Posts by Liz Li

Optimizing DeepseekV3 Inference on SGLang Using ROCm Profiling Tools

As LLMs are growing in size and complexity, ensuring proper utilization of compute resources becomes of prime importance. Performance profiling and kernel-level analysis are essential techniques for diagnosing runtime bottlenecks, such as GPU time, memory-bound operations, and inefficient device-host memory transfers etc. By using profiling tools like RocmProfileData (RPD) and TorchProfiler (PyTorch Profiler) developers have access to granular level insight into kernel execution timelines, data movement patterns, and computational hotspots. In this blog, we delve into how profiling and kernel diagnostics can expose inefficiencies in components like attention mechanisms and Mixture-of-Experts (MoE) layers — and guide targeted optimizations at the kernel level.

Read more ...


Boosting Llama 4 Inference Performance with AMD Instinct MI300X GPUs

In our previous blog post, we explored how to deploy Llama 4 using AMD Instinct™ MI300X GPUs with vLLM. We also highlighted that MI300X and MI325X GPUs are capable of running the full 400B-parameter Llama 4 Maverick model in BF16 precision on a single node, significantly reducing infrastructure complexity. Their substantial HBM memory capacity further supports extended context lengths, enabling high throughput and efficient model execution.

Read more ...


Power Up Llama 4 with AMD Instinct: A Developer’s Day 0 Quickstart

AMD is excited to announce Day 0 support for Meta’s latest leading multimodal intelligence Models — the Llama 4 Maverick and Scout models — on our AMD Instinct™ MI300X and MI325X GPU accelerators using vLLM. In this blog we will walk you through a step-by-step guide on deploying Meta’s Llama4 model using vLLM, docker setup, dependencies, and inference testing.

Read more ...


Bring FLUX to Life on MI300X: Run and Optimize with Hugging Face Diffusers

AI based text-to-image generation is pushing the boundaries of creative and visual storytelling, enabling the critical mass to draw like an artist. Stability AI introduced stable diffusion models which was a breakthrough in text to image generation. However, FLUX - a new state-of-the-art open-source model released by Black Forest Labs, is gaining popularity for its flexibility and controllability.

Read more ...


Supercharge DeepSeek-R1 Inference on AMD Instinct MI300X

Our previous blog post on this topic discussed how DeepSeek-R1 achieves competitive performance on AMD Instinct™ MI300X GPUs. We also included performance comparisons against Nvidia H200 GPUs and a short demo application illustrating real-world usage. In this blog we will delve into how using the SGLang framework, critical kernel optimizations like AI Tensor Engine for ROCm™, and hyperparameter tuning helps to achieve performance boosts.

Read more ...


AITER: AI Tensor Engine For ROCm

Performance optimization is critical when working with GPUs, especially for tasks involving artificial intelligence, which can be extremely demanding. To fully leverage the capabilities of advanced hardware, it’s essential to master optimization strategies and ensure every available resource is utilized efficiently. In this blog we will provide an overview of AMD’s AI Tensor Engine for ROCm (AITER) and show you how easy it is to integrate AITER kernels in basic LLM training and inference workload. AITER helps developers to focus on creating operators while allowing customers to seamlessly integrate this operator collection into their own private, public, or any custom framework.

Read more ...


AITER: AI Tensor Engine For ROCm

Performance optimization is critical when working with GPUs, especially for tasks involving artificial intelligence, which can be extremely demanding. To fully leverage the capabilities of advanced hardware, it’s essential to master optimization strategies and ensure every available resource is utilized efficiently. In this blog we will provide an overview of AMD’s AI Tensor Engine for ROCm (AITER) and show you how easy it is to integrate AITER kernels in basic LLM training and inference workload. AITER helps developers to focus on creating operators while allowing customers to seamlessly integrate this operator collection into their own private, public, or any custom framework.

Read more ...