github/ROCm
1
1
Fork 0
mirror of https://github.com/ROCm/ROCm.git synced 2026-01-07 22:03:58 -05:00
Code Issues Packages Projects Releases Wiki Activity

Main Docs: references of accelerator removal and change to GPU (#5495)

Browse Source 
* Docs: references of accelerator removal and change to GPU

Co-authored-by: Leo Paoletti <164940351+lpaoletti@users.noreply.github.com>
Co-authored-by: Pratik Basyal <pratik.basyal@amd.com>
This commit is contained in:
anisha-amd
2025-10-16 11:22:10 -04:00
committed by GitHub
parent 5cb6bfe151
commit a98236a4e3
95 changed files with 414 additions and 422 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
CHANGELOG.md
View File

@@ -4116,7 +4116,7 @@ memory partition modes upon an invalid argument return from memory partition mod
- JSON output plugin for `rocprofv2`. The JSON file matches Google Trace Format making it easy to load on Perfetto, Chrome tracing, or Speedscope. For Speedscope, use `--disable-json-data-flows` option as speedscope doesn't work with data flows.
- `--no-serialization` flag to disable kernel serialization when `rocprofv2` is in counter collection mode. This allows `rocprofv2` to avoid deadlock when profiling certain programs in counter collection mode.
- `FP64_ACTIVE` and `ENGINE_ACTIVE` metrics to AMD Instinct MI300 accelerator
- `FP64_ACTIVE` and `ENGINE_ACTIVE` metrics to AMD Instinct MI300 GPU
- New HIP APIs with struct defined inside union.
- Early checks to confirm the eligibility of ELF file in ATT plugin
- Support for kernel name filtering in `rocprofv2`
@@ -4140,18 +4140,18 @@ memory partition modes upon an invalid argument return from memory partition mod
#### Resolved issues
- Bandwidth measurement in AMD Instinct MI300 accelerator
- Bandwidth measurement in AMD Instinct MI300 GPU
- Perfetto plugin issue of `roctx` trace not getting displayed
- `--help` for counter collection
- Signal management issues in `queue.cpp`
- Perfetto tracks for multi-GPU
- Perfetto plugin usage with `rocsys`
- Incorrect number of columns in the output CSV files for counter collection and kernel tracing
- The ROCProfiler hang issue when running kernel trace, thread trace, or counter collection on Iree benchmark for AMD Instinct MI300 accelerator
- The ROCProfiler hang issue when running kernel trace, thread trace, or counter collection on Iree benchmark for AMD Instinct MI300 GPU
- Build errors thrown during parsing of unions
- The system hang caused while running `--kernel-trace` with Perfetto for certain applications
- Missing profiler records issue caused while running `--trace-period`
- The hang issue of `ProfilerAPITest` of `runFeatureTests` on AMD Instinct MI300 accelerator
- The hang issue of `ProfilerAPITest` of `runFeatureTests` on AMD Instinct MI300 GPU
- Segmentation fault on Navi32
@@ -5548,7 +5548,7 @@ See [issue #3499](https://github.com/ROCm/ROCm/issues/3499) on GitHub.
intermediary script to call the application with the necessary arguments, then call the script with Omniperf. This
issue is fixed in a future release of Omniperf. See [#347](https://github.com/ROCm/rocprofiler-compute/issues/347).
- Omniperf might not work with AMD Instinct MI300 accelerators out of the box, resulting in the following error:
- Omniperf might not work with AMD Instinct MI300 GPUs out of the box, resulting in the following error:
"*ERROR gfx942 is not enabled rocprofv1. Available profilers include: ['rocprofv2']*". As a workaround, add the
environment variable `export ROCPROF=rocprofv2`.
@@ -5664,7 +5664,7 @@ See [issue #3498](https://github.com/ROCm/ROCm/issues/3498) on GitHub.
#### Optimized
* Improved performance of Level 1 `dot_batched` and `dot_strided_batched` for all precisions. Performance enhanced by 6 times for bigger problem sizes, as measured on an Instinct MI210 accelerator.
* Improved performance of Level 1 `dot_batched` and `dot_strided_batched` for all precisions. Performance enhanced by 6 times for bigger problem sizes, as measured on an Instinct MI210 GPU.
#### Removed

2
docs/compatibility/compatibility-matrix.rst
View File

@@ -10,7 +10,7 @@ Use this matrix to view the ROCm compatibility and system requirements across su
You can also refer to the :ref:`past versions of ROCm compatibility matrix<past-rocm-compatibility-matrix>`.
Accelerators and GPUs listed in the following table support compute workloads (no display
GPUs listed in the following table support compute workloads (no display
information or graphics). If you’re using ROCm with AMD Radeon GPUs or Ryzen APUs for graphics
workloads, see the :docs:`Use ROCm on Radeon and Ryzen <radeon:index.html>` to verify
compatibility and system requirements.

2
docs/compatibility/ml-compatibility/jax-compatibility.rst
View File

@@ -79,7 +79,7 @@ Use cases and recommendations
* The `MI300X workload optimization guide <https://rocm.docs.amd.com/en/latest/how-to/tuning-guides/mi300x/workload.html>`_
provides detailed guidance on optimizing workloads for the AMD Instinct MI300X
accelerator using ROCm. The page is aimed at helping users achieve optimal
GPU using ROCm. The page is aimed at helping users achieve optimal
performance for deep learning and other high-performance computing tasks on
the MI300X GPU.

6
docs/compatibility/ml-compatibility/pytorch-compatibility.rst
View File

@@ -73,9 +73,9 @@ Use cases and recommendations
* The :doc:`Instinct MI300X workload optimization guide </how-to/rocm-for-ai/inference-optimization/workload>`
provides detailed guidance on optimizing workloads for the AMD Instinct MI300X
accelerator using ROCm. This guide helps users achieve optimal performance for
GPU using ROCm. This guide helps users achieve optimal performance for
deep learning and other high-performance computing tasks on the MI300X
accelerator.
GPU.
* The :doc:`Inception with PyTorch documentation </conceptual/ai-pytorch-inception>`
describes how PyTorch integrates with ROCm for AI workloads It outlines the
@@ -417,7 +417,7 @@ Key features and enhancements for PyTorch 2.7 with ROCm 7.0
- Expanded GPU architecture support: Provides optimized support for newer GPU
architectures, including gfx1200 and gfx1201 with preferred hipBLASLt backend
selection, along with improvements for gfx950 and gfx1100 series GPUs.
selection, along with improvements for gfx950 and gfx1100 Series GPUs.
- Advanced Triton Integration: AOTriton 0.10b introduces official support for
gfx950 and gfx1201, along with experimental support for gfx1101, gfx1151,

4
docs/compatibility/ml-compatibility/taichi-compatibility.rst
View File

@@ -30,8 +30,8 @@ visual effects in film and gaming, and general-purpose computing.
Supported devices and features
===============================================================================
There is support through the ROCm software stack for all Taichi GPU features on AMD Instinct MI250X and MI210X series GPUs with the exception of Taichi’s GPU rendering system, CGUI.
AMD Instinct MI300X series GPUs will be supported by November.
There is support through the ROCm software stack for all Taichi GPU features on AMD Instinct MI250X and MI210X Series GPUs with the exception of Taichi’s GPU rendering system, CGUI.
AMD Instinct MI300X Series GPUs will be supported by November.
.. _taichi-recommendations:

12
docs/conceptual/gpu-arch.md
View File

@@ -13,22 +13,22 @@
:gutter: 1
:::{grid-item-card}
**AMD Instinct MI300 series**
**AMD Instinct MI300 Series**
Review hardware aspects of the AMD Instinct™ MI300 series of GPU accelerators and the CDNA™ 3
Review hardware aspects of the AMD Instinct™ MI300 Series GPUs and the CDNA™ 3
architecture.
* [AMD Instinct™ MI300 microarchitecture](./gpu-arch/mi300.md)
* [AMD Instinct MI300/CDNA3 ISA](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/instruction-set-architectures/amd-instinct-mi300-cdna3-instruction-set-architecture.pdf)
* [White paper](https://www.amd.com/content/dam/amd/en/documents/instinct-tech-docs/white-papers/amd-cdna-3-white-paper.pdf)
* [MI300 performance counters](./gpu-arch/mi300-mi200-performance-counters.rst)
* [MI350 series performance counters](./gpu-arch/mi350-performance-counters.rst)
* [MI350 Series performance counters](./gpu-arch/mi350-performance-counters.rst)
:::
:::{grid-item-card}
**AMD Instinct MI200 series**
**AMD Instinct MI200 Series**
Review hardware aspects of the AMD Instinct™ MI200 series of GPU accelerators and the CDNA™ 2
Review hardware aspects of the AMD Instinct™ MI200 Series GPUs and the CDNA™ 2
architecture.
* [AMD Instinct™ MI250 microarchitecture](./gpu-arch/mi250.md)
@@ -41,7 +41,7 @@ architecture.
:::{grid-item-card}
**AMD Instinct MI100**
Review hardware aspects of the AMD Instinct™ MI100 series of GPU accelerators and the CDNA™ 1
Review hardware aspects of the AMD Instinct™ MI100 Series GPUs and the CDNA™ 1
architecture.
* [AMD Instinct™ MI100 microarchitecture](./gpu-arch/mi100.md)

24
docs/conceptual/gpu-arch/mi100.md
View File

@@ -1,14 +1,14 @@
---
myst:
html_meta:
"description lang=en": "Learn about the AMD Instinct MI100 series architecture."
"description lang=en": "Learn about the AMD Instinct MI100 Series architecture."
"keywords": "Instinct, MI100, microarchitecture, AMD, ROCm"
---
# AMD Instinct™ MI100 microarchitecture
The following image shows the node-level architecture of a system that
comprises two AMD EPYC™ processors and (up to) eight AMD Instinct™ accelerators.
comprises two AMD EPYC™ processors and (up to) eight AMD Instinct™ GPUs.
The two EPYC processors are connected to each other with the AMD Infinity™
fabric which provides a high-bandwidth (up to 18 GT/sec) and coherent links such
that each processor can access the available node memory as a single
@@ -18,29 +18,29 @@ available to connect the processors plus one PCIe Gen 4 x16 link per processor
can attach additional I/O devices such as the host adapters for the network
fabric.
![Structure of a single GCD in the AMD Instinct MI100 accelerator](../../data/conceptual/gpu-arch/image004.png "Node-level system architecture with two AMD EPYC™ processors and eight AMD Instinct™ accelerators.")
![Structure of a single GCD in the AMD Instinct MI100 GPU](../../data/conceptual/gpu-arch/image004.png "Node-level system architecture with two AMD EPYC™ processors and eight AMD Instinct™ GPUs.")
In a typical node configuration, each processor can host up to four AMD
Instinct™ accelerators that are attached using PCIe Gen 4 links at 16 GT/sec,
Instinct™ GPUs that are attached using PCIe Gen 4 links at 16 GT/sec,
which corresponds to a peak bidirectional link bandwidth of 32 GB/sec. Each hive
of four accelerators can participate in a fully connected, coherent AMD
Instinct™ fabric that connects the four accelerators using 23 GT/sec AMD
of four GPUs can participate in a fully connected, coherent AMD
Instinct™ fabric that connects the four GPUs using 23 GT/sec AMD
Infinity fabric links that run at a higher frequency than the inter-processor
links. This inter-GPU link can be established in certified server systems if the
GPUs are mounted in neighboring PCIe slots by installing the AMD Infinity
Fabric™ bridge for the AMD Instinct™ accelerators.
Fabric™ bridge for the AMD Instinct™ GPUs.
## Microarchitecture
The microarchitecture of the AMD Instinct accelerators is based on the AMD CDNA
The microarchitecture of the AMD Instinct GPUs is based on the AMD CDNA
architecture, which targets compute applications such as high-performance
computing (HPC) and AI & machine learning (ML) that run on everything from
individual servers to the world's largest exascale supercomputers. The overall
system architecture is designed for extreme scalability and compute performance.
![Structure of the AMD Instinct accelerator (MI100 generation)](../../data/conceptual/gpu-arch/image005.png "Structure of the AMD Instinct accelerator (MI100 generation)")
![Structure of the AMD Instinct GPU (MI100 generation)](../../data/conceptual/gpu-arch/image005.png "Structure of the AMD Instinct GPU (MI100 generation)")
The above image shows the AMD Instinct accelerator with its PCIe Gen 4 x16
The above image shows the AMD Instinct GPU with its PCIe Gen 4 x16
link (16 GT/sec, at the bottom) that connects the GPU to (one of) the host
processor(s). It also shows the three AMD Infinity Fabric ports that provide
high-speed links (23 GT/sec, also at the bottom) to the other GPUs of the local
@@ -48,7 +48,7 @@ hive.
On the left and right of the floor plan, the High Bandwidth Memory (HBM)
attaches via the GPU memory controller. The MI100 generation of the AMD
Instinct accelerator offers four stacks of HBM generation 2 (HBM2) for a total
Instinct GPU offers four stacks of HBM generation 2 (HBM2) for a total
of 32GB with a 4,096bit-wide memory interface. The peak memory bandwidth of the
attached HBM2 is 1.228 TB/sec at a memory clock frequency of 1.2 GHz.
@@ -64,7 +64,7 @@ Therefore, the theoretical maximum FP64 peak performance is 11.5 TFLOPS
![Block diagram of an MI100 compute unit with detailed SIMD view of the AMD CDNA architecture](../../data/conceptual/gpu-arch/image006.png "An MI100 compute unit with detailed SIMD view of the AMD CDNA architecture")
The preceding image shows the block diagram of a single CU of an AMD Instinct™
MI100 accelerator and summarizes how instructions flow through the execution
MI100 GPU and summarizes how instructions flow through the execution
engines. The CU fetches the instructions via a 32KB instruction cache and moves
them forward to execution via a dispatcher. The CU can handle up to ten
wavefronts at a time and feed their instructions into the execution unit. The

23
docs/conceptual/gpu-arch/mi250.md
View File

@@ -1,13 +1,13 @@
---
myst:
html_meta:
"description lang=en": "Learn about the AMD Instinct MI250 series architecture."
"description lang=en": "Learn about the AMD Instinct MI250 Series architecture."
"keywords": "Instinct, MI250, microarchitecture, AMD, ROCm"
---
# AMD Instinct™ MI250 microarchitecture
The microarchitecture of the AMD Instinct MI250 accelerators is based on the
The microarchitecture of the AMD Instinct MI250 GPU is based on the
AMD CDNA 2 architecture that targets compute applications such as HPC,
artificial intelligence (AI), and machine learning (ML) and that run on
everything from individual servers to the world’s largest exascale
@@ -40,7 +40,7 @@ execution units (also called matrix cores), which are geared toward executing
matrix operations like matrix-matrix multiplications. For FP64, the peak
performance of these units amounts to 90.5 TFLOPS.
![Structure of a single GCD in the AMD Instinct MI250 accelerator.](../../data/conceptual/gpu-arch/image001.png "Structure of a single GCD in the AMD Instinct MI250 accelerator.")
![Structure of a single GCD in the AMD Instinct MI250 GPU.](../../data/conceptual/gpu-arch/image001.png "Structure of a single GCD in the AMD Instinct MI250 GPU.")
```{list-table} Peak-performance capabilities of the MI250 OAM for different data types.
:header-rows: 1
@@ -84,16 +84,9 @@ performance of these units amounts to 90.5 TFLOPS.
- 362.1
```
The above table summarizes the aggregated peak performance of the AMD
Instinct MI250 OCP Open Accelerator Modules (OAM, OCP is short for Open Compute
Platform) and its two GCDs for different data types and execution units. The
middle column lists the peak performance (number of data elements processed in a
single instruction) of a single compute unit if a SIMD (or matrix) instruction
is being retired in each clock cycle. The third column lists the theoretical
peak performance of the OAM module. The theoretical aggregated peak memory
bandwidth of the GPU is 3.2 TB/sec (1.6 TB/sec per GCD).
The above table summarizes the aggregated peak performance of the AMD Instinct MI250 Open Compute Platform (OCP) Open Accelerator Modules (OAMs) and its two GCDs for different data types and execution units. The middle column lists the peak performance (number of data elements processed in a single instruction) of a single compute unit if a SIMD (or matrix) instruction is being retired in each clock cycle. The third column lists the theoretical peak performance of the OAM module. The theoretical aggregated peak memory bandwidth of the GPU is 3.2 TB/sec (1.6 TB/sec per GCD).
![Dual-GCD architecture of the AMD Instinct MI250 accelerators](../../data/conceptual/gpu-arch/image002.png "Dual-GCD architecture of the AMD Instinct MI250 accelerators")
![Dual-GCD architecture of the AMD Instinct MI250 GPUs](../../data/conceptual/gpu-arch/image002.png "Dual-GCD architecture of the AMD Instinct MI250 GPUs")
The following image shows the block diagram of an OAM package that consists
of two GCDs, each of which constitutes one GPU device in the system. The two
@@ -105,18 +98,18 @@ between the two GCDs of an OAM, or a bidirectional peak transfer bandwidth of
## Node-level architecture
The following image shows the node-level architecture of a system that is
based on the AMD Instinct MI250 accelerator. The MI250 OAMs attach to the host
based on the AMD Instinct MI250 GPU. The MI250 OAMs attach to the host
system via PCIe Gen 4 x16 links (yellow lines). Each GCD maintains its own PCIe
x16 link to the host part of the system. Depending on the server platform, the
GCD can attach to the AMD EPYC processor directly or via an optional PCIe switch
. Note that some platforms may offer an x8 interface to the GCDs, which reduces
the available host-to-GPU bandwidth.
![Block diagram of AMD Instinct MI250 Accelerators with 3rd Generation AMD EPYC processor](../../data/conceptual/gpu-arch/image003.png "Block diagram of AMD Instinct MI250 Accelerators with 3rd Generation AMD EPYC processor")
![Block diagram of AMD Instinct MI250 GPUs with 3rd Generation AMD EPYC processor](../../data/conceptual/gpu-arch/image003.png "Block diagram of AMD Instinct MI250 GPUs with 3rd Generation AMD EPYC processor")
The preceding image shows the node-level architecture of a system with AMD
EPYC processors in a dual-socket configuration and four AMD Instinct MI250
accelerators. The MI250 OAMs attach to the host processors system via PCIe Gen 4
GPUs. The MI250 OAMs attach to the host processors system via PCIe Gen 4
x16 links (yellow lines). Depending on the system design, a PCIe switch may
exist to make more PCIe lanes available for additional components like network
interfaces and/or storage devices. Each GCD maintains its own PCIe x16 link to

18
docs/conceptual/gpu-arch/mi300-mi200-performance-counters.rst
View File

@@ -1,16 +1,16 @@
.. meta::
:description: MI300 and MI200 series performance counters and metrics
:description: MI300 and MI200 Series performance counters and metrics
:keywords: MI300, MI200, performance counters, command processor counters
***************************************************************************************************
MI300 and MI200 series performance counters and metrics
MI300 and MI200 Series performance counters and metrics
***************************************************************************************************
This document lists and describes the hardware performance counters and derived metrics available
for the AMD Instinct™ MI300 and MI200 GPU. You can also access this information using the
:doc:`ROCprofiler-SDK <rocprofiler-sdk:how-to/using-rocprofv3>`.
MI300 and MI200 series performance counters
MI300 and MI200 Series performance counters
===============================================================
Series performance counters include the following categories:
@@ -27,7 +27,7 @@ The following sections provide additional details for each category.
.. note::
Preliminary validation of all MI300 and MI200 series performance counters is in progress. Those with
Preliminary validation of all MI300 and MI200 Series performance counters is in progress. Those with
an asterisk (*) require further evaluation.
.. _command-processor-counters:
@@ -171,7 +171,7 @@ Instruction mix
"``SQ_INSTS_SMEM``", "Instr", "Number of scalar memory instructions issued"
"``SQ_INSTS_SMEM_NORM``", "Instr", "Number of scalar memory instructions normalized to match ``smem_level`` issued"
"``SQ_INSTS_FLAT``", "Instr", "Number of flat instructions issued"
"``SQ_INSTS_FLAT_LDS_ONLY``", "Instr", "**MI200 series only** Number of FLAT instructions that read/write only from/to LDS issued. Works only if ``EARLY_TA_DONE`` is enabled."
"``SQ_INSTS_FLAT_LDS_ONLY``", "Instr", "**MI200 Series only** Number of FLAT instructions that read/write only from/to LDS issued. Works only if ``EARLY_TA_DONE`` is enabled."
"``SQ_INSTS_LDS``", "Instr", "Number of LDS instructions issued **(MI200: includes flat; MI300: does not include flat)**"
"``SQ_INSTS_GDS``", "Instr", "Number of global data share instructions issued"
"``SQ_INSTS_EXP_GDS``", "Instr", "Number of EXP and global data share instructions excluding skipped export instructions issued"
@@ -396,9 +396,9 @@ Texture cache per pipe counters
"``TCP_UTCL1_TRANSLATION_MISS[n]``", "Req", "Number of unified translation cache (L1) translation misses", "0-15"
"``TCP_UTCL1_PERMISSION_MISS[n]``", "Req", "Number of unified translation cache (L1) permission misses", "0-15"
"``TCP_TOTAL_CACHE_ACCESSES[n]``", "Req", "Number of vector L1d cache accesses including hits and misses", "0-15"
"``TCP_TCP_LATENCY[n]``", "Cycles", "**MI200 series only** Accumulated wave access latency to vL1D over all wavefronts", "0-15"
"``TCP_TCC_READ_REQ_LATENCY[n]``", "Cycles", "**MI200 series only** Total vL1D to L2 request latency over all wavefronts for reads and atomics with return", "0-15"
"``TCP_TCC_WRITE_REQ_LATENCY[n]``", "Cycles", "**MI200 series only** Total vL1D to L2 request latency over all wavefronts for writes and atomics without return", "0-15"
"``TCP_TCP_LATENCY[n]``", "Cycles", "**MI200 Series only** Accumulated wave access latency to vL1D over all wavefronts", "0-15"
"``TCP_TCC_READ_REQ_LATENCY[n]``", "Cycles", "**MI200 Series only** Total vL1D to L2 request latency over all wavefronts for reads and atomics with return", "0-15"
"``TCP_TCC_WRITE_REQ_LATENCY[n]``", "Cycles", "**MI200 Series only** Total vL1D to L2 request latency over all wavefronts for writes and atomics without return", "0-15"
"``TCP_TCC_READ_REQ[n]``", "Req", "Number of read requests to L2 cache", "0-15"
"``TCP_TCC_WRITE_REQ[n]``", "Req", "Number of write requests to L2 cache", "0-15"
"``TCP_TCC_ATOMIC_WITH_RET_REQ[n]``", "Req", "Number of atomic requests to L2 cache with return", "0-15"
@@ -560,7 +560,7 @@ Note the following:
``TCC_TAG_STALL[n]``, probes can stall the pipeline at a variety of places. There is no single point that
can accurately measure the total stalls
MI300 and MI200 series derived metrics list
MI300 and MI200 Series derived metrics list
==============================================================
.. csv-table::

20
docs/conceptual/gpu-arch/mi300.md
View File

@@ -1,21 +1,21 @@
---
myst:
html_meta:
"description lang=en": "Learn about the AMD Instinct MI300 series architecture."
"description lang=en": "Learn about the AMD Instinct MI300 Series architecture."
"keywords": "Instinct, MI300X, MI300A, microarchitecture, AMD, ROCm"
---
# AMD Instinct™ MI300 series microarchitecture
# AMD Instinct™ MI300 Series microarchitecture
The AMD Instinct MI300 series accelerators are based on the AMD CDNA 3
The AMD Instinct MI300 Series GPUs are based on the AMD CDNA 3
architecture which was designed to deliver leadership performance for HPC, artificial intelligence (AI), and machine
learning (ML) workloads. The AMD Instinct MI300 series accelerators are well-suited for extreme scalability and compute performance, running
learning (ML) workloads. The AMD Instinct MI300 Series GPUs are well-suited for extreme scalability and compute performance, running
on everything from individual servers to the world’s largest exascale supercomputers.
With the MI300 series, AMD is introducing the Accelerator Complex Die (XCD), which contains the
With the MI300 Series, AMD is introducing the Accelerator Complex Die (XCD), which contains the
GPU computational elements of the processor along with the lower levels of the cache hierarchy.
The following image depicts the structure of a single XCD in the AMD Instinct MI300 accelerator series.
The following image depicts the structure of a single XCD in the AMD Instinct MI300 GPU Series.
```{figure} ../../data/shared/xcd-sys-arch.png
---
@@ -39,7 +39,7 @@ infrastructure) using the AMD Infinity Fabric™ technology as interconnect.
The Matrix Cores inside the CDNA 3 CUs have significant improvements, emphasizing AI and machine
learning, enhancing throughput of existing data types while adding support for new data types.
CDNA 2 Matrix Cores support FP16 and BF16, while offering INT8 for inference. Compared to MI250X
accelerators, CDNA 3 Matrix Cores triple the performance for FP16 and BF16, while providing a
GPUs, CDNA 3 Matrix Cores triple the performance for FP16 and BF16, while providing a
performance gain of 6.8 times for INT8. FP8 has a performance gain of 16 times compared to FP32,
while TF32 has a gain of 4 times compared to FP32.
@@ -105,7 +105,7 @@ name: mi300-arch
alt:
align: center
---
MI300 series system architecture showing MI300A (left) with 6 XCDs and 3 CCDs, while the MI300X (right) has 8 XCDs.
MI300 Series system architecture showing MI300A (left) with 6 XCDs and 3 CCDs, while the MI300X (right) has 8 XCDs.
```
## Node-level architecture
@@ -116,11 +116,11 @@ name: mi300-node
align: center
---
MI300 series node-level architecture showing 8 fully interconnected MI300X OAM modules connected to (optional) PCIEe switches via retimers and HGX connectors.
MI300 Series node-level architecture showing 8 fully interconnected MI300X OAM modules connected to (optional) PCIEe switches via retimers and HGX connectors.
```
The image above shows the node-level architecture of a system with AMD EPYC processors in a
dual-socket configuration and eight AMD Instinct MI300X accelerators. The MI300X OAMs attach to the
dual-socket configuration and eight AMD Instinct MI300X GPUs. The MI300X OAMs attach to the
host system via PCIe Gen 5 x16 links (yellow lines). The GPUs are using seven high-bandwidth,
low-latency AMD Infinity Fabric™ links (red lines) to form a fully connected 8-GPU system.

6
docs/conceptual/gpu-arch/mi350-performance-counters.rst
View File

@@ -1,12 +1,12 @@
.. meta::
:description: MI355 series performance counters and metrics
:description: MI355 Series performance counters and metrics
:keywords: MI355, MI355X, MI3XX
***********************************
MI350 series performance counters
MI350 Series performance counters
***********************************
This topic lists and describes the hardware performance counters and derived metrics available on the AMD Instinct MI350 and MI355 accelerators. These counters are available for profiling using `ROCprofiler-SDK <https://rocm.docs.amd.com/projects/rocprofiler-sdk/en/latest/index.html>`_ and `ROCm Compute Profiler <https://rocm.docs.amd.com/projects/rocprofiler-compute/en/latest/>`_.
This topic lists and describes the hardware performance counters and derived metrics available on the AMD Instinct MI350 and MI355 GPUs. These counters are available for profiling using `ROCprofiler-SDK <https://rocm.docs.amd.com/projects/rocprofiler-sdk/en/latest/index.html>`_ and `ROCm Compute Profiler <https://rocm.docs.amd.com/projects/rocprofiler-compute/en/latest/>`_.
The following sections list the performance counters based on the IP blocks.

2
docs/conf.py
View File

@@ -234,7 +234,7 @@ suppress_warnings = ["autosectionlabel.*"]
html_context = {
"project_path" : {project_path},
"gpu_type" : [('AMD Instinct accelerators', 'intrinsic'), ('AMD gfx families', 'gfx'), ('NVIDIA families', 'nvidia') ],
"gpu_type" : [('AMD Instinct GPUs', 'intrinsic'), ('AMD gfx families', 'gfx'), ('NVIDIA families', 'nvidia') ],
"atomics_type" : [('HW atomics', 'hw-atomics'), ('CAS emulation', 'cas-atomics')],
"pcie_type" : [('No PCIe atomics', 'nopcie'), ('PCIe atomics', 'pcie')],
"memory_type" : [('Device DRAM', 'device-dram'), ('Migratable Host DRAM', 'migratable-host-dram'), ('Pinned Host DRAM', 'pinned-host-dram')],

2
docs/data/reference/gpu-atomics-operation/cas-atomics_nopcie_instinct.csv
View File

@@ -1,4 +1,4 @@
Atomic,MI100,MI200 PCIe,MI200 A+A,MI300X series,MI300A,MI350X series
Atomic,MI100,MI200 PCIe,MI200 A+A,MI300X Series,MI300A,MI350X Series
32 bit atomicAdd,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS
32 bit atomicSub,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS
32 bit atomicMin,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS
1 Atomic MI100 MI200 PCIe MI200 A+A MI300X series MI300X Series MI300A MI350X series MI350X Series
2 32 bit atomicAdd ✅ CAS ✅ CAS ✅ CAS ✅ CAS ✅ CAS ✅ CAS
3 32 bit atomicSub ✅ CAS ✅ CAS ✅ CAS ✅ CAS ✅ CAS ✅ CAS
4 32 bit atomicMin ✅ CAS ✅ CAS ✅ CAS ✅ CAS ✅ CAS ✅ CAS

2
docs/data/reference/gpu-atomics-operation/cas-atomics_pcie_instinct.csv
View File

@@ -1,4 +1,4 @@
Atomic,MI100,MI200 PCIe,MI200 A+A,MI300X series,MI300A,MI350X series
Atomic,MI100,MI200 PCIe,MI200 A+A,MI300X Series,MI300A,MI350X Series
32 bit atomicAdd,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS
32 bit atomicSub,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS
32 bit atomicMin,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS,✅ CAS
1 Atomic MI100 MI200 PCIe MI200 A+A MI300X series MI300X Series MI300A MI350X series MI350X Series
2 32 bit atomicAdd ✅ CAS ✅ CAS ✅ CAS ✅ CAS ✅ CAS ✅ CAS
3 32 bit atomicSub ✅ CAS ✅ CAS ✅ CAS ✅ CAS ✅ CAS ✅ CAS
4 32 bit atomicMin ✅ CAS ✅ CAS ✅ CAS ✅ CAS ✅ CAS ✅ CAS

2
docs/data/reference/gpu-atomics-operation/hw-atomics_nopcie_instinct.csv
View File

@@ -1,4 +1,4 @@
Atomic,MI100,MI200 PCIe,MI200 A+A,MI300X series,MI300A,MI350X series
Atomic,MI100,MI200 PCIe,MI200 A+A,MI300X Series,MI300A,MI350X Series
32 bit atomicAdd,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native
32 bit atomicSub,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native
32 bit atomicMin,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native
1 Atomic MI100 MI200 PCIe MI200 A+A MI300X series MI300X Series MI300A MI350X series MI350X Series
2 32 bit atomicAdd ✅ Native ✅ Native ✅ Native ✅ Native ✅ Native ✅ Native
3 32 bit atomicSub ✅ Native ✅ Native ✅ Native ✅ Native ✅ Native ✅ Native
4 32 bit atomicMin ✅ Native ✅ Native ✅ Native ✅ Native ✅ Native ✅ Native

2
docs/data/reference/gpu-atomics-operation/hw-atomics_pcie_instinct.csv
View File

@@ -1,4 +1,4 @@
Atomic,MI100,MI200 PCIe,MI200 A+A,MI300X series,MI300A,MI350X series
Atomic,MI100,MI200 PCIe,MI200 A+A,MI300X Series,MI300A,MI350X Series
32 bit atomicAdd,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native
32 bit atomicSub,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native
32 bit atomicMin,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native,✅ Native
1 Atomic MI100 MI200 PCIe MI200 A+A MI300X series MI300X Series MI300A MI350X series MI350X Series
2 32 bit atomicAdd ✅ Native ✅ Native ✅ Native ✅ Native ✅ Native ✅ Native
3 32 bit atomicSub ✅ Native ✅ Native ✅ Native ✅ Native ✅ Native ✅ Native
4 32 bit atomicMin ✅ Native ✅ Native ✅ Native ✅ Native ✅ Native ✅ Native

2
docs/how-to/deep-learning-rocm.rst
View File

@@ -10,7 +10,7 @@ Deep learning frameworks provide environments for machine learning, training, fi
ROCm offers a complete ecosystem for developing and running deep learning applications efficiently. It also provides ROCm-compatible versions of popular frameworks and libraries, such as PyTorch, TensorFlow, JAX, and others.
The AMD ROCm organization actively contributes to open-source development and collaborates closely with framework organizations. This collaboration ensures that framework-specific optimizations effectively leverage AMD GPUs and accelerators.
The AMD ROCm organization actively contributes to open-source development and collaborates closely with framework organizations. This collaboration ensures that framework-specific optimizations effectively leverage AMD GPUs.
The table below summarizes information about ROCm-enabled deep learning frameworks. It includes details on ROCm compatibility and third-party tool support, installation steps and options, and links to GitHub resources. For a complete list of supported framework versions on ROCm, see the :doc:`Compatibility matrix <../compatibility/compatibility-matrix>` topic.

10
docs/how-to/gpu-performance/mi300x.rst
View File

@@ -1,5 +1,5 @@
.. meta::
:description: How to configure MI300X accelerators to fully leverage their capabilities and achieve optimal performance.
:description: How to configure MI300X GPUs to fully leverage their capabilities and achieve optimal performance.
:keywords: ROCm, AI, machine learning, MI300X, LLM, usage, tutorial, optimization, tuning
**************************************
@@ -7,11 +7,11 @@ AMD Instinct MI300X performance guides
**************************************
The following performance guides provide essential guidance on the necessary
steps to properly `configure your system for AMD Instinct™ MI300X accelerators
steps to properly `configure your system for AMD Instinct™ MI300X GPUs
<https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
They include detailed instructions on system settings and application
:doc:`workload tuning </how-to/rocm-for-ai/inference-optimization/workload>` to
help you leverage the maximum capabilities of these accelerators and achieve
help you leverage the maximum capabilities of these GPUs and achieve
superior performance.
* `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`__
@@ -19,9 +19,9 @@ superior performance.
your AMD Instinct MI300X system for performance.
* :doc:`/how-to/rocm-for-ai/inference-optimization/workload` covers steps to
optimize the performance of AMD Instinct MI300X series accelerators for HPC
optimize the performance of AMD Instinct MI300X Series GPUs for HPC
and deep learning operations.
* :doc:`/how-to/rocm-for-ai/inference/benchmark-docker/vllm` introduces a preconfigured
environment for LLM inference, designed to help you test performance with
popular models on AMD Instinct MI300X series accelerators.
popular models on AMD Instinct MI300X Series GPUs.

2
docs/how-to/programming_guide.rst
View File

@@ -25,7 +25,7 @@ execute on AMD GPUs while maintaining compatibility with CUDA-based systems.
OpenCL (Open Computing Language) is an open standard for cross-platform,
parallel programming of diverse processors. ROCm supports OpenCL for developers
who want to use standard frameworks across different hardware platforms,
including CPUs, GPUs, and other accelerators. For more information, see
including CPUs, GPUs, and APUs. For more information, see
`OpenCL <https://www.khronos.org/opencl/>`_.
Python bindings can be found at https://github.com/ROCm/hip-python.

4
docs/how-to/rocm-for-ai/fine-tuning/fine-tuning-and-inference.rst
View File

@@ -11,10 +11,10 @@ Fine-tuning using ROCm involves leveraging AMD's GPU-accelerated :doc:`libraries
ecosystem for deep learning development, including open-source libraries for optimized deep learning operations and
ROCm-aware versions of :doc:`deep learning frameworks <../../deep-learning-rocm>` such as PyTorch, TensorFlow, and JAX.
Single-accelerator systems, such as a machine equipped with a single accelerator or GPU, are commonly used for
Single-accelerator systems, such as a machine equipped with a single GPU, are commonly used for
smaller-scale deep learning tasks, including fine-tuning pre-trained models and running inference on moderately
sized datasets. See :doc:`single-gpu-fine-tuning-and-inference`.
Multi-accelerator systems, on the other hand, consist of multiple accelerators working in parallel. These systems are
Multi-accelerator systems, on the other hand, consist of multiple GPUs working in parallel. These systems are
typically used in LLMs and other large-scale deep learning tasks where performance, scalability, and the handling of
massive datasets are crucial. See :doc:`multi-gpu-fine-tuning-and-inference`.

24
docs/how-to/rocm-for-ai/fine-tuning/multi-gpu-fine-tuning-and-inference.rst
View File

@@ -3,11 +3,11 @@
:keywords: ROCm, LLM, fine-tuning, usage, tutorial, multi-GPU, distributed, inference, accelerators, PyTorch, HuggingFace, torchtune
*****************************************************
Fine-tuning and inference using multiple accelerators
Fine-tuning and inference using multiple GPUs
*****************************************************
This section explains how to fine-tune a model on a multi-accelerator system. See
:doc:`Single-accelerator fine-tuning <single-gpu-fine-tuning-and-inference>` for a single accelerator or GPU setup.
:doc:`Single-accelerator fine-tuning <single-gpu-fine-tuning-and-inference>` for a single GPU setup.
.. _fine-tuning-llms-multi-gpu-env:
@@ -20,7 +20,7 @@ This section was tested using the following hardware and software environment.
:stub-columns: 1
* - Hardware
- 4 AMD Instinct MI300X accelerators
- 4 AMD Instinct MI300X GPUs
* - Software
- ROCm 6.1, Ubuntu 22.04, PyTorch 2.1.2, Python 3.10
@@ -40,13 +40,13 @@ Setting up the base implementation environment
:doc:`PyTorch installation guide <rocm-install-on-linux:install/3rd-party/pytorch-install>`. For consistent
installation, it’s recommended to use official ROCm prebuilt Docker images with the framework pre-installed.
#. In the Docker container, check the availability of ROCM-capable accelerators using the following command.
#. In the Docker container, check the availability of ROCm-capable GPUs using the following command.
.. code-block:: shell
rocm-smi --showproductname
#. Check that your accelerators are available to PyTorch.
#. Check that your GPUs are available to PyTorch.
.. code-block:: python
@@ -66,7 +66,7 @@ Setting up the base implementation environment
.. tip::
During training and inference, you can check the memory usage by running the ``rocm-smi`` command in your terminal.
This tool helps you see shows which accelerators or GPUs are involved.
This tool helps you see shows which GPUs are involved.
.. _fine-tuning-llms-multi-gpu-hugging-face-accelerate:
@@ -74,9 +74,9 @@ Setting up the base implementation environment
Hugging Face Accelerate for fine-tuning and inference
===========================================================
`Hugging Face Accelerate <https://huggingface.co/docs/accelerate/en/index>`_ is a library that simplifies turning raw
PyTorch code for a single accelerator into code for multiple accelerators for LLM fine-tuning and inference. It is
integrated with `Transformers <https://huggingface.co/docs/transformers/en/index>`_ allowing you to scale your PyTorch
`Hugging Face Accelerate <https://huggingface.co/docs/accelerate/en/index>`__ is a library that simplifies turning raw
PyTorch code for a single GPU into code for multiple GPUs for LLM fine-tuning and inference. It is
integrated with `Transformers <https://huggingface.co/docs/transformers/en/index>`__, so you can scale your PyTorch
code while maintaining performance and flexibility.
As a brief example of model fine-tuning and inference using multiple GPUs, let's use Transformers and load in the Llama
@@ -107,7 +107,7 @@ Now, it's important to adjust how you load the model. Add the ``device_map`` par
(``"auto"``, ``"balanced"``, ``"balanced_low_0"``, ``"sequential"``).
It's recommended to set the ``device_map`` parameter to ``“auto”`` to allow Accelerate to automatically and
efficiently allocate the model given the available resources (4 accelerators in this case).
efficiently allocate the model given the available resources (four GPUs in this case).
When you have more GPU memory available than the model size, here is the difference between each ``device_map``
option:
@@ -130,8 +130,8 @@ After loading the model in this way, the model is fully ready to use the resourc
torchtune for fine-tuning and inference
=============================================
`torchtune <https://pytorch.org/torchtune/main/>`_ is a PyTorch-native library for easy single and multi-accelerator or
GPU model fine-tuning and inference with LLMs.
`torchtune <https://pytorch.org/torchtune/main/>`_ is a PyTorch-native library for easy single and multi-GPU
model fine-tuning and inference with LLMs.
#. Install torchtune using pip.

6
docs/how-to/rocm-for-ai/fine-tuning/overview.rst
View File

@@ -30,7 +30,7 @@ The challenge of fine-tuning models
However, the computational cost of fine-tuning is still high, especially for complex models and large datasets, which
poses distinct challenges related to substantial computational and memory requirements. This might be a barrier for
accelerators or GPUs with low computing power or limited device memory resources.
GPUs with low computing power or limited device memory resources.
For example, suppose we have a language model with 7 billion (7B) parameters, represented by a weight matrix :math:`W`.
During backpropagation, the model needs to learn a :math:`ΔW` matrix, which updates the original weights to minimize the
@@ -84,8 +84,8 @@ Walkthrough
===========
To demonstrate the benefits of LoRA and the ideal compute compatibility of using PEFT and TRL libraries on AMD
ROCm-compatible accelerators and GPUs, let's step through a comprehensive implementation of the fine-tuning process
using the Llama 2 7B model with LoRA tailored specifically for question-and-answer tasks on AMD MI300X accelerators.
ROCm-compatible GPUs, let's step through a comprehensive implementation of the fine-tuning process
using the Llama 2 7B model with LoRA tailored specifically for question-and-answer tasks on AMD MI300X GPUs.
Before starting, review and understand the key components of this walkthrough:

17
docs/how-to/rocm-for-ai/fine-tuning/single-gpu-fine-tuning-and-inference.rst
View File

@@ -3,12 +3,11 @@
:keywords: ROCm, LLM, fine-tuning, usage, tutorial, single-GPU, LoRA, PEFT, inference, SFTTrainer
****************************************************
Fine-tuning and inference using a single accelerator
Fine-tuning and inference using a single GPU
****************************************************
This section explains model fine-tuning and inference techniques on a single-accelerator system. See
:doc:`Multi-accelerator fine-tuning <multi-gpu-fine-tuning-and-inference>` for a setup with multiple accelerators or
GPUs.
:doc:`Multi-accelerator fine-tuning <multi-gpu-fine-tuning-and-inference>` for a setup with multiple GPUs.
.. _fine-tuning-llms-single-gpu-env:
@@ -21,7 +20,7 @@ This section was tested using the following hardware and software environment.
:stub-columns: 1
* - Hardware
- AMD Instinct MI300X accelerator
- AMD Instinct MI300X GPU
* - Software
- ROCm 6.1, Ubuntu 22.04, PyTorch 2.1.2, Python 3.10
@@ -41,7 +40,7 @@ Setting up the base implementation environment
:doc:`PyTorch installation guide <rocm-install-on-linux:install/3rd-party/pytorch-install>`. For a consistent
installation, it’s recommended to use official ROCm prebuilt Docker images with the framework pre-installed.
#. In the Docker container, check the availability of ROCm-capable accelerators using the following command.
#. In the Docker container, check the availability of ROCm-capable GPUs using the following command.
.. code-block:: shell
@@ -53,14 +52,14 @@ Setting up the base implementation environment
============================ ROCm System Management Interface ============================
====================================== Product Info ======================================
GPU[0] : Card series: AMD Instinct MI300X OAM
GPU[0] : Card Series: AMD Instinct MI300X OAM
GPU[0] : Card model: 0x74a1
GPU[0] : Card vendor: Advanced Micro Devices, Inc. [AMD/ATI]
GPU[0] : Card SKU: MI3SRIOV
==========================================================================================
================================== End of ROCm SMI Log ===================================
#. Check that your accelerators are available to PyTorch.
#. Check that your GPUs are available to PyTorch.
.. code-block:: python
@@ -502,9 +501,9 @@ Let's look at achieving model inference using these types of models.
# Token generation
print(pipe("What is a large language model?")[0]["generated_text"])
If using multiple accelerators, see
If using multiple GPUs, see
:ref:`Multi-accelerator fine-tuning and inference <fine-tuning-llms-multi-gpu-hugging-face-accelerate>` to explore
popular libraries that simplify fine-tuning and inference in a multi-accelerator system.
popular libraries that simplify fine-tuning and inference in a multiple-GPU system.
Read more about inference frameworks like vLLM and Hugging Face TGI in
:doc:`LLM inference frameworks <../inference/llm-inference-frameworks>`.

4
docs/how-to/rocm-for-ai/inference-optimization/model-acceleration-libraries.rst
View File

@@ -45,7 +45,7 @@ ROCm provides two different implementations of Flash Attention 2 modules. They c
# Install from source
git clone https://github.com/ROCm/flash-attention.git
cd flash-attention/
GPU_ARCHS=gfx942 python setup.py install #MI300 series
GPU_ARCHS=gfx942 python setup.py install #MI300 Series
Hugging Face Transformers can easily deploy the CK Flash Attention 2 module by passing an argument
``attn_implementation="flash_attention_2"`` in the ``from_pretrained`` class.
@@ -526,7 +526,7 @@ follow these instructions:
python -m pytest -v -rsx -s -W ignore::pytest.PytestCollectionWarning split_table_batched_embeddings_test.py
To run the FBGEMM_GPU ``uvm`` test, use these commands. These tests only support the AMD MI210 and
more recent accelerators.
more recent GPUs.
.. code-block:: shell

6
docs/how-to/rocm-for-ai/inference-optimization/model-quantization.rst
View File

@@ -7,7 +7,7 @@ Model quantization techniques
*****************************
Quantization reduces the model size compared to its native full-precision version, making it easier to fit large models
onto accelerators or GPUs with limited memory usage. This section explains how to perform LLM quantization using AMD Quark, GPTQ
onto GPUs with limited memory usage. This section explains how to perform LLM quantization using AMD Quark, GPTQ
and bitsandbytes on AMD Instinct hardware.
.. _quantize-llms-quark:
@@ -311,7 +311,7 @@ ExLlama-v2 support
ExLlama is a Python/C++/CUDA implementation of the Llama model that is
designed for faster inference with 4-bit GPTQ weights. The ExLlama
kernel is activated by default when users create a ``GPTQConfig`` object. To
boost inference speed even further on Instinct accelerators, use the ExLlama-v2
boost inference speed even further on Instinct GPUs, use the ExLlama-v2
kernels by configuring the ``exllama_config`` parameter as the following.
.. code-block:: python
@@ -332,7 +332,7 @@ The `ROCm-aware bitsandbytes <https://github.com/ROCm/bitsandbytes>`_ library is
a lightweight Python wrapper around CUDA custom functions, in particular 8-bit optimizer, matrix multiplication, and
8-bit and 4-bit quantization functions. The library includes quantization primitives for 8-bit and 4-bit operations
through ``bitsandbytes.nn.Linear8bitLt`` and ``bitsandbytes.nn.Linear4bit`` and 8-bit optimizers through the
``bitsandbytes.optim`` module. These modules are supported on AMD Instinct accelerators.
``bitsandbytes.optim`` module. These modules are supported on AMD Instinct GPUs.
Installing bitsandbytes
-----------------------

20
docs/how-to/rocm-for-ai/inference-optimization/optimizing-with-composable-kernel.md
View File

@@ -9,13 +9,13 @@ myst:
The AMD ROCm Composable Kernel (CK) library provides a programming model for writing performance-critical kernels for machine learning workloads. It generates a general-purpose kernel during the compilation phase through a C++ template, enabling developers to achieve operation fusions on different data precisions.
This article gives a high-level overview of CK General Matrix Multiplication (GEMM) kernel based on the design example of `03_gemm_bias_relu`. It also outlines the steps to construct the kernel and run it. Moreover, the article provides a detailed implementation of running SmoothQuant quantized INT8 models on AMD Instinct MI300X accelerators using CK.
This article gives a high-level overview of CK General Matrix Multiplication (GEMM) kernel based on the design example of `03_gemm_bias_relu`. It also outlines the steps to construct the kernel and run it. Moreover, the article provides a detailed implementation of running SmoothQuant quantized INT8 models on AMD Instinct MI300X GPUs using CK.
## High-level overview: a CK GEMM instance
GEMM is a fundamental block in linear algebra, machine learning, and deep neural networks. It is defined as the operation:
{math}`E = α \times (A \times B) + β \times (D)`, with A and B as matrix inputs, α and β as scalar inputs, and D as a pre-existing matrix.
Take the commonly used linear transformation in a fully connected layer as an example. These terms correspond to input activation (A), weight (B), bias (D), and output (E), respectively. The example employs a `DeviceGemmMultipleD_Xdl_CShuffle` struct from CK library as the fundamental instance to explore the compute capability of AMD Instinct accelerators for the computation of GEMM. The implementation of the instance contains two phases:
Take the commonly used linear transformation in a fully connected layer as an example. These terms correspond to input activation (A), weight (B), bias (D), and output (E), respectively. The example employs a `DeviceGemmMultipleD_Xdl_CShuffle` struct from CK library as the fundamental instance to explore the compute capability of AMD Instinct GPUs for the computation of GEMM. The implementation of the instance contains two phases:
- [Template parameter definition](#template-parameter-definition)
- [Instantiating and running the templated kernel](#instantiating-and-running-the-templated-kernel)
@@ -108,7 +108,7 @@ These parameters include Block Size, M/N/K Per Block, M/N per XDL, AK1, BK1, etc
- Block Size determines the number of threads in the thread block.
- M/N/K Per Block determines the size of tile that each thread block is responsible for calculating.
- M/N Per XDL refers to M/N size for Instinct accelerator Matrix Fused Multiply Add (MFMA) instructions operating on a per-wavefront basis.
- M/N Per XDL refers to M/N size for Instinct GPU Matrix Fused Multiply Add (MFMA) instructions operating on a per-wavefront basis.
- A/B K1 is related to the data type. It can be any value ranging from 1 to K Per Block. To achieve the optimal load/store performance, 128bit per load is suggested. In addition, the A/B loading parameters must be changed accordingly to match the A/B K1 value; otherwise, it will result in compilation errors.
Conditions for achieving computational load balancing on different hardware platforms can vary.
@@ -133,7 +133,7 @@ Templated kernel launching consists of kernel instantiation, making arguments by
## Developing fused INT8 kernels for SmoothQuant models
[SmoothQuant](https://github.com/mit-han-lab/smoothquant) (SQ) is a quantization algorithm that enables an INT8 quantization of both weights and activations for all the matrix multiplications in LLM. The required GPU kernel functionalities used to accelerate the inference of SQ models on Instinct accelerators are shown in the following table.
[SmoothQuant](https://github.com/mit-han-lab/smoothquant) (SQ) is a quantization algorithm that enables an INT8 quantization of both weights and activations for all the matrix multiplications in LLM. The required GPU kernel functionalities used to accelerate the inference of SQ models on Instinct GPUs are shown in the following table.
:::{table} Functionalities used to implement SmoothQuant model inference.
@@ -164,7 +164,7 @@ The CK library contains many fundamental instances that implement different func
Second, consider whether the format of input data meets your actual calculation needs. For SQ models, the 8-bit integer data format (INT8) is applied for matrix calculations.
Third, consider the platform for implementing CK instances. The instances suffixed with `xdl` only run on AMD Instinct accelerators after being compiled and cannot run on Radeon-series GPUs. This is due to the underlying device-specific instruction sets for implementing these basic instances.
Third, consider the platform for implementing CK instances. The instances suffixed with `xdl` only run on AMD Instinct GPUs after being compiled and cannot run on Radeon-Series GPUs. This is due to the underlying device-specific instruction sets for implementing these basic instances.
Here, we use [DeviceBatchedGemmMultiD_Xdl](https://github.com/ROCm/composable_kernel/tree/develop/example/24_batched_gemm) as the fundamental instance to implement the functionalities in the previous table.
@@ -435,7 +435,7 @@ The implementation architecture of running SmoothQuant models on MI300X GPUs is
### Figure 7
================ -->
```{figure} ../../../data/how-to/llm-fine-tuning-optimization/ck-inference_flow.jpg
The implementation architecture of running SmoothQuant models on AMD MI300X accelerators.
The implementation architecture of running SmoothQuant models on AMD MI300X GPUs.
```
For the target [SQ quantized model](https://huggingface.co/mit-han-lab/opt-13b-smoothquant), each decoder layer contains three major components: attention calculation, layer normalization, and linear transformation in fully connected layers. The corresponding implementation classes for these components are:
@@ -447,21 +447,21 @@ For the target [SQ quantized model](https://huggingface.co/mit-han-lab/opt-13b-s
These classes' underlying implementation logits will harness the functions in previous table. Note that for the example, the `LayerNormQ` module is implemented by the torch native module.
Testing environment:
The hardware platform used for testing equips with 256 AMD EPYC 9534 64-Core Processor, 8 AMD Instinct MI300X accelerators and 1.5T memory. The testing was done in a publicly available Docker image from Docker Hub:
The hardware platform used for testing equips with 256 AMD EPYC 9534 64-Core Processor, 8 AMD Instinct MI300X GPUs and 1.5T memory. The testing was done in a publicly available Docker image from Docker Hub:
[`rocm/pytorch:rocm6.1_ubuntu22.04_py3.10_pytorch_2.1.2`](https://hub.docker.com/layers/rocm/pytorch/rocm6.1_ubuntu22.04_py3.10_pytorch_2.1.2/images/sha256-f6ea7cee8aae299c7f6368187df7beed29928850c3929c81e6f24b34271d652b)
The tested models are OPT-1.3B, 2.7B, 6.7B and 13B FP16 models and the corresponding SmoothQuant INT8 OPT models were obtained from Hugging Face.
Note that since the default values were used for the tunable parameters of the fundamental instance, the performance of the INT8 kernel is suboptimal.
Figure 8 shows the performance comparisons between the original FP16 and the SmoothQuant-quantized INT8 models on a single MI300X accelerator. The GPU memory footprints of SmoothQuant-quantized models are significantly reduced. It also indicates the per-sample inference latency is significantly reduced for all SmoothQuant-quantized OPT models (illustrated in (b)). Notably, the performance of the CK instance-based INT8 kernel steadily improves with an increase in model size.
Figure 8 shows the performance comparisons between the original FP16 and the SmoothQuant-quantized INT8 models on a single MI300X GPU. The GPU memory footprints of SmoothQuant-quantized models are significantly reduced. It also indicates the per-sample inference latency is significantly reduced for all SmoothQuant-quantized OPT models (illustrated in (b)). Notably, the performance of the CK instance-based INT8 kernel steadily improves with an increase in model size.
<!--
================
### Figure 8
================ -->
```{figure} ../../../data/how-to/llm-fine-tuning-optimization/ck-comparisons.jpg
Performance comparisons between the original FP16 and the SmoothQuant-quantized INT8 models on a single MI300X accelerator.
Performance comparisons between the original FP16 and the SmoothQuant-quantized INT8 models on a single MI300X GPU.
```
For accuracy comparisons between the original FP16 and INT8 models, the evaluation is done by using the first 1,000 samples from the LAMBADA dataset's validation set. We employ the same Last Token Prediction Accuracy method introduced in [SmoothQuant Real-INT8 Inference for PyTorch](https://github.com/mit-han-lab/smoothquant/blob/main/examples/smoothquant_opt_real_int8_demo.ipynb) as our evaluation metric. The comparison results are shown in Table 2.
@@ -482,4 +482,4 @@ CK provides a rich set of template parameters for generating flexible accelerate
CK supports multiple instruction sets of AMD Instinct GPUs, operator fusion and different data precisions. Its composability helps users quickly construct operator performance verification.
With CK, you can build more effective AI applications with higher flexibility and better performance on different AMD accelerator platforms.
With CK, you can build more effective AI applications with higher flexibility and better performance on different AMD GPU platforms.

50
docs/how-to/rocm-for-ai/inference-optimization/workload.rst
View File

@@ -1,15 +1,15 @@
.. meta::
:description: Learn about workload tuning on AMD Instinct MI300X accelerators for optimal performance.
:description: Learn about workload tuning on AMD Instinct MI300X GPUs for optimal performance.
:keywords: AMD, Instinct, MI300X, HPC, tuning, BIOS settings, NBIO, ROCm,
environment variable, performance, HIP, Triton, PyTorch TunableOp, vLLM, RCCL,
MIOpen, accelerator, GPU, resource utilization
MIOpen, GPU, resource utilization
*****************************************
AMD Instinct MI300X workload optimization
*****************************************
This document provides guidelines for optimizing the performance of AMD
Instinct™ MI300X accelerators, with a particular focus on GPU kernel
Instinct™ MI300X GPUs, with a particular focus on GPU kernel
programming, high-performance computing (HPC), and deep learning operations
using PyTorch. It delves into specific workloads such as
:ref:`model inference <mi300x-vllm-optimization>`, offering strategies to
@@ -25,7 +25,7 @@ Workload tuning strategy
By following a structured approach, you can systematically address
performance issues and enhance the efficiency of your workloads on AMD Instinct
MI300X accelerators.
MI300X GPUs.
Measure the current workload
----------------------------
@@ -86,7 +86,7 @@ Optimize model inference with vLLM
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
vLLM provides tools and techniques specifically designed for efficient model
inference on AMD Instinct MI300X accelerators. See :ref:`fine-tuning-llms-vllm`
inference on AMD Instinct MI300X GPUs. See :ref:`fine-tuning-llms-vllm`
for installation guidance. Optimizing performance with vLLM
involves configuring tensor parallelism, leveraging advanced features, and
ensuring efficient execution. Here’s how to optimize vLLM performance:
@@ -239,7 +239,7 @@ benchmarking process.
With AMD's profiling tools, developers are able to gain important insight into how efficiently their application is
using hardware resources and effectively diagnose potential bottlenecks contributing to poor performance. Developers
working with AMD Instinct accelerators have multiple tools depending on their specific profiling needs; these include:
working with AMD Instinct GPUs have multiple tools depending on their specific profiling needs; these include:
* :ref:`ROCProfiler <mi300x-rocprof>`
@@ -257,11 +257,11 @@ metrics, commonly called *performance counters*. These counters quantify the per
showcasing which pieces of the computational pipeline and memory hierarchy are being utilized.
Your ROCm installation contains a script or executable command called ``rocprof`` which provides the ability to list all
available hardware counters for your specific accelerator or GPU, and run applications while collecting counters during
available hardware counters for your specific GPU, and run applications while collecting counters during
their execution.
This ``rocprof`` utility also depends on the :doc:`ROCTracer and ROC-TX libraries <roctracer:index>`, giving it the
ability to collect timeline traces of the accelerator software stack as well as user-annotated code regions.
ability to collect timeline traces of the GPU software stack as well as user-annotated code regions.
.. note::
@@ -276,16 +276,16 @@ ROCm Compute Profiler
^^^^^^^^^^^^^^^^^^^^^
:doc:`ROCm Compute Profiler <rocprofiler-compute:index>` is a system performance profiler for high-performance computing (HPC) and
machine learning (ML) workloads using Instinct accelerators. Under the hood, ROCm Compute Profiler uses
machine learning (ML) workloads using Instinct GPUs. Under the hood, ROCm Compute Profiler uses
:ref:`ROCProfiler <mi300x-rocprof>` to collect hardware performance counters. The ROCm Compute Profiler tool performs
system profiling based on all approved hardware counters for Instinct
accelerator architectures. It provides high level performance analysis features including System Speed-of-Light, IP
GPU architectures. It provides high level performance analysis features including System Speed-of-Light, IP
block Speed-of-Light, Memory Chart Analysis, Roofline Analysis, Baseline Comparisons, and more.
ROCm Compute Profiler takes the guesswork out of profiling by removing the need to provide text input files with lists of counters
to collect and analyze raw CSV output files as is the case with ROCProfiler. Instead, ROCm Compute Profiler automates the collection
of all available hardware counters in one command and provides graphical interfaces to help users understand and
analyze bottlenecks and stressors for their computational workloads on AMD Instinct accelerators.
analyze bottlenecks and stressors for their computational workloads on AMD Instinct GPUs.
.. note::
@@ -411,7 +411,7 @@ for additional performance tips. :ref:`fine-tuning-llms-vllm` describes vLLM
usage with ROCm.
ROCm provides a prebuilt optimized Docker image for validating the performance
of LLM inference with vLLM on MI300X series accelerators. The Docker image includes
of LLM inference with vLLM on MI300X Series GPUs. The Docker image includes
ROCm, vLLM, and PyTorch. For more information, see
:doc:`/how-to/rocm-for-ai/inference/benchmark-docker/vllm`.
@@ -449,7 +449,7 @@ Maximizing vLLM instances on a single node
The general guideline is to maximize per-node throughput by running as many vLLM instances as possible.
However, running too many instances might lead to insufficient memory for the KV-cache, which can affect performance.
The Instinct MI300X accelerator is equipped with 192GB of HBM3 memory capacity and bandwidth.
The Instinct MI300X GPU is equipped with 192 GB of HBM3 memory capacity and bandwidth.
For models that fit in one GPU -- to maximize the accumulated throughput -- you can run as many as eight vLLM instances
simultaneously on one MI300X node (with eight GPUs). To do so, use the GPU isolation environment
variable ``CUDA_VISIBLE_DEVICES``.
@@ -468,7 +468,7 @@ The total throughput achieved by running ``N`` instances of vLLM is generally mu
single vLLM instance across ``N`` GPUs simultaneously (that is, configuring ``tensor_parallel_size`` as N or
using the ``-tp`` N option, where ``1 < N ≤ 8``).
vLLM on MI300X accelerators can run a variety of model weights, including Llama 2 (7b, 13b, 70b), Llama 3 (8b, 70b), Qwen2 (7b, 72b), Mixtral-8x7b, Mixtral-8x22b, and so on.
vLLM on MI300X GPUs can run a variety of model weights, including Llama 2 (7b, 13b, 70b), Llama 3 (8b, 70b), Qwen2 (7b, 72b), Mixtral-8x7b, Mixtral-8x22b, and so on.
Notable configurations include Llama2-70b and Llama3-70b models on a single MI300X GPU, and the Llama3.1 405b model can fit on one single node with 8 MI300X GPUs.
.. _mi300x-vllm-gpu-memory-utilization:
@@ -917,7 +917,7 @@ ROCm library tuning involves optimizing the performance of routine computational
operations (such as ``GEMM``) provided by ROCm libraries like
:ref:`hipBLASLt <mi300x-hipblaslt>`, :ref:`Composable Kernel <mi300x-ck>`,
:ref:`MIOpen <mi300x-miopen>`, and :ref:`RCCL <mi300x-rccl>`. This tuning aims
to maximize efficiency and throughput on Instinct MI300X accelerators to gain
to maximize efficiency and throughput on Instinct MI300X GPUs to gain
improved application performance.
.. _mi300x-library-gemm:
@@ -1451,7 +1451,7 @@ you can only use a fraction of the potential bandwidth on the node.
The following figure shows an
:doc:`MI300X node-level architecture </conceptual/gpu-arch/mi300>` of a
system with AMD EPYC processors in a dual-socket configuration and eight
AMD Instinct MI300X accelerators. The MI300X OAMs attach to the host system via
AMD Instinct MI300X GPUs. The MI300X OAMs attach to the host system via
PCIe Gen 5 x16 links (yellow lines). The GPUs use seven high-bandwidth,
low-latency AMD Infinity Fabric™ links (red lines) to form a fully connected
8-GPU system.
@@ -1460,7 +1460,7 @@ low-latency AMD Infinity Fabric™ links (red lines) to form a fully connected
.. figure:: ../../../data/shared/mi300-node-level-arch.png
MI300 series node-level architecture showing 8 fully interconnected MI300X
MI300 Series node-level architecture showing 8 fully interconnected MI300X
OAM modules connected to (optional) PCIe switches via re-timers and HGX
connectors.
@@ -1653,7 +1653,7 @@ Auto-tunable kernel configuration involves adjusting memory access and computati
resources assigned to each compute unit. It encompasses the usage of
:ref:`LDS <mi300x-cu-fig>`, register, and task scheduling on a compute unit.
The accelerator or GPU contains global memory, local data share (LDS), and
The GPU contains global memory, local data share (LDS), and
registers. Global memory has high access latency, but is large. LDS access has
much lower latency, but is smaller. It is a fast on-CU software-managed memory
that can be used to efficiently share data between all work items in a block.
@@ -1666,11 +1666,11 @@ Register access is the fastest yet smallest among the three.
Schematic representation of a CU in the CDNA2 or CDNA3 architecture.
The following is a list of kernel arguments used for tuning performance and
resource allocation on AMD accelerators, which helps in optimizing the
resource allocation on AMD GPUs, which helps in optimizing the
efficiency and throughput of various computational kernels.
``num_stages=n``
Adjusts the number of pipeline stages for different types of kernels. On AMD accelerators, set ``num_stages``
Adjusts the number of pipeline stages for different types of kernels. On AMD GPUs, set ``num_stages``
according to the following rules:
* For kernels with a single GEMM, set to ``2``.
@@ -1697,15 +1697,15 @@ efficiency and throughput of various computational kernels.
* The occupancy of the kernel is limited by VGPR usage, and
* The current VGPR usage is only a few above a boundary in
:ref:`Occupancy related to VGPR usage in an Instinct MI300X accelerator <mi300x-occupancy-vgpr-table>`.
:ref:`Occupancy related to VGPR usage in an Instinct MI300X GPU <mi300x-occupancy-vgpr-table>`.
.. _mi300x-occupancy-vgpr-table:
.. figure:: ../../../data/shared/occupancy-vgpr.png
:alt: Occupancy related to VGPR usage in an Instinct MI300X accelerator.
:alt: Occupancy related to VGPR usage in an Instinct MI300X GPU.
:align: center
Occupancy related to VGPRs usage on an Instinct MI300X accelerator
Occupancy related to VGPRs usage on an Instinct MI300X GPU
For example, according to the table, each Execution Unit (EU) has 512 available
VGPRs, which are allocated in blocks of 16. If the current VGPR usage is 170,
@@ -1730,7 +1730,7 @@ VGPR usage so that it might fit 3 waves per EU.
- ``matrix_instr_nonkdim = 32``: ``mfma_32x32`` is used.
For GEMM kernels on an MI300X accelerator, ``mfma_16x16`` typically outperforms ``mfma_32x32``, even for large
For GEMM kernels on an MI300X GPU, ``mfma_16x16`` typically outperforms ``mfma_32x32``, even for large
tile/GEMM sizes.
@@ -1749,7 +1749,7 @@ the number of CUs a kernel can distribute its task across.
XCD-level system architecture showing 40 compute units,
each with 32 KB L1 cache, a unified compute system with 4 ACE compute
accelerators, shared 4MB of L2 cache, and a hardware scheduler (HWS).
GPUs, shared 4MB of L2 cache, and a hardware scheduler (HWS).
You can query hardware resources with the command ``rocminfo`` in the
``/opt/rocm/bin`` directory. For instance, query the number of CUs, number of

14
docs/how-to/rocm-for-ai/inference/benchmark-docker/previous-versions/vllm-0.10.0-20250812.rst
View File

@@ -1,7 +1,7 @@
:orphan:
.. meta::
:description: Learn how to validate LLM inference performance on MI300X accelerators using AMD MAD and the
:description: Learn how to validate LLM inference performance on MI300X GPUs using AMD MAD and the
ROCm vLLM Docker image.
:keywords: model, MAD, automation, dashboarding, validate
@@ -23,9 +23,9 @@ vLLM inference performance testing
The `ROCm vLLM Docker <{{ unified_docker.docker_hub_url }}>`_ image offers
a prebuilt, optimized environment for validating large language model (LLM)
inference performance on AMD Instinct™ MI300X series accelerators. This ROCm vLLM
Docker image integrates vLLM and PyTorch tailored specifically for MI300X series
accelerators and includes the following components:
inference performance on AMD Instinct™ MI300X Series GPUs. This ROCm vLLM
Docker image integrates vLLM and PyTorch tailored specifically for MI300X Series
GPUs and includes the following components:
.. list-table::
:header-rows: 1
@@ -47,7 +47,7 @@ vLLM inference performance testing
With this Docker image, you can quickly test the :ref:`expected
inference performance numbers <vllm-benchmark-performance-measurements-812>` for
MI300X series accelerators.
MI300X Series GPUs.
What's new
==========
@@ -139,7 +139,7 @@ page provides reference throughput and serving measurements for inferencing popu
The performance data presented in
`Performance results with AMD ROCm software <https://www.amd.com/en/developer/resources/rocm-hub/dev-ai/performance-results.html>`_
only reflects the latest version of this inference benchmarking environment.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X accelerators or ROCm software.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X GPUs or ROCm software.
System validation
=================
@@ -424,7 +424,7 @@ Further reading
- To learn more about MAD and the ``madengine`` CLI, see the `MAD usage guide <https://github.com/ROCm/MAD?tab=readme-ov-file#usage-guide>`__.
- To learn more about system settings and management practices to configure your system for
AMD Instinct MI300X series accelerators, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
AMD Instinct MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
- For application performance optimization strategies for HPC and AI workloads,
including inference with vLLM, see :doc:`/how-to/rocm-for-ai/inference-optimization/workload`.

8
docs/how-to/rocm-for-ai/inference/benchmark-docker/previous-versions/vllm-0.10.1-20250909.rst
View File

@@ -21,8 +21,8 @@ vLLM inference performance testing
The `ROCm vLLM Docker <{{ docker.docker_hub_url }}>`_ image offers
a prebuilt, optimized environment for validating large language model (LLM)
inference performance on AMD Instinct™ MI300X series accelerators. This ROCm vLLM
Docker image integrates vLLM and PyTorch tailored specifically for MI300X series
inference performance on AMD Instinct™ MI300X Series accelerators. This ROCm vLLM
Docker image integrates vLLM and PyTorch tailored specifically for MI300X Series
accelerators and includes the following components:
.. list-table::
@@ -38,7 +38,7 @@ vLLM inference performance testing
With this Docker image, you can quickly test the :ref:`expected
inference performance numbers <vllm-benchmark-performance-measurements-909>` for
MI300X series accelerators.
MI300X Series accelerators.
What's new
==========
@@ -430,7 +430,7 @@ Further reading
- To learn more about MAD and the ``madengine`` CLI, see the `MAD usage guide <https://github.com/ROCm/MAD?tab=readme-ov-file#usage-guide>`__.
- To learn more about system settings and management practices to configure your system for
AMD Instinct MI300X series accelerators, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
AMD Instinct MI300X Series accelerators, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
- See :ref:`fine-tuning-llms-vllm` and :ref:`mi300x-vllm-optimization` for
a brief introduction to vLLM and optimization strategies.

14
docs/how-to/rocm-for-ai/inference/benchmark-docker/previous-versions/vllm-0.4.3.rst
View File

@@ -1,7 +1,7 @@
:orphan:
.. meta::
:description: Learn how to validate LLM inference performance on MI300X accelerators using AMD MAD and the unified
:description: Learn how to validate LLM inference performance on MI300X GPUs using AMD MAD and the unified
ROCm Docker image.
:keywords: model, MAD, automation, dashboarding, validate
@@ -18,9 +18,9 @@ vLLM inference performance testing
The `ROCm vLLM Docker <https://hub.docker.com/r/rocm/vllm/tags>`_ image offers
a prebuilt, optimized environment designed for validating large language model
(LLM) inference performance on the AMD Instinct™ MI300X accelerator. This
(LLM) inference performance on the AMD Instinct™ MI300X GPU. This
ROCm vLLM Docker image integrates vLLM and PyTorch tailored specifically for the
MI300X accelerator and includes the following components:
MI300X GPU and includes the following components:
* `ROCm 6.2.0 <https://github.com/ROCm/ROCm>`_
@@ -31,7 +31,7 @@ MI300X accelerator and includes the following components:
* Tuning files (in CSV format)
With this Docker image, you can quickly validate the expected inference
performance numbers on the MI300X accelerator. This topic also provides tips on
performance numbers on the MI300X GPU. This topic also provides tips on
optimizing performance with popular AI models.
.. _vllm-benchmark-vllm:
@@ -51,7 +51,7 @@ Getting started
===============
Use the following procedures to reproduce the benchmark results on an
MI300X accelerator with the prebuilt vLLM Docker image.
MI300X GPU with the prebuilt vLLM Docker image.
.. _vllm-benchmark-get-started:
@@ -267,7 +267,7 @@ Options
.. _vllm-benchmark-run-benchmark-v043:
Running the benchmark on the MI300X accelerator
Running the benchmark on the MI300X GPU
-----------------------------------------------
Here are some examples of running the benchmark with various options.
@@ -328,7 +328,7 @@ Further reading
see `<https://github.com/ROCm/vllm/tree/main/benchmarks>`_.
- To learn more about system settings and management practices to configure your system for
MI300X series accelerators, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_
MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_
- To learn how to run community models from Hugging Face on AMD GPUs, see
:doc:`Running models from Hugging Face </how-to/rocm-for-ai/inference/hugging-face-models>`.

14
docs/how-to/rocm-for-ai/inference/benchmark-docker/previous-versions/vllm-0.6.4.rst
View File

@@ -1,7 +1,7 @@
:orphan:
.. meta::
:description: Learn how to validate LLM inference performance on MI300X accelerators using AMD MAD and the unified
:description: Learn how to validate LLM inference performance on MI300X GPUs using AMD MAD and the unified
ROCm Docker image.
:keywords: model, MAD, automation, dashboarding, validate
@@ -18,9 +18,9 @@ vLLM inference performance testing
The `ROCm vLLM Docker <https://hub.docker.com/r/rocm/vllm/tags>`_ image offers
a prebuilt, optimized environment designed for validating large language model
(LLM) inference performance on the AMD Instinct™ MI300X accelerator. This
(LLM) inference performance on the AMD Instinct™ MI300X GPU. This
ROCm vLLM Docker image integrates vLLM and PyTorch tailored specifically for the
MI300X accelerator and includes the following components:
MI300X GPU and includes the following components:
* `ROCm 6.2.1 <https://github.com/ROCm/ROCm>`_
@@ -31,7 +31,7 @@ MI300X accelerator and includes the following components:
* Tuning files (in CSV format)
With this Docker image, you can quickly validate the expected inference
performance numbers on the MI300X accelerator. This topic also provides tips on
performance numbers on the MI300X GPU. This topic also provides tips on
optimizing performance with popular AI models.
.. hlist::
@@ -74,7 +74,7 @@ Getting started
===============
Use the following procedures to reproduce the benchmark results on an
MI300X accelerator with the prebuilt vLLM Docker image.
MI300X GPU with the prebuilt vLLM Docker image.
.. _vllm-benchmark-get-started:
@@ -332,7 +332,7 @@ Options
.. _vllm-benchmark-run-benchmark-v064:
Running the benchmark on the MI300X accelerator
Running the benchmark on the MI300X GPU
-----------------------------------------------
Here are some examples of running the benchmark with various options.
@@ -398,7 +398,7 @@ Further reading
see `<https://github.com/ROCm/vllm/tree/main/benchmarks>`_.
- To learn more about system settings and management practices to configure your system for
MI300X series accelerators, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_
MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_
- To learn how to run community models from Hugging Face on AMD GPUs, see
:doc:`Running models from Hugging Face </how-to/rocm-for-ai/inference/hugging-face-models>`.

14
docs/how-to/rocm-for-ai/inference/benchmark-docker/previous-versions/vllm-0.6.6.rst
View File

@@ -1,7 +1,7 @@
:orphan:
.. meta::
:description: Learn how to validate LLM inference performance on MI300X accelerators using AMD MAD and the
:description: Learn how to validate LLM inference performance on MI300X GPUs using AMD MAD and the
ROCm vLLM Docker image.
:keywords: model, MAD, automation, dashboarding, validate
@@ -18,9 +18,9 @@ LLM inference performance validation on AMD Instinct MI300X
The `ROCm vLLM Docker <https://hub.docker.com/r/rocm/vllm/tags>`_ image offers
a prebuilt, optimized environment for validating large language model (LLM)
inference performance on the AMD Instinct™ MI300X accelerator. This ROCm vLLM
inference performance on the AMD Instinct™ MI300X GPU. This ROCm vLLM
Docker image integrates vLLM and PyTorch tailored specifically for the MI300X
accelerator and includes the following components:
GPU and includes the following components:
* `ROCm 6.3.1 <https://github.com/ROCm/ROCm>`_
@@ -29,7 +29,7 @@ accelerator and includes the following components:
* `PyTorch 2.7.0 (2.7.0a0+git3a58512) <https://github.com/pytorch/pytorch>`_
With this Docker image, you can quickly validate the expected inference
performance numbers for the MI300X accelerator. This topic also provides tips on
performance numbers for the MI300X GPU. This topic also provides tips on
optimizing performance with popular AI models. For more information, see the lists of
:ref:`available models for MAD-integrated benchmarking <vllm-benchmark-mad-v066-models>`
and :ref:`standalone benchmarking <vllm-benchmark-standalone-v066-options>`.
@@ -47,7 +47,7 @@ Getting started
===============
Use the following procedures to reproduce the benchmark results on an
MI300X accelerator with the prebuilt vLLM Docker image.
MI300X GPU with the prebuilt vLLM Docker image.
.. _vllm-benchmark-get-started:
@@ -377,7 +377,7 @@ Options and available models
.. _vllm-benchmark-run-benchmark-v066:
Running the benchmark on the MI300X accelerator
Running the benchmark on the MI300X GPU
-----------------------------------------------
Here are some examples of running the benchmark with various options.
@@ -443,7 +443,7 @@ Further reading
see `<https://github.com/ROCm/vllm/tree/main/benchmarks>`_.
- To learn more about system settings and management practices to configure your system for
MI300X series accelerators, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_
MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_
- To learn how to run community models from Hugging Face on AMD GPUs, see
:doc:`Running models from Hugging Face </how-to/rocm-for-ai/inference/hugging-face-models>`.

16
docs/how-to/rocm-for-ai/inference/benchmark-docker/previous-versions/vllm-0.7.3-20250325.rst
View File

@@ -1,7 +1,7 @@
:orphan:
.. meta::
:description: Learn how to validate LLM inference performance on MI300X accelerators using AMD MAD and the
:description: Learn how to validate LLM inference performance on MI300X GPUs using AMD MAD and the
ROCm vLLM Docker image.
:keywords: model, MAD, automation, dashboarding, validate
@@ -23,9 +23,9 @@ vLLM inference performance testing
The `ROCm vLLM Docker <{{ unified_docker.docker_hub_url }}>`_ image offers
a prebuilt, optimized environment for validating large language model (LLM)
inference performance on AMD Instinct™ MI300X series accelerator. This ROCm vLLM
Docker image integrates vLLM and PyTorch tailored specifically for MI300X series
accelerators and includes the following components:
inference performance on AMD Instinct™ MI300X Series GPU. This ROCm vLLM
Docker image integrates vLLM and PyTorch tailored specifically for MI300X Series
GPUs and includes the following components:
* `ROCm {{ unified_docker.rocm_version }} <https://github.com/ROCm/ROCm>`_
@@ -37,7 +37,7 @@ vLLM inference performance testing
With this Docker image, you can quickly test the :ref:`expected
inference performance numbers <vllm-benchmark-performance-measurements-v073>` for
MI300X series accelerators.
MI300X Series GPUs.
.. _vllm-benchmark-available-models-v073:
@@ -110,7 +110,7 @@ vLLM inference performance testing
The performance data presented in
`Performance results with AMD ROCm software <https://www.amd.com/en/developer/resources/rocm-hub/dev-ai/performance-results.html>`_
only reflects the :doc:`latest version of this inference benchmarking environment <../vllm>`.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X accelerators or ROCm software.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X GPUs or ROCm software.
Advanced features and known issues
==================================
@@ -122,7 +122,7 @@ vLLM inference performance testing
===============
Use the following procedures to reproduce the benchmark results on an
MI300X accelerator with the prebuilt vLLM Docker image.
MI300X GPU with the prebuilt vLLM Docker image.
.. _vllm-benchmark-get-started:
@@ -311,7 +311,7 @@ Further reading
see `<https://github.com/ROCm/vllm/tree/main/benchmarks>`_.
- To learn more about system settings and management practices to configure your system for
MI300X series accelerators, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_
MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_
- To learn how to run community models from Hugging Face on AMD GPUs, see
:doc:`Running models from Hugging Face </how-to/rocm-for-ai/inference/hugging-face-models>`.

14
docs/how-to/rocm-for-ai/inference/benchmark-docker/previous-versions/vllm-0.8.3-20250415.rst
View File

@@ -1,7 +1,7 @@
:orphan:
.. meta::
:description: Learn how to validate LLM inference performance on MI300X accelerators using AMD MAD and the
:description: Learn how to validate LLM inference performance on MI300X GPUs using AMD MAD and the
ROCm vLLM Docker image.
:keywords: model, MAD, automation, dashboarding, validate
@@ -18,9 +18,9 @@ vLLM inference performance testing
The `ROCm vLLM Docker <{{ unified_docker.docker_hub_url }}>`_ image offers
a prebuilt, optimized environment for validating large language model (LLM)
inference performance on AMD Instinct™ MI300X series accelerators. This ROCm vLLM
Docker image integrates vLLM and PyTorch tailored specifically for MI300X series
accelerators and includes the following components:
inference performance on AMD Instinct™ MI300X Series GPUs. This ROCm vLLM
Docker image integrates vLLM and PyTorch tailored specifically for MI300X Series
GPUs and includes the following components:
* `ROCm {{ unified_docker.rocm_version }} <https://github.com/ROCm/ROCm>`_
@@ -32,7 +32,7 @@ vLLM inference performance testing
With this Docker image, you can quickly test the :ref:`expected
inference performance numbers <vllm-benchmark-performance-measurements-v083>` for
MI300X series accelerators.
MI300X Series GPUs.
.. _vllm-benchmark-available-models-v083:
@@ -105,7 +105,7 @@ vLLM inference performance testing
The performance data presented in
`Performance results with AMD ROCm software <https://www.amd.com/en/developer/resources/rocm-hub/dev-ai/performance-results.html>`_
only reflects the :doc:`latest version of this inference benchmarking environment <../vllm>`.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X accelerators or ROCm software.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X GPUs or ROCm software.
Advanced features and known issues
==================================
@@ -327,7 +327,7 @@ Further reading
see `<https://github.com/ROCm/vllm/tree/main/benchmarks>`_.
- To learn more about system settings and management practices to configure your system for
MI300X series accelerators, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_
MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_
- To learn how to run community models from Hugging Face on AMD GPUs, see
:doc:`Running models from Hugging Face </how-to/rocm-for-ai/inference/hugging-face-models>`.

14
docs/how-to/rocm-for-ai/inference/benchmark-docker/previous-versions/vllm-0.8.5-20250513.rst
View File

@@ -1,7 +1,7 @@
:orphan:
.. meta::
:description: Learn how to validate LLM inference performance on MI300X accelerators using AMD MAD and the
:description: Learn how to validate LLM inference performance on MI300X GPUs using AMD MAD and the
ROCm vLLM Docker image.
:keywords: model, MAD, automation, dashboarding, validate
@@ -23,9 +23,9 @@ vLLM inference performance testing
The `ROCm vLLM Docker <{{ unified_docker.docker_hub_url }}>`_ image offers
a prebuilt, optimized environment for validating large language model (LLM)
inference performance on AMD Instinct™ MI300X series accelerators. This ROCm vLLM
Docker image integrates vLLM and PyTorch tailored specifically for MI300X series
accelerators and includes the following components:
inference performance on AMD Instinct™ MI300X Series GPUs. This ROCm vLLM
Docker image integrates vLLM and PyTorch tailored specifically for MI300X Series
GPUs and includes the following components:
* `ROCm {{ unified_docker.rocm_version }} <https://github.com/ROCm/ROCm>`_
@@ -37,7 +37,7 @@ vLLM inference performance testing
With this Docker image, you can quickly test the :ref:`expected
inference performance numbers <vllm-benchmark-performance-measurements-v085-20250513>` for
MI300X series accelerators.
MI300X Series GPUs.
.. _vllm-benchmark-available-models-v085-20250513:
@@ -114,7 +114,7 @@ vLLM inference performance testing
The performance data presented in
`Performance results with AMD ROCm software <https://www.amd.com/en/developer/resources/rocm-hub/dev-ai/performance-results.html>`_
only reflects the :doc:`latest version of this inference benchmarking environment <../vllm>`.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X accelerators or ROCm software.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X GPUs or ROCm software.
Advanced features and known issues
==================================
@@ -333,7 +333,7 @@ Further reading
see `<https://github.com/ROCm/vllm/tree/main/benchmarks>`_.
- To learn more about system settings and management practices to configure your system for
MI300X series accelerators, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_
MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_
- For application performance optimization strategies for HPC and AI workloads,
including inference with vLLM, see :doc:`/how-to/rocm-for-ai/inference-optimization/workload`.

14
docs/how-to/rocm-for-ai/inference/benchmark-docker/previous-versions/vllm-0.8.5-20250521.rst
View File

@@ -1,7 +1,7 @@
:orphan:
.. meta::
:description: Learn how to validate LLM inference performance on MI300X accelerators using AMD MAD and the
:description: Learn how to validate LLM inference performance on MI300X GPUs using AMD MAD and the
ROCm vLLM Docker image.
:keywords: model, MAD, automation, dashboarding, validate
@@ -23,9 +23,9 @@ vLLM inference performance testing
The `ROCm vLLM Docker <{{ unified_docker.docker_hub_url }}>`_ image offers
a prebuilt, optimized environment for validating large language model (LLM)
inference performance on AMD Instinct™ MI300X series accelerators. This ROCm vLLM
Docker image integrates vLLM and PyTorch tailored specifically for MI300X series
accelerators and includes the following components:
inference performance on AMD Instinct™ MI300X Series GPUs. This ROCm vLLM
Docker image integrates vLLM and PyTorch tailored specifically for MI300X Series
GPUs and includes the following components:
* `ROCm {{ unified_docker.rocm_version }} <https://github.com/ROCm/ROCm>`_
@@ -37,7 +37,7 @@ vLLM inference performance testing
With this Docker image, you can quickly test the :ref:`expected
inference performance numbers <vllm-benchmark-performance-measurements-v085-20250521>` for
MI300X series accelerators.
MI300X Series GPUs.
.. _vllm-benchmark-available-models-v085-20250521:
@@ -114,7 +114,7 @@ vLLM inference performance testing
The performance data presented in
`Performance results with AMD ROCm software <https://www.amd.com/en/developer/resources/rocm-hub/dev-ai/performance-results.html>`_
should not be interpreted as the peak performance achievable by AMD
Instinct MI325X and MI300X accelerators or ROCm software.
Instinct MI325X and MI300X GPUs or ROCm software.
Advanced features and known issues
==================================
@@ -333,7 +333,7 @@ Further reading
see `<https://github.com/ROCm/vllm/tree/main/benchmarks>`_.
- To learn more about system settings and management practices to configure your system for
MI300X series accelerators, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_
MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_
- For application performance optimization strategies for HPC and AI workloads,
including inference with vLLM, see :doc:`/how-to/rocm-for-ai/inference-optimization/workload`.

14
docs/how-to/rocm-for-ai/inference/benchmark-docker/previous-versions/vllm-0.9.0.1-20250605.rst
View File

@@ -1,7 +1,7 @@
:orphan:
.. meta::
:description: Learn how to validate LLM inference performance on MI300X accelerators using AMD MAD and the
:description: Learn how to validate LLM inference performance on MI300X GPUs using AMD MAD and the
ROCm vLLM Docker image.
:keywords: model, MAD, automation, dashboarding, validate
@@ -23,9 +23,9 @@ vLLM inference performance testing
The `ROCm vLLM Docker <{{ unified_docker.docker_hub_url }}>`_ image offers
a prebuilt, optimized environment for validating large language model (LLM)
inference performance on AMD Instinct™ MI300X series accelerators. This ROCm vLLM
Docker image integrates vLLM and PyTorch tailored specifically for MI300X series
accelerators and includes the following components:
inference performance on AMD Instinct™ MI300X Series GPUs. This ROCm vLLM
Docker image integrates vLLM and PyTorch tailored specifically for MI300X Series
GPUs and includes the following components:
* `ROCm {{ unified_docker.rocm_version }} <https://github.com/ROCm/ROCm>`_
@@ -37,7 +37,7 @@ vLLM inference performance testing
With this Docker image, you can quickly test the :ref:`expected
inference performance numbers <vllm-benchmark-performance-measurements-v0901-20250605>` for
MI300X series accelerators.
MI300X Series GPUs.
.. _vllm-benchmark-available-models-v0901-20250605:
@@ -113,7 +113,7 @@ vLLM inference performance testing
The performance data presented in
`Performance results with AMD ROCm software <https://www.amd.com/en/developer/resources/rocm-hub/dev-ai/performance-results.html>`_
only reflects the latest version of this inference benchmarking environment.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X accelerators or ROCm software.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X GPUs or ROCm software.
Advanced features and known issues
==================================
@@ -332,7 +332,7 @@ Further reading
see `<https://github.com/ROCm/vllm/tree/main/benchmarks>`_.
- To learn more about system settings and management practices to configure your system for
MI300X accelerators, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_
MI300X GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_
- For application performance optimization strategies for HPC and AI workloads,
including inference with vLLM, see :doc:`/how-to/rocm-for-ai/inference-optimization/workload`.

14
docs/how-to/rocm-for-ai/inference/benchmark-docker/previous-versions/vllm-0.9.1-20250702.rst
View File

@@ -1,7 +1,7 @@
:orphan:
.. meta::
:description: Learn how to validate LLM inference performance on MI300X accelerators using AMD MAD and the
:description: Learn how to validate LLM inference performance on MI300X GPUs using AMD MAD and the
ROCm vLLM Docker image.
:keywords: model, MAD, automation, dashboarding, validate
@@ -23,9 +23,9 @@ vLLM inference performance testing
The `ROCm vLLM Docker <{{ unified_docker.docker_hub_url }}>`_ image offers
a prebuilt, optimized environment for validating large language model (LLM)
inference performance on AMD Instinct™ MI300X series accelerators. This ROCm vLLM
Docker image integrates vLLM and PyTorch tailored specifically for MI300X series
accelerators and includes the following components:
inference performance on AMD Instinct™ MI300X Series GPUs. This ROCm vLLM
Docker image integrates vLLM and PyTorch tailored specifically for MI300X Series
GPUs and includes the following components:
* `ROCm {{ unified_docker.rocm_version }} <https://github.com/ROCm/ROCm>`_
@@ -37,7 +37,7 @@ vLLM inference performance testing
With this Docker image, you can quickly test the :ref:`expected
inference performance numbers <vllm-benchmark-performance-measurements-20250702>` for
MI300X series accelerators.
MI300X Series GPUs.
.. _vllm-benchmark-available-models-20250702:
@@ -113,7 +113,7 @@ vLLM inference performance testing
The performance data presented in
`Performance results with AMD ROCm software <https://www.amd.com/en/developer/resources/rocm-hub/dev-ai/performance-results.html>`_
only reflects the latest version of this inference benchmarking environment.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X accelerators or ROCm software.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X GPUs or ROCm software.
Advanced features and known issues
==================================
@@ -332,7 +332,7 @@ Further reading
see `<https://github.com/ROCm/vllm/tree/main/benchmarks>`_.
- To learn more about system settings and management practices to configure your system for
MI300X series accelerators, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_
MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_
- For application performance optimization strategies for HPC and AI workloads,
including inference with vLLM, see :doc:`/how-to/rocm-for-ai/inference-optimization/workload`.

14
docs/how-to/rocm-for-ai/inference/benchmark-docker/previous-versions/vllm-0.9.1-20250715.rst
View File

@@ -1,7 +1,7 @@
:orphan:
.. meta::
:description: Learn how to validate LLM inference performance on MI300X accelerators using AMD MAD and the
:description: Learn how to validate LLM inference performance on MI300X GPUs using AMD MAD and the
ROCm vLLM Docker image.
:keywords: model, MAD, automation, dashboarding, validate
@@ -23,9 +23,9 @@ vLLM inference performance testing
The `ROCm vLLM Docker <{{ unified_docker.docker_hub_url }}>`_ image offers
a prebuilt, optimized environment for validating large language model (LLM)
inference performance on AMD Instinct™ MI300X series accelerators. This ROCm vLLM
Docker image integrates vLLM and PyTorch tailored specifically for MI300X series
accelerators and includes the following components:
inference performance on AMD Instinct™ MI300X Series GPUs. This ROCm vLLM
Docker image integrates vLLM and PyTorch tailored specifically for MI300X Series
GPUs and includes the following components:
.. list-table::
:header-rows: 1
@@ -47,7 +47,7 @@ vLLM inference performance testing
With this Docker image, you can quickly test the :ref:`expected
inference performance numbers <vllm-benchmark-performance-measurements-715>` for
MI300X series accelerators.
MI300X Series GPUs.
What's new
==========
@@ -145,7 +145,7 @@ page provides reference throughput and latency measurements for inferencing popu
The performance data presented in
`Performance results with AMD ROCm software <https://www.amd.com/en/developer/resources/rocm-hub/dev-ai/performance-results.html>`_
only reflects the latest version of this inference benchmarking environment.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X accelerators or ROCm software.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X GPUs or ROCm software.
System validation
=================
@@ -429,7 +429,7 @@ Further reading
- To learn more about MAD and the ``madengine`` CLI, see the `MAD usage guide <https://github.com/ROCm/MAD?tab=readme-ov-file#usage-guide>`__.
- To learn more about system settings and management practices to configure your system for
AMD Instinct MI300X series accelerators, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
AMD Instinct MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
- For application performance optimization strategies for HPC and AI workloads,
including inference with vLLM, see :doc:`/how-to/rocm-for-ai/inference-optimization/workload`.

6
docs/how-to/rocm-for-ai/inference/benchmark-docker/pytorch-inference.rst
View File

@@ -1,5 +1,5 @@
.. meta::
:description: Learn how to validate LLM inference performance on MI300X accelerators using AMD MAD and the
:description: Learn how to validate LLM inference performance on MI300X GPUs using AMD MAD and the
ROCm PyTorch Docker image.
:keywords: model, MAD, automation, dashboarding, validate, pytorch
@@ -15,7 +15,7 @@ PyTorch inference performance testing
{% set model_groups = data.pytorch_inference_benchmark.model_groups %}
The `ROCm PyTorch Docker <https://hub.docker.com/r/rocm/pytorch/tags>`_ image offers a prebuilt,
optimized environment for testing model inference performance on AMD Instinct™ MI300X series
optimized environment for testing model inference performance on AMD Instinct™ MI300X Series
GPUs. This guide demonstrates how to use the AMD Model Automation and Dashboarding (MAD)
tool with the ROCm PyTorch container to test inference performance on various models efficiently.
@@ -175,7 +175,7 @@ Further reading
- To learn more about MAD and the ``madengine`` CLI, see the `MAD usage guide <https://github.com/ROCm/MAD?tab=readme-ov-file#usage-guide>`__.
- To learn more about system settings and management practices to configure your system for
AMD Instinct MI300X series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
AMD Instinct MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
- For application performance optimization strategies for HPC and AI workloads,
including inference with vLLM, see :doc:`../../inference-optimization/workload`.

6
docs/how-to/rocm-for-ai/inference/benchmark-docker/sglang-distributed.rst
View File

@@ -22,7 +22,7 @@ improved efficiency and throughput.
`SGLang <https://docs.sglang.ai>`__ is a high-performance inference and
serving engine for large language models (LLMs) and vision models. The
ROCm-enabled `SGLang base Docker image <{{ docker.docker_hub_url }}>`__
bundles SGLang with PyTorch, which is optimized for AMD Instinct MI300X series
bundles SGLang with PyTorch, which is optimized for AMD Instinct MI300X Series
GPUs. It includes the following software components:
.. list-table::
@@ -37,7 +37,7 @@ improved efficiency and throughput.
{% endfor %}
The following guides on setting up and running SGLang and Mooncake for disaggregated
distributed inference on a Slurm cluster using AMD Instinct MI300X series GPUs backed by
distributed inference on a Slurm cluster using AMD Instinct MI300X Series GPUs backed by
Mellanox CX-7 NICs.
Prerequisites
@@ -236,7 +236,7 @@ Further reading
- See the base upstream Docker image on `Docker Hub <https://hub.docker.com/layers/lmsysorg/sglang/v0.5.2rc1-rocm700-mi30x/images/sha256-10c4ee502ddba44dd8c13325e6e03868bfe7f43d23d0a44780a8ee8b393f4729>`__.
- To learn more about system settings and management practices to configure your system for
MI300X series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`__.
MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`__.
- For application performance optimization strategies for HPC and AI workloads,
including inference with vLLM, see :doc:`/how-to/rocm-for-ai/inference-optimization/workload`.

8
docs/how-to/rocm-for-ai/inference/benchmark-docker/sglang.rst
View File

@@ -1,5 +1,5 @@
.. meta::
:description: Learn how to validate LLM inference performance on MI300X accelerators using AMD MAD and SGLang
:description: Learn how to validate LLM inference performance on MI300X GPUs using AMD MAD and SGLang
:keywords: model, MAD, automation, dashboarding, validate
*****************************************************************
@@ -15,8 +15,8 @@ SGLang inference performance testing DeepSeek-R1-Distill-Qwen-32B
`SGLang <https://docs.sglang.ai>`__ is a high-performance inference and
serving engine for large language models (LLMs) and vision models. The
ROCm-enabled `SGLang Docker image <{{ docker.docker_hub_url }}>`__
bundles SGLang with PyTorch, optimized for AMD Instinct MI300X series
accelerators. It includes the following software components:
bundles SGLang with PyTorch, optimized for AMD Instinct MI300X Series
GPUs. It includes the following software components:
.. list-table::
:header-rows: 1
@@ -255,7 +255,7 @@ Further reading
- To learn more about MAD and the ``madengine`` CLI, see the `MAD usage guide <https://github.com/ROCm/MAD?tab=readme-ov-file#usage-guide>`__.
- To learn more about system settings and management practices to configure your system for
MI300X series accelerators, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`__.
MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`__.
- For application performance optimization strategies for HPC and AI workloads,
including inference with vLLM, see :doc:`/how-to/rocm-for-ai/inference-optimization/workload`.

4
docs/how-to/rocm-for-ai/inference/benchmark-docker/vllm.rst
View File

@@ -1,5 +1,5 @@
.. meta::
:description: Learn how to validate LLM inference performance on MI300X accelerators using AMD MAD and the ROCm vLLM Docker image.
:description: Learn how to validate LLM inference performance on MI300X GPUs using AMD MAD and the ROCm vLLM Docker image.
:keywords: model, MAD, automation, dashboarding, validate
**********************************
@@ -457,7 +457,7 @@ Further reading
- To learn more about MAD and the ``madengine`` CLI, see the `MAD usage guide <https://github.com/ROCm/MAD?tab=readme-ov-file#usage-guide>`__.
- To learn more about system settings and management practices to configure your system for
AMD Instinct MI300X series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
AMD Instinct MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
- See :ref:`fine-tuning-llms-vllm` and :ref:`mi300x-vllm-optimization` for
a brief introduction to vLLM and optimization strategies.

6
docs/how-to/rocm-for-ai/inference/deploy-your-model.rst
View File

@@ -44,9 +44,9 @@ Validating vLLM performance
---------------------------
ROCm provides a prebuilt optimized Docker image for validating the performance of LLM inference with vLLM
on the MI300X accelerator. The Docker image includes ROCm, vLLM, PyTorch, and tuning files in the CSV
on the MI300X GPU. The Docker image includes ROCm, vLLM, PyTorch, and tuning files in the CSV
format. For more information, see the guide to
`LLM inference performance testing with vLLM on the AMD Instinct™ MI300X accelerator <https://github.com/ROCm/MAD/blob/develop/benchmark/vllm/README.md>`_
`LLM inference performance testing with vLLM on the AMD Instinct™ MI300X GPU <https://github.com/ROCm/MAD/blob/develop/benchmark/vllm/README.md>`_
on the ROCm GitHub repository.
.. _rocm-for-ai-serve-hugging-face-tgi:
@@ -61,7 +61,7 @@ The `Hugging Face Text Generation Inference <https://huggingface.co/docs/text-ge
TGI installation
----------------
The easiest way to use Hugging Face TGI with ROCm on AMD Instinct accelerators is to use the official Docker image at
The easiest way to use Hugging Face TGI with ROCm on AMD Instinct GPUs is to use the official Docker image at
`<https://github.com/huggingface/text-generation-inference/pkgs/container/text-generation-inference>`__.
TGI walkthrough

8
docs/how-to/rocm-for-ai/inference/hugging-face-models.rst
View File

@@ -10,7 +10,7 @@ Running models from Hugging Face
transformer models. Hugging Face models and tools significantly enhance productivity, performance, and accessibility in
developing and deploying AI solutions.
This section describes how to run popular community transformer models from Hugging Face on AMD accelerators and GPUs.
This section describes how to run popular community transformer models from Hugging Face on AMD GPUs.
.. _rocm-for-ai-hugging-face-transformers:
@@ -62,11 +62,11 @@ Using Hugging Face with Optimum-AMD
Optimum-AMD is the interface between Hugging Face libraries and the ROCm software stack.
For a deeper dive into using Hugging Face libraries on AMD accelerators and GPUs, refer to the
For a deeper dive into using Hugging Face libraries on AMD GPUs, refer to the
`Optimum-AMD <https://huggingface.co/docs/optimum/main/en/amd/amdgpu/overview>`_ page on Hugging Face for guidance on
using Flash Attention 2, GPTQ quantization and the ONNX Runtime integration.
Hugging Face libraries natively support AMD Instinct accelerators. For other
Hugging Face libraries natively support AMD Instinct GPUs. For other
:doc:`ROCm-capable hardware <rocm-install-on-linux:reference/system-requirements>`, support is currently not
validated, but most features are expected to work without issues.
@@ -139,7 +139,7 @@ To enable `GPTQ <https://arxiv.org/abs/2210.17323>`_, hosted wheels are availabl
pip install auto-gptq --no-build-isolation --extra-index-url https://huggingface.github.io/autogptq-index/whl/rocm573/
Or, to install from source for AMD accelerators supporting ROCm, specify the ``ROCM_VERSION`` environment variable.
Or, to install from source for AMD GPUs supporting ROCm, specify the ``ROCM_VERSION`` environment variable.
.. code-block:: shell

2
docs/how-to/rocm-for-ai/inference/index.rst
View File

@@ -9,7 +9,7 @@ AI inference is a process of deploying a trained machine learning model to make
Understanding the ROCm™ software platform’s architecture and capabilities is vital for running AI inference. By leveraging the ROCm platform's capabilities, you can harness the power of high-performance computing and efficient resource management to run inference workloads, leading to faster predictions and classifications on real-time data.
Throughout the following topics, this section provides a comprehensive guide to setting up and deploying AI inference on AMD GPUs. This includes instructions on how to install ROCm, how to use Hugging Face Transformers to manage pre-trained models for natural language processing (NLP) tasks, how to validate vLLM on AMD Instinct™ MI300X accelerators and illustrate how to deploy trained models in production environments.
Throughout the following topics, this section provides a comprehensive guide to setting up and deploying AI inference on AMD GPUs. This includes instructions on how to install ROCm, how to use Hugging Face Transformers to manage pre-trained models for natural language processing (NLP) tasks, how to validate vLLM on AMD Instinct™ MI300X GPUs and illustrate how to deploy trained models in production environments.
The AI Developer Hub contains `AMD ROCm tutorials <https://rocm.docs.amd.com/projects/ai-developer-hub/en/latest/>`_ for
training, fine-tuning, and inference. It leverages popular machine learning frameworks on AMD GPUs.

10
docs/how-to/rocm-for-ai/inference/llm-inference-frameworks.rst
View File

@@ -60,7 +60,7 @@ Installing vLLM
vllm-rocm \
bash
3. Inside the container, start the API server to run on a single accelerator on port 8000 using the following command.
3. Inside the container, start the API server to run on a single GPU on port 8000 using the following command.
.. code-block:: shell
@@ -113,7 +113,7 @@ Installing vLLM
python -m vllm.entrypoints.api_server --model /app/model --dtype float16 -tp 2 --port 8000 &
4. To run multiple instances of API Servers, specify different ports for each server, and use ``ROCR_VISIBLE_DEVICES`` to
isolate each instance to a different accelerator.
isolate each instance to a different GPU.
For example, to run two API servers, one on port 8000 using GPU 0 and 1, one on port 8001 using GPU 2 and 3, use a
a command like the following.
@@ -140,7 +140,7 @@ Installing vLLM
See :ref:`mi300x-vllm-optimization` for performance optimization tips.
ROCm provides a prebuilt optimized Docker image for validating the performance of LLM inference with vLLM
on the MI300X accelerator. The Docker image includes ROCm, vLLM, and PyTorch.
on the MI300X GPU. The Docker image includes ROCm, vLLM, and PyTorch.
For more information, see :doc:`/how-to/rocm-for-ai/inference/benchmark-docker/vllm`.
.. _fine-tuning-llms-tgi:
@@ -178,7 +178,7 @@ Install TGI
.. tab-item:: TGI on a single-accelerator system
:sync: single
2. Inside the container, launch a model using TGI server on a single accelerator.
2. Inside the container, launch a model using TGI server on a single GPU.
.. code-block:: shell
@@ -199,7 +199,7 @@ Install TGI
.. tab-item:: TGI on a multi-accelerator system
2. Inside the container, launch a model using TGI server on multiple accelerators (4 in this case).
2. Inside the container, launch a model using TGI server on multiple GPUs (four in this case).
.. code-block:: shell

10
docs/how-to/rocm-for-ai/system-setup/multi-node-setup.rst
View File

@@ -1,6 +1,6 @@
.. meta::
:description: Multi-node setup for AI training
:keywords: gpu, accelerator, system, health, validation, bench, perf, performance, rvs, rccl, babel, mi300x, mi325x, flops, bandwidth, rbt, training
:keywords: gpu, system, health, validation, bench, perf, performance, rvs, rccl, babel, mi300x, mi325x, flops, bandwidth, rbt, training
.. _rocm-for-ai-multi-node-setup:
@@ -21,7 +21,7 @@ Before starting, ensure your environment meets the following requirements:
* Multi-node networking: your cluster should have a configured multi-node network. For setup
instructions, see the `Multi-node network configuration for AMD Instinct
accelerators
GPUs
<https://instinct.docs.amd.com/projects/gpu-cluster-networking/en/latest/how-to/multi-node-config.html>`__
guide in the Instinct documentation.
@@ -54,8 +54,8 @@ Compile and install the RoCE library
If you're using Broadcom NICs, you need to compile and install the RoCE (RDMA
over Converged Ethernet) library. See `RoCE cluster network configuration guide
for AMD Instinct accelerators
<https://instinct.docs.amd.com/projects/gpu-cluster-networking/en/latest/how-to/roce-network-config.html#roce-cluster-network-configuration-guide-for-amd-instinct-accelerators>`__
for AMD Instinct GPUs
<https://instinct.docs.amd.com/projects/gpu-cluster-networking/en/latest/how-to/roce-network-config.html>`__
for more information.
See the `Ethernet networking guide for AMD
@@ -315,6 +315,6 @@ Megatron-LM
Further reading
===============
* `Multi-node network configuration for AMD Instinct accelerators <https://instinct.docs.amd.com/projects/gpu-cluster-networking/en/latest/how-to/multi-node-config.html>`__
* `Multi-node network configuration for AMD Instinct GPUs <https://instinct.docs.amd.com/projects/gpu-cluster-networking/en/latest/how-to/multi-node-config.html>`__
* `Ethernet networking guide for AMD Instinct MI300X GPU clusters: Compiling Broadcom NIC software from source <https://docs.broadcom.com/doc/957608-AN2XX#page=81>`__

6
docs/how-to/rocm-for-ai/training/benchmark-docker/jax-maxtext.rst
View File

@@ -10,7 +10,7 @@ MaxText is a high-performance, open-source framework built on the Google JAX
machine learning library to train LLMs at scale. The MaxText framework for
ROCm is an optimized fork of the upstream
`<https://github.com/AI-Hypercomputer/maxtext>`__ enabling efficient AI workloads
on AMD MI300X series GPUs.
on AMD MI300X Series GPUs.
The MaxText for ROCm training Docker image
provides a prebuilt environment for training on AMD Instinct MI300X and MI325X GPUs,
@@ -69,7 +69,7 @@ Supported models
================
The following models are pre-optimized for performance on AMD Instinct MI300
series GPUs. Some instructions, commands, and available training
Series GPUs. Some instructions, commands, and available training
configurations in this documentation might vary by model -- select one to get
started.
@@ -347,7 +347,7 @@ Further reading
- To learn more about MAD and the ``madengine`` CLI, see the `MAD usage guide <https://github.com/ROCm/MAD?tab=readme-ov-file#usage-guide>`__.
- To learn more about system settings and management practices to configure your system for
AMD Instinct MI300X series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
AMD Instinct MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
- For a list of other ready-made Docker images for AI with ROCm, see
`AMD Infinity Hub <https://www.amd.com/en/developer/resources/infinity-hub.html#f-amd_hub_category=AI%20%26%20ML%20Models>`_.

10
docs/how-to/rocm-for-ai/training/benchmark-docker/megatron-lm.rst
View File

@@ -17,13 +17,13 @@ Training a model with Megatron-LM on ROCm
The `Megatron-LM framework for ROCm <https://github.com/ROCm/Megatron-LM>`_ is
a specialized fork of the robust Megatron-LM, designed to enable efficient
training of large-scale language models on AMD GPUs. By leveraging AMD
Instinct™ MI300X series GPUs, Megatron-LM delivers enhanced
Instinct™ MI300X Series GPUs, Megatron-LM delivers enhanced
scalability, performance, and resource utilization for AI workloads. It is
purpose-built to support models like Llama, DeepSeek, and Mixtral,
enabling developers to train next-generation AI models more
efficiently.
AMD provides ready-to-use Docker images for MI300X series GPUs containing
AMD provides ready-to-use Docker images for MI300X Series GPUs containing
essential components, including PyTorch, ROCm libraries, and Megatron-LM
utilities. It contains the following software components to accelerate training
workloads:
@@ -61,7 +61,7 @@ workloads:
================
The following models are supported for training performance benchmarking with Megatron-LM and ROCm
on AMD Instinct MI300X series GPUs.
on AMD Instinct MI300X Series GPUs.
Some instructions, commands, and training recommendations in this documentation might
vary by model -- select one to get started.
@@ -138,7 +138,7 @@ Environment setup
=================
Use the following instructions to set up the environment, configure the script to train models, and
reproduce the benchmark results on MI300X series GPUs with the AMD Megatron-LM Docker
reproduce the benchmark results on MI300X Series GPUs with the AMD Megatron-LM Docker
image.
.. _amd-megatron-lm-requirements:
@@ -533,7 +533,7 @@ Run training
Use the following example commands to set up the environment, configure
:ref:`key options <amd-megatron-lm-benchmark-test-vars>`, and run training on
MI300X series GPUs with the AMD Megatron-LM environment.
MI300X Series GPUs with the AMD Megatron-LM environment.
Single node training
--------------------

4
docs/how-to/rocm-for-ai/training/benchmark-docker/mpt-llm-foundry.rst
View File

@@ -16,7 +16,7 @@ environment for the MPT-30B model using the ``rocm/pytorch-training:v25.5``
base `Docker image <https://hub.docker.com/layers/rocm/pytorch-training/v25.5/images/sha256-d47850a9b25b4a7151f796a8d24d55ea17bba545573f0d50d54d3852f96ecde5>`_
and the `LLM Foundry <https://github.com/mosaicml/llm-foundry>`_ framework.
This environment packages the following software components to train
on AMD Instinct MI300X series accelerators:
on AMD Instinct MI300X Series GPUs:
+--------------------------+--------------------------------+
| Software component | Version |
@@ -182,7 +182,7 @@ Further reading
- To learn more about MAD and the ``madengine`` CLI, see the `MAD usage guide <https://github.com/ROCm/MAD?tab=readme-ov-file#usage-guide>`__.
- To learn more about system settings and management practices to configure your system for
AMD Instinct MI300X series accelerators, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
AMD Instinct MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
- For a list of other ready-made Docker images for AI with ROCm, see
`AMD Infinity Hub <https://www.amd.com/en/developer/resources/infinity-hub.html#f-amd_hub_category=AI%20%26%20ML%20Models>`_.

6
docs/how-to/rocm-for-ai/training/benchmark-docker/previous-versions/jax-maxtext-v25.4.rst
View File

@@ -17,10 +17,10 @@ MaxText is a high-performance, open-source framework built on the Google JAX
machine learning library to train LLMs at scale. The MaxText framework for
ROCm is an optimized fork of the upstream
`<https://github.com/AI-Hypercomputer/maxtext>`__ enabling efficient AI workloads
on AMD MI300X series accelerators.
on AMD MI300X Series GPUs.
The MaxText for ROCm training Docker (``rocm/jax-training:maxtext-v25.4``) image
provides a prebuilt environment for training on AMD Instinct MI300X and MI325X accelerators,
provides a prebuilt environment for training on AMD Instinct MI300X and MI325X GPUs,
including essential components like JAX, XLA, ROCm libraries, and MaxText utilities.
It includes the following software components:
@@ -53,7 +53,7 @@ MaxText provides the following key features to train large language models effic
.. _amd-maxtext-model-support-v254:
The following models are pre-optimized for performance on AMD Instinct MI300X series accelerators.
The following models are pre-optimized for performance on AMD Instinct MI300X Series GPUs.
* Llama 3.1 8B

6
docs/how-to/rocm-for-ai/training/benchmark-docker/previous-versions/jax-maxtext-v25.5.rst
View File

@@ -17,10 +17,10 @@ MaxText is a high-performance, open-source framework built on the Google JAX
machine learning library to train LLMs at scale. The MaxText framework for
ROCm is an optimized fork of the upstream
`<https://github.com/AI-Hypercomputer/maxtext>`__ enabling efficient AI workloads
on AMD MI300X series accelerators.
on AMD MI300X Series GPUs.
The MaxText for ROCm training Docker (``rocm/jax-training:maxtext-v25.5``) image
provides a prebuilt environment for training on AMD Instinct MI300X and MI325X accelerators,
provides a prebuilt environment for training on AMD Instinct MI300X and MI325X GPUs,
including essential components like JAX, XLA, ROCm libraries, and MaxText utilities.
It includes the following software components:
@@ -53,7 +53,7 @@ MaxText provides the following key features to train large language models effic
.. _amd-maxtext-model-support-v255:
The following models are pre-optimized for performance on AMD Instinct MI300X series accelerators.
The following models are pre-optimized for performance on AMD Instinct MI300X Series GPUs.
* Llama 3.3 70B

10
docs/how-to/rocm-for-ai/training/benchmark-docker/previous-versions/megatron-lm-v24.12-dev.rst
View File

@@ -17,12 +17,12 @@ Training a model with ROCm Megatron-LM
The ROCm Megatron-LM framework is a specialized fork of the robust Megatron-LM, designed to
enable efficient training of large-scale language models on AMD GPUs. By leveraging AMD Instinct™ MI300X
accelerators, AMD Megatron-LM delivers enhanced scalability, performance, and resource utilization for AI
GPUs, AMD Megatron-LM delivers enhanced scalability, performance, and resource utilization for AI
workloads. It is purpose-built to :ref:`support models <amd-megatron-lm-model-support-24-12>`
like Meta's Llama 2, Llama 3, and Llama 3.1, enabling developers to train next-generation AI models with greater
efficiency. See the GitHub repository at `<https://github.com/ROCm/Megatron-LM>`__.
For ease of use, AMD provides a ready-to-use Docker image for MI300X accelerators containing essential
For ease of use, AMD provides a ready-to-use Docker image for MI300X GPUs containing essential
components, including PyTorch, PyTorch Lightning, ROCm libraries, and Megatron-LM utilities. It contains the
following software to accelerate training workloads:
@@ -69,7 +69,7 @@ Megatron-LM provides the following key features to train large language models e
.. _amd-megatron-lm-model-support-24-12:
The following models are pre-optimized for performance on the AMD Instinct MI300X accelerator.
The following models are pre-optimized for performance on the AMD Instinct MI300X GPU.
* Llama 2 7B
@@ -208,14 +208,14 @@ Use the following script to run the RCCL test for four MI300X GPU nodes. Modify
.. _mi300x-amd-megatron-lm-training-v2412:
Start training on MI300X accelerators
Start training on MI300X GPUs
=====================================
The pre-built ROCm Megatron-LM environment allows users to quickly validate system performance, conduct
training benchmarks, and achieve superior performance for models like Llama 2 and Llama 3.1.
Use the following instructions to set up the environment, configure the script to train models, and
reproduce the benchmark results on the MI300X accelerators with the AMD Megatron-LM Docker
reproduce the benchmark results on the MI300X GPUs with the AMD Megatron-LM Docker
image.
.. _amd-megatron-lm-requirements-v2412:

14
docs/how-to/rocm-for-ai/training/benchmark-docker/previous-versions/megatron-lm-v25.3.rst
View File

@@ -15,13 +15,13 @@ Training a model with Megatron-LM for ROCm
The Megatron-LM framework for ROCm is a specialized fork of the robust Megatron-LM,
designed to enable efficient training of large-scale language models on AMD
GPUs. By leveraging AMD Instinct™ MI300X series accelerators, Megatron-LM delivers
GPUs. By leveraging AMD Instinct™ MI300X Series GPUs, Megatron-LM delivers
enhanced scalability, performance, and resource utilization for AI workloads.
It is purpose-built to support models like Llama 2, Llama 3, Llama 3.1, and
DeepSeek, enabling developers to train next-generation AI models more
efficiently. See the GitHub repository at `<https://github.com/ROCm/Megatron-LM>`__.
AMD provides a ready-to-use Docker image for MI300X accelerators containing
AMD provides a ready-to-use Docker image for MI300X GPUs containing
essential components, including PyTorch, ROCm libraries, and Megatron-LM
utilities. It contains the following software components to accelerate training
workloads:
@@ -69,7 +69,7 @@ Megatron-LM provides the following key features to train large language models e
.. _amd-megatron-lm-model-support-25-3:
The following models are pre-optimized for performance on the AMD Instinct MI300X accelerator.
The following models are pre-optimized for performance on the AMD Instinct MI300X GPU.
* Llama 2 7B
@@ -123,7 +123,7 @@ The pre-built ROCm Megatron-LM environment allows users to quickly validate syst
training benchmarks, and achieve superior performance for models like Llama 3.1, Llama 2, and DeepSeek V2.
Use the following instructions to set up the environment, configure the script to train models, and
reproduce the benchmark results on the MI300X accelerators with the AMD Megatron-LM Docker
reproduce the benchmark results on the MI300X GPUs with the AMD Megatron-LM Docker
image.
.. _amd-megatron-lm-requirements-v253:
@@ -334,7 +334,7 @@ Multi-node training
inside a Docker, either install the drivers inside the Docker container or pass the network
drivers from the host while creating the Docker container.
Start training on AMD Instinct accelerators
Start training on AMD Instinct GPUs
===========================================
The prebuilt Megatron-LM with ROCm training environment allows users to quickly validate
@@ -345,8 +345,8 @@ can expect the container to perform in the model configurations described in
the following section, but other configurations are not validated by AMD.
Use the following instructions to set up the environment, configure the script
to train models, and reproduce the benchmark results on MI300X series
accelerators with the AMD Megatron-LM Docker image.
to train models, and reproduce the benchmark results on MI300X Series
GPUs with the AMD Megatron-LM Docker image.
.. tab-set::

16
docs/how-to/rocm-for-ai/training/benchmark-docker/previous-versions/megatron-lm-v25.4.rst
View File

@@ -15,13 +15,13 @@ Training a model with Megatron-LM for ROCm
The Megatron-LM framework for ROCm is a specialized fork of the robust Megatron-LM,
designed to enable efficient training of large-scale language models on AMD
GPUs. By leveraging AMD Instinct™ MI300X series accelerators, Megatron-LM delivers
GPUs. By leveraging AMD Instinct™ MI300X Series GPUs, Megatron-LM delivers
enhanced scalability, performance, and resource utilization for AI workloads.
It is purpose-built to support models like Llama 2, Llama 3, Llama 3.1, and
DeepSeek, enabling developers to train next-generation AI models more
efficiently. See the GitHub repository at `<https://github.com/ROCm/Megatron-LM>`__.
AMD provides a ready-to-use Docker image for MI300X series accelerators containing
AMD provides a ready-to-use Docker image for MI300X Series GPUs containing
essential components, including PyTorch, ROCm libraries, and Megatron-LM
utilities. It contains the following software components to accelerate training
workloads:
@@ -69,7 +69,7 @@ Megatron-LM provides the following key features to train large language models e
.. _amd-megatron-lm-model-support-25-4:
The following models are pre-optimized for performance on AMD Instinct MI300X series accelerators.
The following models are pre-optimized for performance on AMD Instinct MI300X Series GPUs.
* Llama 3.1 8B
@@ -105,7 +105,7 @@ popular AI models.
The performance data presented in
`Performance results with AMD ROCm software <https://www.amd.com/en/developer/resources/rocm-hub/dev-ai/performance-results.html>`__
only reflects the :doc:`latest version of this training benchmarking environment <../megatron-lm>`.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X accelerators or ROCm software.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X GPUs or ROCm software.
System validation
=================
@@ -124,7 +124,7 @@ The prebuilt ROCm Megatron-LM environment allows users to quickly validate syste
training benchmarks, and achieve superior performance for models like Llama 3.1, Llama 2, and DeepSeek V2.
Use the following instructions to set up the environment, configure the script to train models, and
reproduce the benchmark results on MI300X series accelerators with the AMD Megatron-LM Docker
reproduce the benchmark results on MI300X Series GPUs with the AMD Megatron-LM Docker
image.
.. _amd-megatron-lm-requirements-v254:
@@ -367,7 +367,7 @@ Multi-node training
# Specify which RDMA interfaces to use for communication
export NCCL_IB_HCA=rdma0,rdma1,rdma2,rdma3,rdma4,rdma5,rdma6,rdma7
Start training on AMD Instinct accelerators
Start training on AMD Instinct GPUs
===========================================
The prebuilt Megatron-LM with ROCm training environment allows users to quickly validate
@@ -378,8 +378,8 @@ can expect the container to perform in the model configurations described in
the following section, but other configurations are not validated by AMD.
Use the following instructions to set up the environment, configure the script
to train models, and reproduce the benchmark results on MI300X series
accelerators with the AMD Megatron-LM Docker image.
to train models, and reproduce the benchmark results on MI300X Series
GPUs with the AMD Megatron-LM Docker image.
.. tab-set::

12
docs/how-to/rocm-for-ai/training/benchmark-docker/previous-versions/megatron-lm-v25.5.rst
View File

@@ -16,13 +16,13 @@ Training a model with Megatron-LM for ROCm
The `Megatron-LM framework for ROCm <https://github.com/ROCm/Megatron-LM>`_ is
a specialized fork of the robust Megatron-LM, designed to enable efficient
training of large-scale language models on AMD GPUs. By leveraging AMD
Instinct™ MI300X series accelerators, Megatron-LM delivers enhanced
Instinct™ MI300X Series GPUs, Megatron-LM delivers enhanced
scalability, performance, and resource utilization for AI workloads. It is
purpose-built to support models like Llama, DeepSeek, and Mixtral,
enabling developers to train next-generation AI models more
efficiently.
AMD provides a ready-to-use Docker image for MI300X series accelerators containing
AMD provides a ready-to-use Docker image for MI300X Series GPUs containing
essential components, including PyTorch, ROCm libraries, and Megatron-LM
utilities. It contains the following software components to accelerate training
workloads:
@@ -69,7 +69,7 @@ Megatron-LM provides the following key features to train large language models e
.. _amd-megatron-lm-model-support-v255:
The following models are pre-optimized for performance on AMD Instinct MI300X series accelerators.
The following models are pre-optimized for performance on AMD Instinct MI300X Series GPUs.
.. datatemplate:yaml:: /data/how-to/rocm-for-ai/training/previous-versions/megatron-lm-v25.5-benchmark-models.yaml
@@ -131,7 +131,7 @@ popular AI models.
The performance data presented in
`Performance results with AMD ROCm software <https://www.amd.com/en/developer/resources/rocm-hub/dev-ai/performance-results.html>`__
only reflects the latest version of this training benchmarking environment.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X accelerators or ROCm software.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X GPUs or ROCm software.
System validation
=================
@@ -154,7 +154,7 @@ Environment setup
=================
Use the following instructions to set up the environment, configure the script to train models, and
reproduce the benchmark results on MI300X series accelerators with the AMD Megatron-LM Docker
reproduce the benchmark results on MI300X Series GPUs with the AMD Megatron-LM Docker
image.
.. _amd-megatron-lm-requirements-v255:
@@ -536,7 +536,7 @@ Run training
Use the following example commands to set up the environment, configure
:ref:`key options <amd-megatron-lm-benchmark-test-vars-v255>`, and run training on
MI300X series accelerators with the AMD Megatron-LM environment.
MI300X Series GPUs with the AMD Megatron-LM environment.
Single node training
^^^^^^^^^^^^^^^^^^^^

12
docs/how-to/rocm-for-ai/training/benchmark-docker/previous-versions/megatron-lm-v25.6.rst
View File

@@ -16,13 +16,13 @@ Training a model with Megatron-LM for ROCm
The `Megatron-LM framework for ROCm <https://github.com/ROCm/Megatron-LM>`__ is
a specialized fork of the robust Megatron-LM, designed to enable efficient
training of large-scale language models on AMD GPUs. By leveraging AMD
Instinct™ MI300X series accelerators, Megatron-LM delivers enhanced
Instinct™ MI300X Series GPUs, Megatron-LM delivers enhanced
scalability, performance, and resource utilization for AI workloads. It is
purpose-built to support models like Llama, DeepSeek, and Mixtral,
enabling developers to train next-generation AI models more
efficiently.
AMD provides ready-to-use Docker images for MI300X series accelerators containing
AMD provides ready-to-use Docker images for MI300X Series GPUs containing
essential components, including PyTorch, ROCm libraries, and Megatron-LM
utilities. It contains the following software components to accelerate training
workloads:
@@ -65,7 +65,7 @@ workloads:
.. _amd-megatron-lm-model-support-v256:
The following models are pre-optimized for performance on AMD Instinct MI300X series accelerators.
The following models are pre-optimized for performance on AMD Instinct MI300X Series GPUs.
Supported models
================
@@ -124,7 +124,7 @@ popular AI models.
The performance data presented in
`Performance results with AMD ROCm software <https://www.amd.com/en/developer/resources/rocm-hub/dev-ai/performance-results.html>`__
only reflects the latest version of this training benchmarking environment.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X accelerators or ROCm software.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X GPUs or ROCm software.
System validation
=================
@@ -147,7 +147,7 @@ Environment setup
=================
Use the following instructions to set up the environment, configure the script to train models, and
reproduce the benchmark results on MI300X series accelerators with the AMD Megatron-LM Docker
reproduce the benchmark results on MI300X Series GPUs with the AMD Megatron-LM Docker
image.
.. _amd-megatron-lm-requirements-v256:
@@ -589,7 +589,7 @@ Run training
Use the following example commands to set up the environment, configure
:ref:`key options <amd-megatron-lm-benchmark-test-vars-v256>`, and run training on
MI300X series accelerators with the AMD Megatron-LM environment.
MI300X Series GPUs with the AMD Megatron-LM environment.
Single node training
--------------------

12
docs/how-to/rocm-for-ai/training/benchmark-docker/previous-versions/megatron-lm-v25.7.rst
View File

@@ -22,13 +22,13 @@ Training a model with Megatron-LM for ROCm
The `Megatron-LM framework for ROCm <https://github.com/ROCm/Megatron-LM>`_ is
a specialized fork of the robust Megatron-LM, designed to enable efficient
training of large-scale language models on AMD GPUs. By leveraging AMD
Instinct™ MI300X series accelerators, Megatron-LM delivers enhanced
Instinct™ MI300X Series GPUs, Megatron-LM delivers enhanced
scalability, performance, and resource utilization for AI workloads. It is
purpose-built to support models like Llama, DeepSeek, and Mixtral,
enabling developers to train next-generation AI models more
efficiently.
AMD provides ready-to-use Docker images for MI300X series accelerators containing
AMD provides ready-to-use Docker images for MI300X Series GPUs containing
essential components, including PyTorch, ROCm libraries, and Megatron-LM
utilities. It contains the following software components to accelerate training
workloads:
@@ -66,7 +66,7 @@ workloads:
================
The following models are supported for training performance benchmarking with Megatron-LM and ROCm
on AMD Instinct MI300X series accelerators.
on AMD Instinct MI300X Series GPUs.
Some instructions, commands, and training recommendations in this documentation might
vary by model -- select one to get started.
@@ -120,7 +120,7 @@ popular AI models.
The performance data presented in
`Performance results with AMD ROCm software <https://www.amd.com/en/developer/resources/rocm-hub/dev-ai/performance-results.html>`__
only reflects the latest version of this training benchmarking environment.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X accelerators or ROCm software.
The listed measurements should not be interpreted as the peak performance achievable by AMD Instinct MI325X and MI300X GPUs or ROCm software.
System validation
=================
@@ -143,7 +143,7 @@ Environment setup
=================
Use the following instructions to set up the environment, configure the script to train models, and
reproduce the benchmark results on MI300X series accelerators with the AMD Megatron-LM Docker
reproduce the benchmark results on MI300X Series GPUs with the AMD Megatron-LM Docker
image.
.. _amd-megatron-lm-requirements-v257:
@@ -592,7 +592,7 @@ Run training
Use the following example commands to set up the environment, configure
:ref:`key options <amd-megatron-lm-benchmark-test-vars-v257>`, and run training on
MI300X series accelerators with the AMD Megatron-LM environment.
MI300X Series GPUs with the AMD Megatron-LM environment.
Single node training
--------------------

10
docs/how-to/rocm-for-ai/training/benchmark-docker/previous-versions/primus-megatron-v25.7.rst
View File

@@ -15,7 +15,7 @@ Training a model with Primus and Megatron-LM
`Primus <https://github.com/AMD-AGI/Primus>`__ is a unified and flexible
LLM training framework designed to streamline training. It streamlines LLM
training on AMD Instinct accelerators using a modular, reproducible configuration paradigm.
training on AMD Instinct GPUs using a modular, reproducible configuration paradigm.
Primus is backend-agnostic and supports multiple training engines -- including Megatron.
.. note::
@@ -25,7 +25,7 @@ Primus is backend-agnostic and supports multiple training engines -- including M
workloads from Megatron-LM to Primus with Megatron, see
:doc:`megatron-lm-primus-migration-guide`.
For ease of use, AMD provides a ready-to-use Docker image for MI300 series accelerators
For ease of use, AMD provides a ready-to-use Docker image for MI300 Series GPUs
containing essential components for Primus and Megatron-LM.
.. note::
@@ -53,7 +53,7 @@ containing essential components for Primus and Megatron-LM.
Supported models
================
The following models are pre-optimized for performance on AMD Instinct MI300X series accelerators.
The following models are pre-optimized for performance on AMD Instinct MI300X Series GPUs.
Some instructions, commands, and training examples in this documentation might
vary by model -- select one to get started.
@@ -120,7 +120,7 @@ system's configuration.
=================
Use the following instructions to set up the environment, configure the script to train models, and
reproduce the benchmark results on MI300X series accelerators with the ``{{ docker.pull_tag }}`` image.
reproduce the benchmark results on MI300X Series GPUs with the ``{{ docker.pull_tag }}`` image.
.. _amd-primus-megatron-lm-requirements-v257:
@@ -231,7 +231,7 @@ Run training
Use the following example commands to set up the environment, configure
:ref:`key options <amd-primus-megatron-lm-benchmark-test-vars>`, and run training on
MI300X series accelerators with the AMD Megatron-LM environment.
MI300X Series GPUs with the AMD Megatron-LM environment.
Single node training
--------------------

10
docs/how-to/rocm-for-ai/training/benchmark-docker/previous-versions/pytorch-training-v25.3.rst
View File

@@ -18,7 +18,7 @@ model training with GPU-optimized components for transformer-based models.
The PyTorch for ROCm training Docker (``rocm/pytorch-training:v25.3``) image
provides a prebuilt optimized environment for fine-tuning and pretraining a
model on AMD Instinct MI325X and MI300X accelerators. It includes the following
model on AMD Instinct MI325X and MI300X GPUs. It includes the following
software components to accelerate training workloads:
+--------------------------+--------------------------------+
@@ -44,7 +44,7 @@ software components to accelerate training workloads:
Supported models
================
The following models are pre-optimized for performance on the AMD Instinct MI300X accelerator.
The following models are pre-optimized for performance on the AMD Instinct MI300X GPU.
* Llama 3.1 8B
@@ -237,7 +237,7 @@ Along with the following datasets:
* `bghira/pseudo-camera-10k <https://huggingface.co/datasets/bghira/pseudo-camera-10k>`_
Start training on AMD Instinct accelerators
Start training on AMD Instinct GPUs
===========================================
The prebuilt PyTorch with ROCm training environment allows users to quickly validate
@@ -248,8 +248,8 @@ can expect the container to perform in the model configurations described in
the following section, but other configurations are not validated by AMD.
Use the following instructions to set up the environment, configure the script
to train models, and reproduce the benchmark results on MI300X series
accelerators with the AMD PyTorch training Docker image.
to train models, and reproduce the benchmark results on MI300X Series
GPUs with the AMD PyTorch training Docker image.
Once your environment is set up, use the following commands and examples to start benchmarking.

8
docs/how-to/rocm-for-ai/training/benchmark-docker/previous-versions/pytorch-training-v25.4.rst
View File

@@ -18,7 +18,7 @@ model training with GPU-optimized components for transformer-based models.
The PyTorch for ROCm training Docker (``rocm/pytorch-training:v25.4``) image
provides a prebuilt optimized environment for fine-tuning and pretraining a
model on AMD Instinct MI325X and MI300X accelerators. It includes the following
model on AMD Instinct MI325X and MI300X GPUs. It includes the following
software components to accelerate training workloads:
+--------------------------+--------------------------------+
@@ -44,7 +44,7 @@ software components to accelerate training workloads:
Supported models
================
The following models are pre-optimized for performance on the AMD Instinct MI325X and MI300X accelerators.
The following models are pre-optimized for performance on the AMD Instinct MI325X and MI300X GPUs.
* Llama 3.1 8B
@@ -76,7 +76,7 @@ popular AI models.
The performance data presented in
`Performance results with AMD ROCm software <https://www.amd.com/en/developer/resources/rocm-hub/dev-ai/performance-results.html#tabs-a8deaeb413-item-21cea50186-tab>`_
should not be interpreted as the peak performance achievable by AMD
Instinct MI325X and MI300X accelerators or ROCm software.
Instinct MI325X and MI300X GPUs or ROCm software.
System validation
=================
@@ -260,7 +260,7 @@ the following section, but other configurations are not validated by AMD.
Use the following instructions to set up the environment, configure the script
to train models, and reproduce the benchmark results on MI325X and MI300X
accelerators with the AMD PyTorch training Docker image.
GPUs with the AMD PyTorch training Docker image.
Once your environment is set up, use the following commands and examples to start benchmarking.

6
docs/how-to/rocm-for-ai/training/benchmark-docker/previous-versions/pytorch-training-v25.5.rst
View File

@@ -19,7 +19,7 @@ model training with GPU-optimized components for transformer-based models.
The `PyTorch for ROCm training Docker <https://hub.docker.com/layers/rocm/pytorch-training/v25.5/images/sha256-d47850a9b25b4a7151f796a8d24d55ea17bba545573f0d50d54d3852f96ecde5>`_
(``rocm/pytorch-training:v25.5``) image
provides a prebuilt optimized environment for fine-tuning and pretraining a
model on AMD Instinct MI325X and MI300X accelerators. It includes the following
model on AMD Instinct MI325X and MI300X GPUs. It includes the following
software components to accelerate training workloads:
+--------------------------+--------------------------------+
@@ -45,7 +45,7 @@ software components to accelerate training workloads:
Supported models
================
The following models are pre-optimized for performance on the AMD Instinct MI325X and MI300X accelerators.
The following models are pre-optimized for performance on the AMD Instinct MI325X and MI300X GPUs.
* Llama 3.3 70B
@@ -79,7 +79,7 @@ popular AI models.
The performance data presented in
`Performance results with AMD ROCm software <https://www.amd.com/en/developer/resources/rocm-hub/dev-ai/performance-results.html#tabs-a8deaeb413-item-21cea50186-tab>`_
should not be interpreted as the peak performance achievable by AMD
Instinct MI325X and MI300X accelerators or ROCm software.
Instinct MI325X and MI300X GPUs or ROCm software.
System validation
=================

8
docs/how-to/rocm-for-ai/training/benchmark-docker/previous-versions/pytorch-training-v25.6.rst
View File

@@ -18,7 +18,7 @@ model training with GPU-optimized components for transformer-based models.
The `PyTorch for ROCm training Docker <https://hub.docker.com/layers/rocm/pytorch-training/v25.6/images/sha256-a4cea3c493a4a03d199a3e81960ac071d79a4a7a391aa9866add3b30a7842661>`_
(``rocm/pytorch-training:v25.6``) image provides a prebuilt optimized environment for fine-tuning and pretraining a
model on AMD Instinct MI325X and MI300X accelerators. It includes the following software components to accelerate
model on AMD Instinct MI325X and MI300X GPUs. It includes the following software components to accelerate
training workloads:
+--------------------------+--------------------------------+
@@ -44,7 +44,7 @@ training workloads:
Supported models
================
The following models are pre-optimized for performance on the AMD Instinct MI325X and MI300X accelerators.
The following models are pre-optimized for performance on the AMD Instinct MI325X and MI300X GPUs.
.. datatemplate:yaml:: /data/how-to/rocm-for-ai/training/previous-versions/pytorch-training-v25.6-benchmark-models.yaml
@@ -99,7 +99,7 @@ The following models are pre-optimized for performance on the AMD Instinct MI325
The performance data presented in
`Performance results with AMD ROCm software <https://www.amd.com/en/developer/resources/rocm-hub/dev-ai/performance-results.html#tabs-a8deaeb413-item-21cea50186-tab>`_
should not be interpreted as the peak performance achievable by AMD
Instinct MI325X and MI300X accelerators or ROCm software.
Instinct MI325X and MI300X GPUs or ROCm software.
System validation
=================
@@ -444,7 +444,7 @@ Further reading
- To learn more about MAD and the ``madengine`` CLI, see the `MAD usage guide <https://github.com/ROCm/MAD?tab=readme-ov-file#usage-guide>`__.
- To learn more about system settings and management practices to configure your system for
AMD Instinct MI300X series accelerators, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
AMD Instinct MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
- For a list of other ready-made Docker images for AI with ROCm, see
`AMD Infinity Hub <https://www.amd.com/en/developer/resources/infinity-hub.html#f-amd_hub_category=AI%20%26%20ML%20Models>`_.

8
docs/how-to/rocm-for-ai/training/benchmark-docker/previous-versions/pytorch-training-v25.7.rst
View File

@@ -22,7 +22,7 @@ model training with GPU-optimized components for transformer-based models.
{% set docker = dockers[0] %}
The `PyTorch for ROCm training Docker <{{ docker.docker_hub_url }}>`__
(``{{ docker.pull_tag }}``) image provides a prebuilt optimized environment for fine-tuning and pretraining a
model on AMD Instinct MI325X and MI300X accelerators. It includes the following software components to accelerate
model on AMD Instinct MI325X and MI300X GPUs. It includes the following software components to accelerate
training workloads:
.. list-table::
@@ -41,7 +41,7 @@ model training with GPU-optimized components for transformer-based models.
Supported models
================
The following models are pre-optimized for performance on the AMD Instinct MI325X and MI300X accelerators.
The following models are pre-optimized for performance on the AMD Instinct MI325X and MI300X GPUs.
Some instructions, commands, and training recommendations in this documentation might
vary by model -- select one to get started.
@@ -124,7 +124,7 @@ popular AI models.
The performance data presented in
`Performance results with AMD ROCm software <https://www.amd.com/en/developer/resources/rocm-hub/dev-ai/performance-results.html#tabs-a8deaeb413-item-21cea50186-tab>`_
should not be interpreted as the peak performance achievable by AMD
Instinct MI325X and MI300X accelerators or ROCm software.
Instinct MI325X and MI300X GPUs or ROCm software.
System validation
=================
@@ -555,7 +555,7 @@ Further reading
- To learn more about MAD and the ``madengine`` CLI, see the `MAD usage guide <https://github.com/ROCm/MAD?tab=readme-ov-file#usage-guide>`__.
- To learn more about system settings and management practices to configure your system for
AMD Instinct MI300X series accelerators, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
AMD Instinct MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
- For a list of other ready-made Docker images for AI with ROCm, see
`AMD Infinity Hub <https://www.amd.com/en/developer/resources/infinity-hub.html#f-amd_hub_category=AI%20%26%20ML%20Models>`_.

10
docs/how-to/rocm-for-ai/training/benchmark-docker/primus-megatron.rst
View File

@@ -17,7 +17,7 @@ Primus is backend-agnostic and supports multiple training engines -- including M
To learn how to migrate workloads from Megatron-LM to Primus with Megatron,
see :doc:`previous-versions/megatron-lm-primus-migration-guide`.
For ease of use, AMD provides a ready-to-use Docker image for MI300 series GPUs
For ease of use, AMD provides a ready-to-use Docker image for MI300 Series GPUs
containing essential components for Primus and Megatron-LM. This Docker is powered by Primus
Turbo optimizations for performance; this release adds support for Primus Turbo
with optimized attention and grouped GEMM kernels.
@@ -47,7 +47,7 @@ with optimized attention and grouped GEMM kernels.
Supported models
================
The following models are pre-optimized for performance on AMD Instinct MI300X series GPUs.
The following models are pre-optimized for performance on AMD Instinct MI300X Series GPUs.
Some instructions, commands, and training examples in this documentation might
vary by model -- select one to get started.
@@ -114,7 +114,7 @@ Environment setup
{% set docker = dockers[0] %}
Use the following instructions to set up the environment, configure the script to train models, and
reproduce the benchmark results on MI300X series GPUs with the ``{{ docker.pull_tag }}`` image.
reproduce the benchmark results on MI300X Series GPUs with the ``{{ docker.pull_tag }}`` image.
.. _amd-primus-megatron-lm-requirements:
@@ -234,7 +234,7 @@ Run training
Use the following example commands to set up the environment, configure
:ref:`key options <amd-primus-megatron-lm-benchmark-test-vars>`, and run training on
MI300X series GPUs with the AMD Megatron-LM environment.
MI300X Series GPUs with the AMD Megatron-LM environment.
Single node training
--------------------
@@ -649,7 +649,7 @@ Further reading
Framework for Large Models on AMD GPUs <https://rocm.blogs.amd.com/software-tools-optimization/primus/README.html>`__.
- To learn more about system settings and management practices to configure your system for
AMD Instinct MI300X series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
AMD Instinct MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
- For a list of other ready-made Docker images for AI with ROCm, see
`AMD Infinity Hub <https://www.amd.com/en/developer/resources/infinity-hub.html#f-amd_hub_category=AI%20%26%20ML%20Models>`_.

4
docs/how-to/rocm-for-ai/training/benchmark-docker/primus-pytorch.rst
View File

@@ -21,7 +21,7 @@ Primus now supports the PyTorch torchtitan backend.
{% set dockers = data.dockers %}
{% set docker = dockers[0] %}
For ease of use, AMD provides a ready-to-use Docker image -- ``{{
docker.pull_tag }}`` -- for MI300X series GPUs containing essential
docker.pull_tag }}`` -- for MI300X Series GPUs containing essential
components for Primus and PyTorch training with
Primus Turbo optimizations.
@@ -296,7 +296,7 @@ Further reading
- To learn more about MAD and the ``madengine`` CLI, see the `MAD usage guide <https://github.com/ROCm/MAD?tab=readme-ov-file#usage-guide>`__.
- To learn more about system settings and management practices to configure your system for
AMD Instinct MI300X series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
AMD Instinct MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
- For a list of other ready-made Docker images for AI with ROCm, see
`AMD Infinity Hub <https://www.amd.com/en/developer/resources/infinity-hub.html#f-amd_hub_category=AI%20%26%20ML%20Models>`_.

2
docs/how-to/rocm-for-ai/training/benchmark-docker/pytorch-training.rst
View File

@@ -576,7 +576,7 @@ Further reading
- To learn more about MAD and the ``madengine`` CLI, see the `MAD usage guide <https://github.com/ROCm/MAD?tab=readme-ov-file#usage-guide>`__.
- To learn more about system settings and management practices to configure your system for
AMD Instinct MI300X series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
AMD Instinct MI300X Series GPUs, see `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_.
- For a list of other ready-made Docker images for AI with ROCm, see
`AMD Infinity Hub <https://www.amd.com/en/developer/resources/infinity-hub.html#f-amd_hub_category=AI%20%26%20ML%20Models>`_.

18
docs/how-to/rocm-for-ai/training/scale-model-training.rst
View File

@@ -6,9 +6,9 @@
Scaling model training
**********************
To train a large-scale model like OpenAI GPT-2 or Meta Llama 2 70B, a single accelerator or GPU cannot store all the
model parameters required for training. This immense scale presents a fundamental challenge: no single GPU or
accelerator can simultaneously store and process the entire model's parameters during training. PyTorch
To train a large-scale model like OpenAI GPT-2 or Meta Llama 2 70B, a single GPU cannot store all the
model parameters required for training. This immense scale presents a fundamental challenge: no single GPU
can simultaneously store and process the entire model's parameters during training. PyTorch
provides an answer to this computational constraint through its distributed training frameworks.
.. _rocm-for-ai-pytorch-distributed:
@@ -26,9 +26,9 @@ Features in ``torch.distributed`` are categorized into three main components:
- `Collective communication <https://pytorch.org/docs/stable/distributed.html>`_
In this topic, the focus is on the distributed data-parallelism strategy as it’s the most popular. To get started with DDP,
you need to first understand how to coordinate the model and its training data across multiple accelerators or GPUs.
you need to first understand how to coordinate the model and its training data across multiple GPUs.
The DDP workflow on multiple accelerators or GPUs is as follows:
The DDP workflow on multiple GPUs is as follows:
#. Split the current global training batch into small local batches on each GPU. For instance, if you have 8 GPUs and
the global batch is set at 32 samples, each of the 8 GPUs will have a local batch size of 4 samples.
@@ -46,7 +46,7 @@ In DDP training, each process or worker owns a replica of the model and processe
See the following developer blogs for more in-depth explanations and examples.
* `Multi GPU training with DDP — PyTorch Tutorials <https://pytorch.org/tutorials/beginner/ddp_series_multigpu.html>`_
* `Multi GPU training with DDP — PyTorch Tutorials <https://pytorch.org/tutorials/beginner/ddp_Series_multigpu.html>`_
* `Building a decoder transformer model on AMD GPUs — ROCm Blogs
<https://rocm.blogs.amd.com/artificial-intelligence/decoder-transformer/README.html#distributed-training-on-multiple-gpus>`_
@@ -82,7 +82,7 @@ the training pillar.
See `Pre-training a large language model with Megatron-DeepSpeed on multiple AMD GPUs
<https://rocm.blogs.amd.com/artificial-intelligence/megatron-deepspeed-pretrain/README.html>`_ for a detailed example of
training with DeepSpeed on an AMD accelerator or GPU.
training with DeepSpeed on an AMD GPU.
.. _rocm-for-ai-automatic-mixed-precision:
@@ -95,7 +95,7 @@ can take to reduce training time and memory usage through `automatic mixed preci
See `Automatic mixed precision in PyTorch using AMD GPUs — ROCm Blogs
<https://rocm.blogs.amd.com/artificial-intelligence/automatic-mixed-precision/README.html#automatic-mixed-precision-in-pytorch-using-amd-gpus>`_
for more information about running AMP on an AMD accelerator.
for more information about running AMP on an AMD Instinct-Series GPU.
.. _rocm-for-ai-fine-tune:
@@ -107,7 +107,7 @@ example, LoRA, QLoRA, PEFT, and FSDP.
Learn more about challenges and solutions for model fine-tuning in :doc:`../fine-tuning/index`.
The following developer blogs showcase examples of fine-tuning a model on an AMD accelerator or GPU.
The following developer blogs showcase examples of fine-tuning a model on an AMD GPU.
* Fine-tuning Llama2 with LoRA

10
docs/how-to/rocm-for-hpc/index.rst
View File

@@ -7,14 +7,14 @@ Using ROCm for HPC
******************
The ROCm open-source software stack is optimized to extract high-performance
computing (HPC) workload performance from AMD Instinct™ accelerators
computing (HPC) workload performance from AMD Instinct™ GPUs
while maintaining compatibility with industry software frameworks.
ROCm enhances support and access for developers by providing streamlined and
improved tools that significantly increase productivity. Being open-source, ROCm
fosters innovation, differentiation, and collaboration within the developer
community, making it a powerful and accessible solution for leveraging the full
potential of AMD accelerators' capabilities in diverse computational
potential of AMD GPUs' capabilities in diverse computational
applications.
* For more information, see :doc:`What is ROCm? <../../what-is-rocm>`.
@@ -24,7 +24,7 @@ applications.
and operating system support.
Some of the most popular HPC frameworks are part of the ROCm platform, including
those to help parallelize operations across multiple accelerators and servers,
those to help parallelize operations across multiple GPUs and servers,
handle memory hierarchies, and solve linear systems.
.. image:: ../../data/how-to/rocm-for-hpc/hpc-stack-2024_6_20.png
@@ -38,7 +38,7 @@ science, genomics, geophysics, molecular dynamics, and physics computing.
Refer to the resources in the following table for instructions on building,
running, and deploying these applications on ROCm-capable systems with AMD
Instinct accelerators. Each build container provides parameters to specify
Instinct GPUs. Each build container provides parameters to specify
different source code branches, release versions of ROCm, OpenMPI, UCX, and
Ubuntu versions.
@@ -96,7 +96,7 @@ Ubuntu versions.
* -
- `QUDA <https://github.com/amd/InfinityHub-CI/tree/main/quda>`_
- Library designed for efficient lattice QCD computations on
accelerators. It includes optimized Dirac operators and a variety of
GPUs. It includes optimized Dirac operators and a variety of
fermion solvers and conjugate gradient (CG) implementations, enhancing
performance and accuracy in lattice QCD simulations.

2
docs/how-to/system-optimization/index.rst
View File

@@ -1,6 +1,6 @@
.. meta::
:description: Learn about AMD hardware optimization for HPC-specific and workstation workloads.
:keywords: high-performance computing, HPC, Instinct accelerators, Radeon,
:keywords: high-performance computing, HPC, Instinct GPUs, Radeon,
tuning, tuning guide, AMD, ROCm
*******************

6
docs/how-to/tuning-guides/mi300x/index.rst
View File

@@ -1,7 +1,7 @@
:orphan:
.. meta::
:description: How to configure MI300X accelerators to fully leverage their capabilities and achieve optimal performance.
:description: How to configure MI300X GPUs to fully leverage their capabilities and achieve optimal performance.
:keywords: ROCm, AI, machine learning, MI300X, LLM, usage, tutorial, optimization, tuning
************************
@@ -10,9 +10,9 @@ AMD MI300X tuning guides
The tuning guides in this section provide a comprehensive summary of the
necessary steps to properly configure your system for AMD Instinct™ MI300X
accelerators. They include detailed instructions on system settings and
GPUs. They include detailed instructions on system settings and
application tuning suggestions to help you fully leverage the capabilities of
these accelerators, thereby achieving optimal performance.
these GPUs, thereby achieving optimal performance.
* :doc:`/how-to/rocm-for-ai/inference-optimization/workload`
* `AMD Instinct MI300X system optimization <https://instinct.docs.amd.com/projects/amdgpu-docs/en/latest/system-optimization/mi300x.html>`_

4
docs/index.md
View File

@@ -8,7 +8,7 @@ myst:
# AMD ROCm documentation
ROCm is an open-source software platform optimized to extract HPC and AI workload
performance from AMD Instinct accelerators and AMD Radeon GPUs while maintaining
performance from AMD Instinct GPUs and AMD Radeon GPUs while maintaining
compatibility with industry software frameworks. For more information, see
[What is ROCm?](./what-is-rocm.rst)
@@ -64,7 +64,7 @@ ROCm documentation is organized into the following categories:
<!-- markdownlint-disable MD051 -->
* [ROCm libraries](./reference/api-libraries.md)
* [ROCm tools, compilers, and runtimes](./reference/rocm-tools.md)
* [Accelerator and GPU hardware specifications](./reference/gpu-arch-specs.rst)
* [GPU hardware specifications](./reference/gpu-arch-specs.rst)
* [Data types and precision support](./reference/precision-support.rst)
* [Graph safe support](./reference/graph-safe-support.rst)
<!-- markdownlint-enable MD051 -->

8
docs/reference/gpu-arch-specs.rst
View File

@@ -1,17 +1,17 @@
.. meta::
:description: AMD Instinct™ accelerator, AMD Radeon PRO™, and AMD Radeon™ GPU architecture information
:description: AMD Instinct™ GPU, AMD Radeon PRO™, and AMD Radeon™ GPU architecture information
:keywords: Instinct, Radeon, accelerator, GCN, CDNA, RDNA, GPU, architecture, VRAM, Compute Units, Cache, Registers, LDS, Register File
Accelerator and GPU hardware specifications
GPU hardware specifications
===========================================
The following tables provide an overview of the hardware specifications for AMD Instinct™ accelerators, and AMD Radeon™ PRO and Radeon™ GPUs.
The following tables provide an overview of the hardware specifications for AMD Instinct™ GPUs, and AMD Radeon™ PRO and Radeon™ GPUs.
For more information about ROCm hardware compatibility, see the ROCm `Compatibility matrix <https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html>`_.
.. tab-set::
.. tab-item:: AMD Instinct accelerators
.. tab-item:: AMD Instinct GPUs
.. list-table::
:header-rows: 1

28
docs/reference/gpu-atomics-operation.rst
View File

@@ -1,5 +1,5 @@
.. meta::
:description: AMD Instinct accelerator, AMD Radeon PRO, and AMD Radeon GPU
:description: AMD Instinct GPU, AMD Radeon PRO, and AMD Radeon GPU
atomics operations information
:keywords: Atomics operations, atomic bitwise functions, atomics add, atomics
subtraction, atomics exchange, atomics min, atomics max
@@ -15,8 +15,8 @@ access to the same memory location could lead to incorrect or undefined
behavior.
This topic summarizes the support of atomic read-modify-write
(atomicRMW) operations on AMD GPUs and accelerators. This includes gfx9, gfx10,
gfx11, and gfx12 targets and the following series of Instinct™ series:
(atomicRMW) operations on AMD GPUs. This includes gfx9, gfx10,
gfx11, and gfx12 targets and the following Instinct™ Series:
- MI100
@@ -79,10 +79,10 @@ Scopes of operations:
Support summary
================================================================================
AMD Instinct accelerators
AMD Instinct GPUs
--------------------------------------------------------------------------------
**MI300 and MI350 series**
**MI300 and MI350 Series**
- All atomicRMW operations are forwarded out to the Infinity Fabric.
- Infinity Fabric supports common integer and bitwise atomics, FP32 atomic add,
@@ -95,7 +95,7 @@ AMD Instinct accelerators
It will seem like atomics to the wave, but the CPU sees it as a non-atomic
load-op-store sequence. This downgrades system-scope atomics to device-scope.
**MI200 series**
**MI200 Series**
- L2 cache and Infinity Fabric both support common integer and bitwise atomics.
- L2 cache supports FP32 atomic add, packed-FP16 atomic add, and FP64 add,
@@ -272,10 +272,10 @@ The tables selectors or options are the following:
- Second-level option:
- "No PCIe atomics" means the system does not support PCIe atomics between
the accelerator and peer/host-memory.
the GPU and peer/host-memory.
- "PCIe atomics" means the system supports PCIe atomics between the
accelerator and peer/host-memory.
GPU and peer/host-memory.
- The third-level option is the memory granularity of the memory target.
@@ -306,11 +306,11 @@ The integer type atomic operations that are supported by different hardware.
- Max
AMD Instinct accelerators
AMD Instinct GPUs
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
The integer type atomic operations that are supported by different AMD
Instinct accelerators listed in the following table.
Instinct GPUs listed in the following table.
.. <!-- spellcheck-disable -->
@@ -481,11 +481,11 @@ The bitwise atomic operations that are supported by different hardware.
128-bit bitwise Exchange and CAS are not supported on AMD GPUs
AMD Instinct accelerators
AMD Instinct GPUs
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
The bitwise atomic operations that are supported by different AMD Instinct
accelerators listed in the following table.
GPUs listed in the following table.
.. <!-- spellcheck-disable -->
@@ -659,11 +659,11 @@ The float types atomic operations that are supported by different hardware.
- Add
AMD Instinct accelerators
AMD Instinct GPUs
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
The float type atomic operations that are supported by different AMD Instinct
accelerators listed in the following table.
GPUs listed in the following table.
.. <!-- spellcheck-disable -->

2
docs/reference/precision-support.rst
View File

@@ -9,7 +9,7 @@
Data types and precision support
*************************************************************
This topic summarizes the data types supported on AMD GPUs and accelerators and
This topic summarizes the data types supported on AMD GPUs and
ROCm libraries, along with corresponding :doc:`HIP <hip:index>` data types.
Integral types

6
docs/sphinx/_toc.yml.in
View File

@@ -97,9 +97,9 @@ subtrees:
subtrees:
- entries:
- file: how-to/rocm-for-ai/fine-tuning/single-gpu-fine-tuning-and-inference.rst
title: Use a single accelerator
title: Use a single GPU
- file: how-to/rocm-for-ai/fine-tuning/multi-gpu-fine-tuning-and-inference.rst
title: Use multiple accelerators
title: Use multiple GPUs
- file: how-to/rocm-for-ai/inference/index.rst
title: Inference
@@ -180,7 +180,7 @@ subtrees:
- file: conceptual/gpu-arch/mi300-mi200-performance-counters.rst
title: MI300 and MI200 performance counters
- file: conceptual/gpu-arch/mi350-performance-counters.rst
title: MI350 series performance counters
title: MI350 Series performance counters
- file: conceptual/gpu-arch/mi250.md
title: MI250 microarchitecture
subtrees:

2
tools/autotag/templates/extra_components/6.1.2.md
View File

@@ -46,4 +46,4 @@ ROCm SMI for ROCm 6.1.2
#### Fixes
* Fixed an issue causing ROCm SMI to incorrectly report GPU utilization for RDNA3 GPUs. See the issue on [GitHub](https://github.com/ROCm/ROCm/issues/3112).
* Fixed the parsing of `pp_od_clk_voltage` in `get_od_clk_volt_info` to work better with MI-series hardware.
* Fixed the parsing of `pp_od_clk_voltage` in `get_od_clk_volt_info` to work better with MI-Series hardware.

4
tools/autotag/templates/highlights/5.3.2.md
View File

@@ -33,10 +33,10 @@ Until a fix is provided, users should rely on ROCm v5.2.3 to support their SRIOV
#### AMD Instinct™ MI200 firmware updates
Customers cannot update the Integrated Firmware Image (IFWI) for AMD Instinct™ MI200 accelerators.
Customers cannot update the Integrated Firmware Image (IFWI) for AMD Instinct™ MI200 GPUs.
An updated firmware maintenance bundle consisting of an installation tool and images specific to
AMD Instinct™ MI200 accelerators is under planning and will be available soon.
AMD Instinct™ MI200 GPUs is under planning and will be available soon.
#### Known issue with rocThrust and rocPRIM libraries

6
tools/autotag/templates/highlights/5.4.1.md
View File

@@ -50,12 +50,12 @@ fixed in this release.
#### AMD Instinct™ MI200 firmware IFWI maintenance update #3
This IFWI release fixes the following issue in AMD Instinct™ MI210/MI250 Accelerators.
This IFWI release fixes the following issue in AMD Instinct™ MI210/MI250 GPUs.
After prolonged periods of operation, certain MI200 Instinct™ Accelerators may perform in a degraded
After prolonged periods of operation, certain MI200 Instinct™ GPUs may perform in a degraded
way resulting in application failures.
In this package, AMD delivers a new firmware version for MI200 GPU accelerators and a firmware
In this package, AMD delivers a new firmware version for MI200 GPU GPUs and a firmware
installation tool – AMD FW FLASH 1.2.
| GPU | Productionp part number | SKU | IFWI name |

4
tools/autotag/templates/highlights/5.7.0.md
View File

@@ -10,10 +10,10 @@ New features include:
* AddressSanitizer for host and device code (GPU) is now available as a beta
Note that ROCm 5.7.0 is EOS for MI50. 5.7 versions of ROCm are the last major releases in the ROCm 5
series. This release is Linux-only.
Series. This release is Linux-only.
:::{important}
The next major ROCm release (ROCm 6.0) will not be backward compatible with the ROCm 5 series.
The next major ROCm release (ROCm 6.0) will not be backward compatible with the ROCm 5 Series.
Changes will include: splitting LLVM packages into more manageable sizes, changes to the HIP runtime
API, splitting rocRAND and hipRAND into separate packages, and reorganizing our file structure.
:::

4
tools/autotag/templates/highlights/6.0.0.md
View File

@@ -3,7 +3,7 @@
ROCm 6.0 is a major release with new performance optimizations, expanded frameworks and library
support, and improved developer experience. This includes initial enablement of the AMD Instinct™
MI300 series. Future releases will further enable and optimize this new platform. Key features include:
MI300 Series. Future releases will further enable and optimize this new platform. Key features include:
* Improved performance in areas like lower precision math and attention layers.
* New hipSPARSELt library to accelerate AI workloads via AMD's sparse matrix core technique.
@@ -18,7 +18,7 @@ the [Changelog](https://rocm.docs.amd.com/en/docs-6.0.0/about/CHANGELOG.html).
### OS and GPU support changes
AMD Instinct™ MI300A and MI300X Accelerator support has been enabled for limited operating
AMD Instinct™ MI300A and MI300X GPU support has been enabled for limited operating
systems.
* Ubuntu 22.04.3 (MI300A and MI300X)

2
tools/autotag/templates/highlights/6.1.1.md
View File

@@ -3,7 +3,7 @@ ROCm™ 6.1.1 introduces minor fixes and improvements to some tools and librarie
### OS support
* ROCm 6.1.1 now supports Oracle Linux. It has been tested against version 8.9 (kernel 5.15.0-205) with AMD Instinct MI300X accelerators.
* ROCm 6.1.1 now supports Oracle Linux. It has been tested against version 8.9 (kernel 5.15.0-205) with AMD Instinct MI300X GPUs.
* ROCm 6.1.1 has been tested against a pre-release version of Ubuntu 22.04.5 (kernel: 5.15 [GA], 6.8 [HWE]).

10
tools/autotag/templates/highlights/6.2.0.md
View File

@@ -28,7 +28,7 @@ This section introduces notable new features and improvements in ROCm 6.2. See t
ROCm 6.2.0 introduces the following new components to the ROCm software stack.
- **Omniperf** -- A kernel-level profiling tool for machine learning and high-performance computing (HPC) workloads
running on AMD Instinct accelerators. Omniperf offers comprehensive profiling and advanced analysis via command line
running on AMD Instinct GPUs. Omniperf offers comprehensive profiling and advanced analysis via command line
or a GUI dashboard. For more information, see
[Omniperf](https://rocm.docs.amd.com/projects/omniperf/en/latest).
@@ -141,7 +141,7 @@ For more information, see [Model quantization techniques](https://rocm.docs.amd.
#### Improved vLLM support
ROCm 6.2.0 enhances vLLM support for inference on AMD Instinct accelerators, adding
ROCm 6.2.0 enhances vLLM support for inference on AMD Instinct GPUs, adding
capabilities for `FP16`/`BF16` precision for LLMs, and `FP8` support for Llama.
ROCm 6.2.0 adds support for the following vLLM features:
@@ -177,12 +177,12 @@ To enable these experimental new features, see
Use the `rocm/vllm` branch when cloning the GitHub repo. The `vllm/ROCm_performance.md` document outlines
all the accessible features, and the `vllm/Dockerfile.rocm` file can be used.
### Enhanced performance tuning on AMD Instinct accelerators
### Enhanced performance tuning on AMD Instinct GPUs
ROCm is pre-tuned for high-performance computing workloads including large language models, generative AI, and scientific computing.
The ROCm documentation provides comprehensive guidance on configuring your system for AMD Instinct accelerators. It includes
The ROCm documentation provides comprehensive guidance on configuring your system for AMD Instinct GPUs. It includes
detailed instructions on system settings and application tuning suggestions to help you fully leverage the capabilities of these
accelerators for optimal performance. For more information, see
GPUs for optimal performance. For more information, see
[AMD MI300X tuning guides](https://rocm.docs.amd.com/en/docs-6.2.0/how-to/tuning-guides/mi300x/index.html) and
[AMD MI300A system optimization](https://rocm.docs.amd.com/en/docs-6.2.0/how-to/system-optimization/mi300x.html).

2
tools/autotag/templates/highlights/6.2.2.md
View File

@@ -22,7 +22,7 @@ The following is a significant fix introduced in ROCm 6.2.2.
### Fixed Instinct MI300X error recovery failure
Improved the reliability of AMD Instinct MI300X accelerators in scenarios involving
Improved the reliability of AMD Instinct MI300X GPUs in scenarios involving
uncorrectable errors. Previously, error recovery did not occur as expected,
potentially leaving the system in an undefined state. This fix ensures that error
recovery functions as expected, maintaining system stability.

2
tools/autotag/templates/highlights/6.2.4.md
View File

@@ -32,7 +32,7 @@ ROCm documentation continues to be updated to provide clearer and more comprehen
a wider variety of user needs and use cases.
* Added a new GPU cluster networking guide. See
[Cluster network performance validation for AMD Instinct accelerators](https://rocm.docs.amd.com/projects/gpu-cluster-networking/en/latest/index.html).
[Cluster network performance validation for AMD Instinct GPUs](https://rocm.docs.amd.com/projects/gpu-cluster-networking/en/latest/index.html).
This documentation provides guidelines on validating network configurations
in single-node and multi-node environments to attain optimal speed and bandwidth

2
tools/autotag/templates/highlights/6.3.0.md
View File

@@ -138,7 +138,7 @@ wider variety of user needs and use cases.
documentation](https://rocm.docs.amd.com/projects/Tensile/en/docs-6.3.0/src/index.html).
- New documentation has been added to explain the advantages of enabling the IOMMU in passthrough
mode for Instinct accelerators and Radeon GPUs. See [Input-Output Memory Management
mode for Instinct GPUs and Radeon GPUs. See [Input-Output Memory Management
Unit](https://rocm.docs.amd.com/en/docs-6.3.0/conceptual/iommu.html).
- The HIP documentation has been updated and includes the following new topics:

6
tools/autotag/templates/highlights/6.3.1.md
View File

@@ -26,9 +26,9 @@ documentation to verify compatibility and system requirements.
The following are notable new features and improvements in ROCm 6.3.1. For changes to individual components, see
[Detailed component changes](#detailed-component-changes).
### Per queue resiliency for Instinct MI300 accelerators
### Per queue resiliency for Instinct MI300 GPUs
The AMDGPU driver now includes enhanced resiliency for misbehaving applications on AMD Instinct MI300 accelerators. This helps isolate the impact of misbehaving applications, ensuring other workloads running on the same accelerator are unaffected.
The AMDGPU driver now includes enhanced resiliency for misbehaving applications on AMD Instinct MI300 GPUs. This helps isolate the impact of misbehaving applications, ensuring other workloads running on the same GPU are unaffected.
### ROCm Runfile Installer
@@ -38,7 +38,7 @@ ROCm 6.3.1 introduces the ROCm Runfile Installer, with initial support for Ubunt
ROCm documentation continues to be updated to provide clearer and more comprehensive guidance for a wider variety of user needs and use cases.
* Added documentation on training a model with ROCm Megatron-LM. AMD offers a Docker image for MI300X accelerators
* Added documentation on training a model with ROCm Megatron-LM. AMD offers a Docker image for MI300X GPUs
containing essential components to get started, including ROCm libraries, PyTorch, and Megatron-LM utilities. See
[Training a model using ROCm Megatron-LM](https://rocm.docs.amd.com/en/latest/how-to/rocm-for-ai/train-a-model.html)
to get started.

4
tools/autotag/templates/known_issues/6.2.0.md
View File

@@ -45,9 +45,9 @@ bash -c "echo taskset -p \$\$"
See [issue #3493](https://github.com/ROCm/ROCm/issues/3493) on GitHub.
### Display issues on servers with Instinct MI300-series accelerators when loading AMDGPU driver
### Display issues on servers with Instinct MI300-Series GPUs when loading AMDGPU driver
AMD Instinct MI300-series accelerators and third-party GPUs such as the Matrox G200 have an issue impacting video
AMD Instinct MI300-Series GPUs and third-party GPUs such as the Matrox G200 have an issue impacting video
output. The issue was reproduced on a Dell server model PowerEdge XE9680. Servers from other vendors utilizing Matrox
G200 cards may be impacted as well. This issue was found with ROCm 6.2.0 but is present in older ROCm versions.

2
tools/autotag/templates/known_issues/6.2.1.md
View File

@@ -5,7 +5,7 @@ individual components are listed in the [Detailed component changes](detailed-co
### Instinct MI300X GPU recovery failure on uncorrectable errors
For the AMD Instinct MI300X accelerator, GPU recovery resets triggered by uncorrectable errors (UE) might not complete
For the AMD Instinct MI300X GPU, GPU recovery resets triggered by uncorrectable errors (UE) might not complete
successfully, which can result in the system being left in an undefined state. A system reboot is needed to recover from
this state. Additionally, error logging might fail in these situations, hindering diagnostics.

8
tools/autotag/templates/known_issues/6.3.0.md
View File

@@ -5,13 +5,13 @@ issues related to individual components, review the [Detailed component changes]
### Instinct MI300X reports incorrect raw GPU timestamps
On MI300X accelerators, the command processor firmware reports incorrect raw GPU timestamps. This
On MI300X GPUs, the command processor firmware reports incorrect raw GPU timestamps. This
issue is under investigation and will be addressed in a future release. See [GitHub issue #4079](https://github.com/ROCm/ROCm/issues/4079).
### Instinct MI300 series: backward weights convolution performance issue
### Instinct MI300 Series: backward weights convolution performance issue
A performance issue affects certain tensor shapes during backward weights convolution when using
FP16 or FP32 data types on Instinct MI300 series accelerators. This issue will be addressed in a future ROCm release.
FP16 or FP32 data types on Instinct MI300 Series GPUs. This issue will be addressed in a future ROCm release.
See [GitHub issue #4080](https://github.com/ROCm/ROCm/issues/4080).
To mitigate the issue during model training, set the following environment variables:
@@ -77,7 +77,7 @@ This issue will be addressed in a future ROCm release. See [GitHub issue #4085](
Canny edge detection kernels might access out-of-bounds memory locations while
computing gradient intensities on edge pixels. This issue is isolated to
Canny-specific use cases on Instinct MI300 series accelerators. This issue is
Canny-specific use cases on Instinct MI300 Series GPUs. This issue is
resolved in the [MIVisionX `develop` branch](https://github.com/ROCm/mivisionx)
and will be part of a future ROCm release. See [GitHub issue #4086](https://github.com/ROCm/ROCm/issues/4086).

4
tools/autotag/templates/resolved_issues/6.3.1.md
View File

@@ -3,9 +3,9 @@
The following are previously known issues resolved in this release. For resolved issues related to
individual components, review the [Detailed component changes](#detailed-component-changes).
### Instinct MI300 series: backward weights convolution performance issue
### Instinct MI300 Series: backward weights convolution performance issue
Fixed a performance issue affecting certain tensor shapes during backward weights convolution when using FP16 or FP32 data types on Instinct MI300 series accelerators. See [GitHub issue #4080](https://github.com/ROCm/ROCm/issues/4080).
Fixed a performance issue affecting certain tensor shapes during backward weights convolution when using FP16 or FP32 data types on Instinct MI300 Series GPUs. See [GitHub issue #4080](https://github.com/ROCm/ROCm/issues/4080).
### ROCm Compute Profiler and ROCm Systems Profiler post-upgrade issues

2
tools/autotag/templates/resolved_issues/6.3.2.md
View File

@@ -19,7 +19,7 @@ This issue has been fixed in the ROCm 6.3.2 release. See [GitHub issue #4085](ht
An issue where Canny edge detection kernels accessed out-of-bounds memory locations while
computing gradient intensities on edge pixels has been fixed. This issue was isolated to
Canny-specific use cases on Instinct MI300 series accelerators. See [GitHub issue #4086](https://github.com/ROCm/ROCm/issues/4086).
Canny-specific use cases on Instinct MI300 Series GPUs. See [GitHub issue #4086](https://github.com/ROCm/ROCm/issues/4086).
### AMD VCN instability with rocDecode

4
tools/autotag/templates/support/6.3.1.md
View File

@@ -1,8 +1,8 @@
## Operating system and hardware support changes
ROCm 6.3.1 adds support for Debian 12 (kernel: 6.1). Debian is supported only on AMD Instinct accelerators. See the installation instructions at [Debian native installation](https://rocm.docs.amd.com/projects/install-on-linux/en/docs-6.3.1/install/native-install/debian.html).
ROCm 6.3.1 adds support for Debian 12 (kernel: 6.1). Debian is supported only on AMD Instinct GPUs. See the installation instructions at [Debian native installation](https://rocm.docs.amd.com/projects/install-on-linux/en/docs-6.3.1/install/native-install/debian.html).
ROCm 6.3.1 enables support for AMD Instinct MI325X accelerator. For more information, see [AMD Instinct™ MI325X Accelerators](https://www.amd.com/en/products/accelerators/instinct/mi300/mi325x.html).
ROCm 6.3.1 enables support for AMD Instinct MI325X GPU. For more information, see [AMD Instinct™ MI325X GPUs](https://www.amd.com/en/products/accelerators/instinct/mi300/mi325x.html).
See the [Compatibility
matrix](https://rocm.docs.amd.com/en/docs-6.3.1/compatibility/compatibility-matrix.html)

2
tools/autotag/templates/support/6.3.2.md
View File

@@ -1,6 +1,6 @@
## Operating system and hardware support changes
ROCm 6.3.2 adds support for Azure Linux 3.0 (kernel: 6.6). Azure Linux is supported only on AMD Instinct accelerators. For more information, see [Azure Linux installation](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/install/quick-start.html).
ROCm 6.3.2 adds support for Azure Linux 3.0 (kernel: 6.6). Azure Linux is supported only on AMD Instinct GPUs. For more information, see [Azure Linux installation](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/install/quick-start.html).
See the [Compatibility
matrix](https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html)