amd-rocm

面向 DeepSeek Harness 的本地图像生成插件（优化版）：经本机 ComfyUI 调用 Qwen-Image 2.1 生成图像，全部流量限于 127.0.0.1。本版增强推理性能、NVIDIA/Intel/AMD 显卡适配与 Windows/Linux 兼容。原作者 @奇迹与你，优化版 @yanfei0725。

拒绝一次性Demo！基于 AMD ROCm 与 LeRobot 框架，在 RDNA3 GPU 上真机跑通通用具身智能（π0.5/VLA）模型前向推理与微调的环境验证与端到端指南。

录音转写与会议纪要 · 自部署 Web 版 · GPU 加速 · 隐私优先 · Recording transcription & meeting notes self-hosted web app

A machine learning compiler for GPUs, CPUs, and ML accelerators

AMD ROCm AI Inference & Training Solutions Library for RNDA2/GFX1030/Radeon Pro v620 includes: compiling vLLM Source, Llama.cpp, HIPFire, and Fine-Tuning paths.

ANALYSIS OF --DLSS 5 Neural Rendering on AMD GPU-- DUMP!

3D, LLM-driven adventure engine: a walkable three.js frontend plus a local-AI backend (AMD Strix Halo GGUF via llama.cpp, optional ComfyUI) where an AI narrator generates explorable 3D worlds and sentient NPCs you talk to. Strictly isolated contract-per-layer architecture. 3D sibling of gamentic.

AI runtime optimized for agentic work with Qwen models on AMD RX 7900 XT/XTX

Run PyTorch natively on Windows 11 with an AMD RX 9070 XT (RDNA4 / gfx1201) on stable ROCm 7.2.1 — no WSL2, no Linux, no ZLUDA. Exact pinned wheel URLs, runtime env vars, documented RDNA4 pitfalls (broken nightlies, no xformers/flash-attn/bitsandbytes), and real benchmarks for ComfyUI, SD, LLMs, RVC/TTS. Verified on one 9070 XT.

Lightweight terminal dashboards for monitoring ComfyUI on Ubuntu with AMD ROCm or NVIDIA CUDA GPUs. Tracks VRAM, system RAM, ComfyUI RSS and swap, memory pressure, GPU activity, temperatures, cooling, power, clocks, and rolling restart warnings.

MinerU on AMD ROCm/Windows-HIP — evaluation-backed port. OmniDocBench v1.6: Linux W7900 pipeline 86.48 + VLM 95.56. Windows Strix Halo: pipeline 86.59 verified + VLM llama.cpp HIP 95.46 verified.

📖 中文小说转有声读物 CLI | AMD 消费级显卡本地运行 Qwen3-TTS + Edge TTS | 可能是 GitHub 上唯一针对 AMD 显卡的中文小说 TTS 完整方案

Autonomous global cargo pathfinder routing agent.

RyFlow is an offline-first AI collaboration workspace. Run local LLMs, collaborate in real time over LAN, create tasks with plain English, index local folders and github repos and build a semantic knowledge graph that lasts. Zero cloud, zero cost, zero compromise.

绝区零「仪玄」角色助手后端 | Qwen3-4B + LoRA 微调 + vLLM 部署 + RAG 检索 + 防 OOC 校验器 | AMD ROCm 全栈实战

One-command OmniDocBench v1.6 evaluation setup on AMD Windows (ROCm/HIP). Model-agnostic: swap any document parsing model via adapters. PaddleOCR-VL-1.6 as reference. CDM formula scoring included.

Standalone HIP Winograd WMMA inference for AMD GPUs

One command to install AMD ROCm 10.0.0 + PyTorch on Ubuntu (WSL2 or native), then verify your GPU + diffusion ready

Portable, privacy-first AI context layer & MCP server. Classifies personal data on-device via fine-tuned Gemma 3 & WebGPU, keeping private data offline.

ACRE (Adaptive Cognitive RAG Engine) is a RAG engine that reasons like a researcher: it decomposes questions, retrieves and critiques evidence, resolves contradictions across sources, and renders its reasoning as a live concept flowchart — not just a wall of text.