search

alumae / kaldi-gstreamer-server

Real-time full-duplex speech recognition server, based on the Kaldi toolkit and the GStreamer framwork.
BSD 2-Clause "Simplified" License
1.07k stars 341 forks source link

Fork post-processors before loading the language model. #37

Closed dwks closed 8 years ago

dwks commented 8 years ago

This prevents out-of-memory errors when the system's RAM is nearly full. With this commit, I can launch a worker which requires 2.4GB of RAM on a server with 2.6GB free; otherwise, I get an out of memory error. Once the LM has been loaded into memory, the footprint is huge so a fork is expensive.

alumae commented 8 years ago

Thanks!