mi50

llama.cpp-gfx906

Open-source local AI server configs, GFX906 runtime maintenance, reproducible benchmarks, and QC methods for affordable AI research infrastructure.

FlashAttention-style custom attention backend for vLLM on AMD MI50/MI60/Radeon VII (gfx906). Downstream fork of mixa3607/ML-gfx906 with replacement HIP kernels and a vllm.general_plugins entry point.

AMD Instinct MI50/MI60 (gfx906, HIP/ROCm) port of NInfer - the specialized single-GPU Qwen inference engine. First AMD target in the ecosystem. Audit complete, port in progress.

Deploy Qwen3.6-27B GGUF on an AMD MI50 (gfx906) eGPU via vLLM + ROCm — docker-compose bundle with patches for Qwen3.5/3.6 hybrid Gated-DeltaNet support.

Strata v0.1.40 inference engine (Qwen3.8-Flash-Next) on AMD gfx906 — Instinct MI50 / MI60, Radeon VII. Docker image + web configurator (per-card VRAM reserve) + setup wizard. OpenAI/Anthropic API, image input.

ROCm/Unsloth/bitsandbytes 4-bit lab and VRAM benchmarks for AMD MI50/gfx906 LLM fine-tuning

Native C++/HIP Qwen3.8-27B inference engine for 4x AMD MI50 (gfx906), with chained MTP speculative decoding.

Qwen3.8-27B on AMD Instinct MI50 (gfx906): 19 -> 48 tok/s. Patched llama.cpp in Docker, one-command setup, EN/RU docs.

openPangu Flash92 — C++/CUDA & C++/HIP ROCm Engines. Fully resident, Blackwell tensor cores, no Python. Powered by openPangu.

Local Paperless-ngx + Ollama integration for AI-based document titles, correspondents, document types, tags, review workflows, and backfill.

Restore videos with pixelated/mosaic regions rocm_gfx906 (mi50 gpu)

Qwen 3.ocho (C++): stochastic trajectory rendering with a decision/commit loop on a native CUDA/HIP/ROCm inference engine (NVIDIA GB10 + 4x AMD MI50)

Run ComfyUI on AMD Radeon VII (gfx906) via Docker. ROCm 5.7 + PyTorch 2.3.1, SDXL 1024×1024 in ~28s/image. Pinned to ComfyUI v0.3.60 — the last gfx906-compatible build.