#3,471 of 5,656 by stars
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.
- Stars
- 14,751
- Forks
- 1,024
- Created
- 5y ago
- Last push
- 20d ago
Momentum
- Per day
- +3
- Per 7 days
- +6
- Per 30 days
- +39
- Rank, 7 days
- -16
Stars over time
View as table
| Date | Stars |
|---|---|
| Sep 17, 2026 | 14,709 |
| Oct 5, 2026 | 14,738 |
| Oct 12, 2026 | 14,745 |
| Oct 19, 2026 | 14,751 |
Understand this repo
Open it in a tool that explains, diagrams or digests the code.
- Explain A plain-language walkthrough of what the project is and does
- Diagram An interactive architecture diagram of the codebase
- Ingest for AI The whole repository as one text digest to paste into an AI assistant
- Reverse prompt A prompt describing how to rebuild the project
These open independent third-party sites, which may be slow or unavailable for very large repositories.
Similar repositories
Ranked by how many distinctive topics they share with this one.
- 3771
Code for loralib, an implementation of "LoRA: Low-Rank Adaptation of Large Language Models"
- 1644
Implementation of Vision Transformer, a simple way to achieve SOTA in vision classification with only a single transformer encoder, in Pytorch
- 3637
A framework for few-shot evaluation of language models.
- 118
Implement a ChatGPT-like LLM in PyTorch from scratch, step by step
- 3109
- 2394
邱锡鹏《神经网络与深度学习》第二版与通识版：电子书、章节目录、学习资源与勘误。