aisa-group / PostTrainBench

Measuring how well CLI agents like Claude Code or Codex CLI can post-train base LLMs on a single H100 GPU in 10 hours

neuralhydrology / neuralhydrology

Python library to train neural networks with a strong focus on hydrological applications.

google / cameratrapai

AI models trained by Google to classify species in images from motion-triggered wildlife cameras.

NVIDIA / cosmos-framework

Our inference and training framework to run on the Cosmos Models

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.

hmorimitsu / ptlflow

PyTorch Lightning Optical Flow models, scripts, and pretrained weights.

thu-ml / SLA

SLA: Beyond Sparsity in Diffusion Transformers via Fine-Tunable Sparse–Linear Attention

AetherLabsAI / RSIAgent

A training-free multi-agent framework for recursive self-improvement in new environments through broad-then-deep autonomous exploration and reusable memory.

meta-pytorch / torchft

Fault tolerance for PyTorch (HSDP, LocalSGD, DiLoCo, Streaming DiLoCo)

guillermo-navas-palencia / optbinning

Optimal binning: monotonic binning with constraints. Support batch & stream optimal binning. Scorecard modelling and counterfactual explanations.

aws / sagemaker-training-toolkit

Train machine learning models within a 🐳 Docker container using 🧠 Amazon SageMaker.

MIC-DKFZ / nnInteractive

nnInteractive is a framework for 3D interactive segmentation, supporting intuitive prompts like points, scribbles, bounding boxes, and lasso. Trained on 120+ diverse 3D datasets, it sets a new standard in accuracy, usability, and adaptability for clinical and research applications.

pathsim / pathsim

A Python native dynamical system simulation framework in the block diagram paradigm.

official-stockfish / nnue-pytorch

Stockfish NNUE (Chess evaluation) trainer in Pytorch

OpenMOSS / EasyWAM

A unified framework for training, fine-tuning, and evaluating World Action Models

jrz97619761 / test-model-thing

A small proof-of-concept language model (not an LLM) that aims to incorporate latent-space prediction, internal state using recurrent trace units, and byte-by-byte output, built with MLX.

DaoyuanLi2816 / mini-verl

verl for a single consumer GPU. PPO, GRPO and on-policy distillation on NVIDIA GPUs.

Hebbian-Robotics / hflow

SDK for robotics teams to verify the quality of their data used for AI model training.

znxzsy / VeriInk

Verifiable handwriting synthesis for failure-driven multimodal training and regression.

jinzijian / EvoTrace

Compile real-world Claude Code and Codex trajectories into verified, tradable post-training assets.

Tencent / AngelSpec

A unified, torch-native training framework for MTP and block-parallel speculative decoding.

lcpu-club / wmhpc-training-camp-x-lcpu-ai-infra-seminars

北京大学未名超算队与北京大学学生 Linux 俱乐部合办的暑期 AI Infra 系列活动仓库

alexiglad / XM

PyTorch Code for Explorative Modeling: Unlocking a Third Pretraining Axis and End-to-End Generation

Mercor-Intelligence / ApexAgents-SkyRL-Recipe

RL-training knowledge-work agents on APEX-Agents data with SkyRL + Harbor

HarlandZZC / OmniVChat

Synthesizing, Benchmarking, and Training for Native Audio-Visual Dialogue

Zihan-Su / Self_Gradient_Forcing_Plus

Autoregressive video generation with decoupled context-writing and denoising parameters, enabling up to 24-hour generation from 5s training windows.

PINTO0309 / High-Angle_Robust_Fast_FaceAlignment

HRFFA: High-Angle Robust Fast Face Alignment. Whole-head face alignment dataset & training pipeline.

LMResiliency / lm-resiliency

Closed-loop fault localization and checkpoint recovery for LLM Pre-training

black-forest-labs / flux-action

FLUX 3 Action: open weights 7B world action model from Black Forest Labs. Train, fine-tune and run action prediction for robots (DROID, SO-101), simulators and games.

ssaaffaakk / Lilly

Bosnian-first offline assistant: type, speak, and snap street text → English. FastAPI app + Kaggle-trained listen/read models. Fail-stop training.

zwliJay / jev-forge

An open training and inference stack for Jev-style decision models. Train models to score dynamic candidate branches from a shared prefix, with support for high-cardinality choice, calibration, and fast batched inference.

zhuang2002 / Self_Gradient_Forcing

Self Gradient Forcing (SGF) recovers the missing context-gradient path for self-generated causal memory through a bounded two-pass replay, enabling models trained with only a 5-second window to extrapolate to minute-scale videos with stronger identity, layout, and temporal stability.

zylwithxy / HRDiT

[ECCV 2026] Official implementation of "HRDiT: Training-Free High-Resolution Image Generation with Off-the-Shelf Diffusion Transformer Models"

AcademiaSD / AcademiaSD_LoRAlab-Krea2

Low-VRAM LoRA Training for Krea-2

s4ntrx / S4ntrx-Bot

The project is a multi-user cybersecurity education chatbot built with Flask, aimed at a school/classroom project.

bandr-ai / bandits

Trace mining for post-training. Filter trajectories, draft verifiers, export SFT and RL data.

Zhao-Yian / MiniWorld

[Tech Report] Democratizing the Training of Video World Models from Scratch. 🔥 🔥 🔥

KurosawaGeeker / femboy-skill

面向 MTF、crossdresser 与性别多元成年人的中文 Agent Skill，基于生如夏花知识库并加入医学安全护栏。

Jalen-Brunson / ComfyUI-MiniMax-H3-PDD-Acc

Official MiniMax-H3 8-step PDD Acc LoRAs (alibaba-pai) in ComfyUI: LoRA + parallel-decoding head bank, euler on trained sigmas, audio+video in 8 steps

denis-pplx / autojev

Full-weight Qwen decision model with calibrated probabilities, training code, REST API, and playground.

menloresearch / cyclotron

Official reinforcement learning training framework for Asimov humanoid robots, built on Isaac Lab.

rzc156 / dreamforge

Turn world-model imagination into corrective training data for robot policies — a single-GPU, falsifiable pipeline (SmolVLA · LIBERO · CEM · paired simulator replays).

fengzhizi715 / YoloWebAgentStarter

Local-first YOLO workbench for annotation, training, evaluation, and AI assistance.

jianglongye / dexterous-astra

Dexterous pen spinning and Rubik's cube: sim environments, zero-shot scripts, trained checkpoints

Dbp-dev12 / vla-preflight

Local VLA workbench: data diagnostics, reproducible preparation and reference training.

GENTEL-lab / Fold2Reason

Official code for Fold2Reason: Enhancing general reasoning in language models through protein-structure post-training.

To3akaRin / VLMForge

Evaluate visual models on your own images, JSON Schema, and production constraints with LangGraph, Pareto analysis, and a no-key Replay demo.

hkproj / torchfeather

A distributed training framework based on torchtitan

nums-ai / causilo

A pretrained tabular foundation model for classification and regression by Nums AI Inc.

aibuildai-inc / aibuildai-llm-posttrain-agent

AIBuildAI LLM-Post-Train Agent: a recursive self-improving (RSI) agent for autonomous LLM post-training

nicobrenner / jeffy

Pretrained text classifiers you can run and retrain on CPU

lucidrains / robo_ttt

Implementation of RoboTTT proposed by Yunfan Jiang et al. of Stanford and Nvidia

Bariona / FACT

[CoRL 2026] Code for "FACT: Failure-Aware Causal Training for World‑Action Models"

Blackoutta / ai-agent-fullstack-training

极客时间 AI Agent 全栈工程训练营课程代码

InkMind-AI / Dyn-3D

TempoVista is the official implementation of Dyn-3D, a benchmark and training pipeline for 3D spatial reasoning in vision-language models. The release contains frame-selection tools, SFT and Kinematic-GSPO training code, benchmark evaluation scripts, Dyn-3D annotations, and checkpoints for Qwen3-VL and InternVL3.5.

DAGroup-PKU / ViGAR

Rethinking World-Action Model for Compositional and In-context Robotic Manipulation

archer-clawbot / omarchy-hermarchy-theme

A restrained research-workstation theme for Omarchy with optional Hermes agent status

AcademiaSD / AcademiaSD_LoRAlab-TrainerStudio

Train LoRAs for Qwen-Image 2.1, FLUX.2 Klein 9B, FLUX.3 Image Klein, Krea 2, Z-Image, MOSS-Image, ERNIE-Image, Ideogram 4, Anima, SDXL (Pony, Illustrious, NoobAI), LTX 2.3, MiniMax-H3, and more on NVIDIA GPUs from 4–8 GB VRAM. One install, one launcher, NF4 models. Windows, Linux and RunPod.

erictrainnlpinxihu / Grader-agent

Grader 的 agent 采用五阶段主循环：perceive→plan→act→observe→respond，由 harness 驱动。模型只决定意图与置信度，路由、工具集、风险等级由代码钉死。act 分只读 ReAct、RAG、人工提案、批量初批。四个写动作不在工具表，只能产提案。批改输出结构化 GradingDraft，逐条绑定 rubric 与段落 hash，保证可重放。

Dreamer-Toby / STEPQuant

STEPQuant: When and Where Errors Matter in Delta-Rule Recurrent State Quantization