deepspeedai / DeepSpeed

DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective.

hpcaitech / ColossalAI

Making large AI models cheaper, faster and more accessible

epoko77-ai / im-not-ai

AI가 쓴 한글을 사람 글처럼 윤문하는 Claude 스킬 — Korean AI-text humanizer: detects and rewrites translationese, mechanical parallelism, and 71 other AI tells

xlite-dev / Awesome-LLM-Inference

📚A curated list of Awesome LLM/VLM Inference Papers with Codes: Flash-Attention, Paged-Attention, WINT8/4, Parallelism, etc.🎉

huggingface / nanotron

Minimalistic large language model 3D-parallelism training

xdit-project / xDiT

xDiT: A Scalable Inference Engine for Diffusion Transformers (DiTs) with Massive Parallelism

nerevu / riko

A Python stream processing engine modeled after Yahoo! Pipes

Lightning-AI / lightning-thunder

PyTorch compiler that accelerates training and inference. Get built-in optimizations for performance, memory, parallelism, and easily write your own.

vipshop / cache-dit

A PyTorch-native inference engine with cache, parallelism, quantization and cpu offload for DiTs.

NVIDIA-NeMo / labs-molt

A scalable, agentic-first, and HuggingFace-native RL framework for research (9k lines).

MoonshotAI / MoonEP

MoonEP: A Perfectly Balanced Expert Parallelism Library via Dynamic Redundant Experts

whitecircle / halo

Halo is an open-source framework built by White Circle for training large language and multimodal models

KarhouTam / FL-bench

Benchmark of federated learning. Dedicated to the community. 🤗

google / paxml

Pax is a Jax-based machine learning framework for training large scale models. Pax allows for advanced and fully configurable experimentation and parallelization, and has demonstrated industry leading model flop utilization rates.

allover326 / deepseek-v4-cmp170hx

DeepSeek-V4-Flash-0731 on 4x CMP 170HX (sm_80): 98 tok/s decode, ~5300 tok/s prefill. DSpark speculative decoding under pipeline parallelism, which vLLM does not support upstream.

agentscope-ai-java / interp-rpc

Typed RPC over a warm pool of PEP 734 subinterpreters — real parallelism, real error propagation, no hanging calls.

christopherowen / dgx-spark-networking

Switchless RoCE collectives, NCCL profiles and fabric tooling for DGX Spark inference recipes

buqi-code / buqi-minimax-h3-multigpu

MiniMax-H3 multi-GPU parallel inference for ComfyUI | 多卡并行加速节点:2-8 GPU Ulysses sequence parallel, bit-identical video+audio generation