alphacep / vosk-server

WebSocket, gRPC and WebRTC speech recognition server based on Vosk and Kaldi libraries
Apache License 2.0
882 stars 243 forks source link

"Failed to create model" with the newest russian model. #153

Closed chamoretto closed 2 years ago

chamoretto commented 2 years ago

Hello. We have some troubles with the newest russian model. Previously we used august 2021 version of the model, now we decided to deploy additional instance of jitsi with ru and en models. The model version "en" is fine, but with the "ru" latest docker image we getting this error: image

We have tested new model on both Ubuntu 20.04 (8 cores, 12 GB RAM) and Debian 10 (4 cores, 12GB RAM), both fails to launch the newest model with the error on the screenshot above.

nshmyrev commented 2 years ago

Try to add more RAM

sskorol commented 2 years ago

Or just remove rnnlm folder from the model, which should also drop the memory consumption.

chamoretto commented 2 years ago

Try to add more RAM

We've added RAM, now it's 4 cores and 14 GB of RAM is enough per one instance, and the model is loaded, but the actual workload does not exceed 10 GB. It's kinda strange, but works now. Thank you!

Feel free to close this issue due to resolved problem, if you have no another advices.

nshmyrev commented 2 years ago

Same as #77