Search

🧠 Train a 64M-parameter LLM from scratch in just 2h!

Official repository for MOSS-Image, a 5B-parameter base DiT model with high-quality text rendering, layout generation, and efficient fine-tuning.

Use PEFT or Full-parameter to CPT/SFT/DPO/GRPO 600+ LLMs (Qwen3.8, DeepSeek-V4, GLM-5.1, InternLM3, Llama4, ...) and 300+ MLLMs (Qwen3-VL, Qwen3-Omni, InternVL3.5, Ovis2.5, GLM5.3, Gemma4, Llava, Phi4, ...) (AAAI 2025).

🤗 PEFT: State-of-the-art Parameter-Efficient Fine-Tuning.

A Flexible and Powerful Parameter Server for large-scale machine learning

Simulation and Parameter Estimation in Geophysics - A python package for simulation and gradient based parameter estimation in the context of geophysical applications.

Set of support modules for Java 8 datatypes (Optionals, date/time) and features (parameter names)

Lazy Predict help build a lot of basic models without much code and helps understand which models works better without any parameter tuning

An anomaly detection library comprising state-of-the-art algorithms and features such as experiment management, hyper-parameter optimization, and edge inference.

Wrapper for 10+ VPR models. Use any SOTA VPR model just by changing one parameter

A high-performance RL training-inference weight synchronization framework, designed to enable second-level parameter updates from training to inference in RL workflows

SoulX-FlashHead: A unified 1.3B-parameter framework designed for high-fidelity, infinite-length, and real-time streaming portrait video generation.

Find the local LLM that actually runs and performs best on your hardware. Ranked by real, recency-aware benchmarks, not parameter count. One command, run it instantly.

A ~9M parameter LLM that talks like a small fish.

A 100M-parameter multilingual TTS model for real-time CPU inference, voice cloning, and 48 kHz stereo generation

A 3B-active-parameter native unified multimodal model for image and video understanding, generation, and editing.

A tiny ~10K-parameter LLM router that learns which open-source model (deepseek-v4-pro / glm-5p2 / kimi-k2p6 via Fireworks) should answer each question and in what role, trained by evolution (sep-CMA-ES).

Run the full 2.78-trillion-parameter Kimi K3 model, DeepSeek V4.1 Flash or GLM-5.3-Flash beyond available RAM by streaming activated weights directly from NVMe. A dependency-free, embeddable C inference engine.

A 2.78-trillion-parameter Kimi K3 running inference on a single CPU in 8.24 GB of RAM. Portable C99: no BLAS, no framework, no GPU.