Dominik Nowak

Strata 0.1.43 (Qwen3.8-Flash-Next) for AMD gfx906: MI50/MI60, Radeon VII, web configurator

Ollama v0.40.2 / MI50 (gfx906), ROCm 7.2, MTP (multi-token prediction) supported.

Optimized llama.cpp for AMD Instinct MI50/MI60/Radeon VII (gfx906).

Ollama 0.15.4 + ROCm 7.1 user-space, Tensile libs gfx906 for AMD Radeon Instinct MI50 32 GB

High-performance CUDA RPC Node for Hybrid LLM Clusters. Aggregates VRAM with AMD nodes.

Optimized RPC Node for AMD MI50 (gfx906) & ROCm 7.1. Part of xxDoman Hybrid Cluster.

Ollama + ROCm 7.1 dla AMD MI50 (gfx906)