vllm-project/vllm
A high-throughput and memory-efficient inference and serving engine for LLMs
A star delta has been computed for 62,252 of 63,106 tracked repositories in this period, and only those can be ranked in it.
A repository missing here has not been measured far enough back yet. That does not mean it did not move.
Star counts measured between 08:04 and 14:26 (BKK).
A high-throughput and memory-efficient inference and serving engine for LLMs
A Next-Generation Training Engine Built for Ultra-Large MoE Models
Model Context Protocol server for DeepSeek's advanced language models
Implementing LLM architectures from scratch in PyTorch
That's all 4 repositories in this period.