.
Sign in

Technology

ROCm

AMD's open-source software stack for high-performance GPU computing and large-scale AI model training.

ROCm (Radeon Open Compute) provides a modular, open-source environment for GPU programming across Linux and Windows. It powers massive deployments like the Frontier supercomputer (1.1 exaflops) and supports industry-standard frameworks: PyTorch, TensorFlow, and JAX. Developers use the HIP (Heterogeneous-compute Interface for Portability) compiler to port CUDA code to AMD hardware with minimal friction. The stack includes specialized libraries (rocBLAS and MIOpen) to squeeze maximum performance out of Instinct MI300X and Radeon RX 7000 series silicon.

https://rocm.docs.amd.com/
4 projects · 4 cities

Related technologies

Llama-2 337 OpenAI API 500 PyTorch 264 CUDA 21 Docker 157 GPT-4 678 Hyprstream 1 Kubernetes 34 LangChain 439 LoRA 26 Mistral 36 Ollama 82 SkyPilot 1 SLURM 2 Transformers 168 vLLM 33

Recent Talks & Demos

Showing 1-4 of 4

Members-Only

Sign in to see who built these projects

Sign in View FAQ
Hyprstream: Secure Versioned AI Models
Milan Feb 24
Hyprstream LoRA
Transformer Lab: Local to Distributed ML
Toronto Jan 29
Kubernetes SLURM
Strix Halo Unified Memory AI
Tokyo Oct 10
GPT-4 Llama-2
Ollama Groq Local Inference
Manizales Jan 22
Llama-2 Mistral
Events Gallery