search

metal3d / docker-snips

Snips with pulseaudio to make it running on Linux
MIT License
3 stars 0 forks source link

Model Version Error #2

Open jhoelzl opened 5 years ago

jhoelzl commented 5 years ago

Hello,

when i start the container by your start script, i always get a model version error:

==> /var/log/supervisor/snips-nlu-stderr---supervisor-xGwu6D.log <== INFO:snips_nlu_hermes: loading nlu engine "/usr/share/snips/assistant/nlu_engine" ERROR:snips_nlu_hermes: nlu thread ended with error: Unable to read file '/usr/share/snips/assistant/nlu_engine/nlu_engine.json' -> caused by: Expected model version 0.17.0 but found 0.19.0

ERROR:snips_nlu : Fail to wait end of pipeline -> caused by: Unable to read file '/usr/share/snips/assistant/nlu_engine/nlu_engine.json' -> caused by: Expected model version 0.17.0 but found 0.19.0

ERROR:rumqtt::client : snips-nlu|23-0013a83cf3e2-1 Disconnected: (Error(SyncMpsc(Disconnected), State { next_error: None, backtrace: None }))

==> /var/log/supervisor/snips-tts-stderr---supervisor-m73csa.log <== INFO:snips_tts_hermes: starting with provider PicoTts

==> /var/log/supervisor/snips-nlu-stderr---supervisor-xGwu6D.log <== INFO:snips_nlu_hermes: loading nlu engine "/usr/share/snips/assistant/nlu_engine" ERROR:snips_nlu_hermes: nlu thread ended with error: Unable to read file '/usr/share/snips/assistant/nlu_engine/nlu_engine.json' -> caused by: Expected model version 0.17.0 but found 0.19.0

ERROR:snips_nlu : Fail to wait end of pipeline -> caused by: Unable to read file '/usr/share/snips/assistant/nlu_engine/nlu_engine.json' -> caused by: Expected model version 0.17.0 but found 0.19.0

ERROR:rumqtt::client : snips-nlu|46-0013a83cf3e2-1 Disconnected: (Error(SyncMpsc(Disconnected), State { next_error: None, backtrace: None })) INFO:snips_nlu_hermes: loading nlu engine "/usr/share/snips/assistant/nlu_engine" ERROR:snips_nlu_hermes: nlu thread ended with error: Unable to read file '/usr/share/snips/assistant/nlu_engine/nlu_engine.json' -> caused by: Expected model version 0.17.0 but found 0.19.0

ERROR:snips_nlu : Fail to wait end of pipeline -> caused by: Unable to read file '/usr/share/snips/assistant/nlu_engine/nlu_engine.json' -> caused by: Expected model version 0.17.0 but found 0.19.0

ERROR:rumqtt::client : snips-nlu|49-0013a83cf3e2-1 Disconnected: (Error(SyncMpsc(Disconnected), State { next_error: None, backtrace: None })) INFO:snips_nlu_hermes: loading nlu engine "/usr/share/snips/assistant/nlu_engine" ERROR:snips_nlu_hermes: nlu thread ended with error: Unable to read file '/usr/share/snips/assistant/nlu_engine/nlu_engine.json' -> caused by: Expected model version 0.17.0 but found 0.19.0

ERROR:snips_nlu : Fail to wait end of pipeline -> caused by: Unable to read file '/usr/share/snips/assistant/nlu_engine/nlu_engine.json' -> caused by: Expected model version 0.17.0 but found 0.19.0

What yo you recommend to overcome this issue? Regards, Josef

metal3d commented 5 years ago

I'm currently checking, Snips changed some things and I'm rebuilding Docker image to make some checks.

Thanks for the issue

metal3d commented 5 years ago

Could you please try to pull the latest version ?

docker pull metal3d/snips

And try again ?

lalten commented 5 years ago

I have a similar error:

==> /var/log/supervisor/snips-nlu-stderr---supervisor-98k_31.log <==
INFO:snips_nlu_hermes: loading nlu engine "/usr/share/snips/assistant/nlu_engine"
ERROR:snips_nlu       : Unable to read file '/usr/share/snips/assistant/nlu_engine/nlu_engine.json'
 -> caused by: Mismatched model version: model is 0.20.0 but runner is 0.19.0
metal3d commented 5 years ago

OK, it seems that Snips has changed assistant version. I re-downloaded my assistant and I've got the same error.

The main problem is that Snips changes model frequently, so I need to find a way to know that and to refresh my Docker image following the model version... not that easey :scream_cat:

I just rebuild Docker image and it's now working (excepting a non-blocking error when hermes-python is installed, but that's not from my side) - I added a label for snips version, but I probably need to setup label + model version in label. But, that's my side, for now the solution is simple: I rebuild image withtout any cache. So :arrow_right: Image is currently rebuilding on Docker Hub.

You should pull image again docker pull metal3d/snips or get the 1.3.0 tag docker pull metal3d/snips:1.3.0 and change the docker version in start.sh if needed.

The simplest is to use the latest image.

Can you please try ?