Find trending repositories by name or description.
Xolotl: run DeepSeek-V4.1-Flash, all 552B, as a local coding agent on one gaming PC (RTX 4090 + 110 GB RAM): ~33 tok/s on agent tasks, ~41 in fast mode, 512K context. Plus GLM-5.3-Flash, Qwen3.8 and LFM3-Omni. MoE experts tiered across VRAM, pinned RAM and NVMe, with Xolotl Direct NVMe->VRAM. OpenAI + Anthropic API.
Run 35B MoE models in ~2.5 GB of RAM. Weights stream from SSD; a trained router predicts loads a step ahead. Mac, iPhone, Android, Windows supported
Naive-N0.5-Flash is an open-weight 309B MoE model with 15.5B active parameters, built for coding and AI R&D.
Pure-Rust inference engine for Qwen GGUF models (Qwen3 to Qwen3.8, dense and MoE up to 770B+). CUDA and Vulkan (NVIDIA, AMD), hybrid CPU+GPU experts, MTP speculative decoding, OpenAI and Anthropic compatible server. Up to ~1.7x llama.cpp decode on a 12 GB RTX 3060.
Lsglang is a special extension of sglang that fully utilizes CPU and GPU computing resources with an efficient GPU parallel + NUMA parallel architecture, suitable for MOE model hybrid inference.
Timestamp Retrieval for Anime Clips Everywhere
A command line based editor inspired by Vim. Written in Nim.
整理的《朱颜血》小说
API server for serving backend requests for trace.moe
Moe counter badge with multiple themes! - 多种风格可选的萌萌计数器
Tutel MoE: Optimized Mixture-of-Experts Library, Support GptOss/DeepSeek/Kimi-K2/Qwen3 using FP8/NVFP4/MXFP4
Efficient Mixture of Experts for LLM Paper List
Best practices for training DeepSeek, Mixtral, Qwen and other MoE models using Megatron Core.
ComfyUI-QwenVL custom node: Integrates the Qwen-VL series, including Qwen2.5-VL, Qwen3-VL, Qwen3.5-VL, Qwen3.6-VL (MoE), and Qwen3.8-VL, with GGUF support for advanced multimodal AI in text generation, image understanding, and video analysis.
LvLLM is a special NUMA extension of vllm that makes full use of CPU and memory resources, reduces GPU memory requirements, and features an efficient GPU parallel and NUMA parallel architecture, supporting hybrid inference for MOE large models.
🎓 系统性大语言模型构建课程｜🛠️ 覆盖预训练数据工程、Tokenizer、Transformer、MoE、GPU 编程 (CUDA/Triton)、分布式训练、Scaling Laws、推理优化及对齐 (SFT/RLHF/GRPO)｜🚀 6 个渐进式作业 + 代码驱动，建立 LLM 全栈认知体系
Accelerating MoE with IO and Tile-aware Optimizations
[CVPR2026]🚀🚀🚀Official code for the paper "YOLO-Master: MOE-Accelerated with Specialized Transformers for Enhanced Real-time Detection." *(YOLO = You Only Look Once)* 🔥🔥🔥
Automated alignment adjustment for LLMs — direct steering, LoRA, and MoE expert-granular abliteration, optimized via multi-objective Optuna TPE.
Running a big model on a small laptop
⚡ Native MLX Swift LLM inference server for Apple Silicon. OpenAI-compatible API, SSD streaming for 100B+ MoE models, TurboQuant KV cache compression, MACOS + iOS iPhone app.
Compact and Agent-Native MoE Training System
Adaptive Precision for EXpert Models: MoE-aware mixed-precision quantization
DeepSeek V4 Pro: Advanced AI desktop app. Features: 1.6T MoE architecture, 1M token context window, Engram memory. Pro coding agent, Think Mode (High/Max), repo-level reasoning, complex bug fixing. Real-time web search, local API key storage, cross-file analysis, low-latency streaming interface.
A hands-on course for building modern LLMs from scratch in PyTorch, with 26 runnable Jupyter Notebooks covering tokenizers, attention, MoE, RLHF, inference, evaluation, and distillation.
Zonos2 is a leading open-weight text-to-speech MoE.
deepseek v4 flash pro chat api app windows 11 desktop client github open source ai assistant chatbot coding agent moe model 1.6t parameters free api setup guide reasoning max mode local deployment
A vLLM patch + hand‑written SM120 SASS kernels: 2‑bit MoE experts + an FP4 "delta" cache that recovers precision — matching the official (NV)FP4 checkpoint's quality on consumer Blackwell cards
Run frontier MoE models on hardware you already own — pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦
Run MoE models bigger than your RAM. Frontier-size MoE on a 12 GB phone, CPU only, lossless, on stock llama.cpp