCm version	UCX version
>= 6.1.1	>= 1.15.0
<= 6.1.0, >= 5.5.0	>= 1.14.1
<= 5.4.x	>= 1.14.0

The Unified Collective Communication (UCC) library also has support for ROCm devices.

ROCm version	UCC version
>= 6.1.0	>= 1.3.0
>= 5.6.0	>= 1.2.0
<= 5.5.x	>= 1.1.0

Note

See the Compatibility matrix for an overview of support for third-party communication libraries across ROCm releases.

Algorithm libraries

ROCm releases provide algorithm libraries with interfaces compatible with contemporary CUDA/NVIDIA HPC SDK alternatives.

• Thrust → rocThrust

• CUB → hipCUB

ROCm	Thrust / CUB	HPC SDK
6.2.x	2.2.0	22.9
6.1.x	2.1.0	22.9
6.0.x	2.0.1	22.9
5.7.x	1.17.2	22.9
5.6.x	1.17.2	22.9
5.5.x	1.17	22.9
5.4.x	1.16	22.9
5.3.x	1.16	22.7
5.2.x	1.15	22.2, 22.3
5.1.3	1.15	22.1
5.0.2	1.14	21.9

For the latest documentation of these libraries, refer to the ROCm API libraries.

Note

See the Compatibility matrix for an overview of algorithm library support across ROCm releases.