LaunchDevelopers
15 days ago
Eagle 3.1 released with collaboration between EAGLE, vLLM, and TorchSpec
Eagle 3.1, a new version of the speculative decoding engine, is released through collaboration between the EAGLE team, vLLM team, and TorchSpec team. The update aims to improve inference efficiency for large language models. vLLM integrates Eagle for faster token generation.