📈 GH Trending Explorer 🎲 Surprise me

Search: eliza · Language: Python · clear

BlinkDL / RWKV-LM

RWKV (pronounced RwaKuv) is an RNN with great LLM performance, which can also be directly trained like a GPT transformer (parallelizable). We are at RWKV-7 "Goose". So it's combining the best of RNN and transformer - great performance, linear time, constant space (no kv-cache), fast training, infinite ctx_len, and free sentence embedding.

▲ ~0/7d ★ 15k Python 23d ago
PaddlePaddle / PARL

A high-performance distributed training framework for Reinforcement Learning

▲ ~0/7d ★ 3.5k Python 21d ago
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.

▲ ~0/7d ★ 560 Python 6d ago
kunifujiwara / VoxCity

Python framework for integrating geospatial data, 3D city modeling, and urban simulation.

▲ ~0/7d ★ 555 Python 16d ago
KohakuBlueleaf / PixelOE

Detail-Oriented Pixelization based on Contrast-Aware Outline Expansion.

▲ ~0/7d ★ 530 Python 22d ago
Phyzicalorg / Phyzical_org

Browser teleoperation data for embodied AI — elizaOS-ready episodes, trajectory_db converter, onchain provenance. The fuel station for agent robot stacks.

▲ ~0/7d ★ 273 Python 1mo ago
PillCrew / claimchain

Verify that an AI agent's on-chain claims are actually true. A claim-level groundedness checker for Solana / memecoin agents.