Pinned Loading

1. OpenBMB/CPM.cu OpenBMB/CPM.cu Public

   CPM.cu is a lightweight, high-performance CUDA implementation for LLMs, optimized for end-device inference and featuring cutting-edge techniques in sparse architecture, speculative sampling and qua…

   Cuda 199 19

2. thunlp/FR-Spec thunlp/FR-Spec Public

   [ACL 2025 main] FR-Spec: Frequency-Ranked Speculative Sampling

   C++ 44 2

3. thunlp/Ouroboros thunlp/Ouroboros Public

   Ouroboros: Speculative Decoding with Large Model Enhanced Drafting (EMNLP 2024 main)

   Python 110 9

4. OpenBMB/MiniCPM OpenBMB/MiniCPM Public

   MiniCPM4 & MiniCPM4.1: Ultra-Efficient LLMs on End Devices, achieving 3+ generation speedup on reasoning tasks

   Jupyter Notebook 8.4k 520

5. OpenBMB/BMTrain OpenBMB/BMTrain Public

   Efficient Training (including pre-training and fine-tuning) for Big Models

   Python 612 82

6. thunlp/OpenPrompt thunlp/OpenPrompt Public

   An Open-Source Framework for Prompt-Learning.

   Python 4.7k 478