Note: This document accompanies limited preview container images designed to validate and reproduce inference and training performance on AMD’s MI355X and MI350X accelerators, announced during the AMD Advancing AI event on June 12, 2025. The images provide access to pre-release version of the ROCm 7.0 software stack and are targeted at early-access users evaluating inference and training workloads using next-generation AMD GPU hardware.
The goal of this preview is to provide hands-on benchmarking capability using representative large-scale language and reasoning models, with optimized compute precisions and configurations.
Benchmark Llama 3.1 405B FP4 inference with vLLM
Benchmark Llama 3.3 70B FP8 inference with vLLM
Benchmark GPT OSS 120B inference with vLLM
Benchmark DeepSeek R1 FP4 inference with SGLang
Benchmark DeepSeek R1 FP8 inference with SGLang
Benchmark Llama 2 70B LoRA fine-tuning with MLPerf
Benchmark Llama 3 pre-training with Megatron-LM
Content type
Image
Digest
sha256:8f232de0e…
Size
10.9 GB
Last updated
12 months ago
docker pull rocm/7.0-preview:rocm7.0_rel_30_ubuntu22.04_py3.10_pytorch_release_2.8.0