Closed basnijholt closed 4 years ago
If you're using a virtual environment, make sure to do a pip3 install --upgrade -r requirements.txt
to upgrade the dependencies. Forgot to add this to the docs.
I have installed Rhasspy via Hass.io.
Oh, I didn't realize I am using a brand new feature :)
Ok, thanks. Something seems to be wrong with one of the pypi packages. Will get fixed soon.
Got problems also with docker container just updated:
While training:
[ERROR:603628] train: number_range_transform Traceback (most recent call last): File "/usr/share/rhasspy/rhasspy/train/init.py", line 233, in number_range_transform step = int(match.group(3)[1:]) TypeError: 'NoneType' object is not subscriptable [ERROR:603627] train: number_range_transform Traceback (most recent call last): File "/usr/share/rhasspy/rhasspy/train/init.py", line 233, in number_range_transform step = int(match.group(3)[1:]) TypeError: 'NoneType' object is not subscriptable
Here is the sentence for example: allume [dans] [le | la] ($house_room){house_room} [pendant | durant] (0..120){tempsminutes} minutes
tried this, does the same:
allume [dans] [le | la] ($house_room){house_room} [pendant | durant] (0..120){duration!int} minutes
Also, got 0..120 as unknown word.
Just pushed a fix. This is why I shouldn't try to code right before bed :man_facepalming:
Ahah !
Seems you can go to bed now ! In which country are you ? Often asked myself about your timezone lol
Updated, and my rhasspy batcher show now:
[ MATCHED] TimeTurnOnJeedom | query: allume dans la cuisine pendant douze minutes | confidence:1.0 | Slots: | duration_mins : 12 | house_room : cuisine [ MATCHED] TimeTurnOnJeedom | query: allume la cuisine pendant trente minutes | confidence:1.0 | Slots: | duration_mins : 30 | house_room : cuisine [ MATCHED] TimeTurnOnJeedom | query: allume le champi d'axel pendant dix minutes | confidence:1.0 | Slots: | device_name : champi_Axel | duration_mins : 10 [ MATCHED] TimeTurnOffJeedom | query: éteins dans la cuisine pendant douze minutes | confidence:1.0 | Slots: | duration_mins : 12 | house_room : cuisine [ MATCHED] TimeTurnOffJeedom | query: laisse le globe d'axel éteint pendant vingt minutes | confidence:1.0 | Slots: | device_name : globe_Axel | duration_mins : 20 [ MATCHED] TimeTurnOffJeedom | query: laisse la cuisine éteinte pendant 25 minutes | confidence:1.0 | Slots: | duration_mins : 25 | house_room : cuisine
matched: 6 | unmatched: 0 | total: 6
Good night dude!
I'm in the U.S., but I finished coding that last (broken) piece last night and forgot to fix it before pushing the new version this morning :/
I'm using docker and get this output
For this setence: setze Lautstärke auf 5 prozent with this ini file:
[SetVolume]
setze [die] Lautstärke auf (0..100){volume} [Prozent]
Even the token values are __convert__int
.
docker image version:
$ docker inspect b87dac217a9d | grep Image
"Image": "sha256:6932f518345479dde443b605a499af89347e3a6493f2a0acc0ed6fcacf4140df",
"Image": "synesthesiam/rhasspy-server:latest",
$ docker images | grep rhasspy
synesthesiam/rhasspy-server latest 6932f5183454 2 days ago 1.3GB
Weird. I just tried this on the de
profile and I get the expected output. Can you pull the latest image (baf9707bd412
), re-train with the "no-cache" option, and try again? Thanks.
I've updated but didn't help: Here is the full output:
{
"entities": [
{
"end": 93,
"entity": "volume",
"raw_end": 93,
"raw_start": 25,
"raw_value": "__convert__int __convert__int null __converted__int __converted__int",
"start": 25,
"value": "__convert__int __convert__int null __converted__int __converted__int"
}
],
"intent": {
"confidence": 0.86,
"name": "SetVolume"
},
"raw_text": "setze die lautstärke auf __convert__int __convert__int null __converted__int __converted__int prozent",
"raw_tokens": [
"setze",
"die",
"lautstärke",
"auf",
"__convert__int",
"__convert__int",
"null",
"__converted__int",
"__converted__int",
"prozent"
],
"siteId": "default",
"slots": {
"volume": "__convert__int __convert__int null __converted__int __converted__int"
},
"speech_confidence": 1,
"text": "setze die lautstärke auf __convert__int __convert__int null __converted__int __converted__int prozent",
"time_sec": 0.2532923221588135,
"tokens": [
"setze",
"die",
"lautstärke",
"auf",
"__convert__int",
"__convert__int",
"null",
"__converted__int",
"__converted__int",
"prozent"
],
"wakeId": ""
}
I've triggered this via the webinterface, put it into the textfield next to sentence.
What's also strange is that this sentence: setze die Lautstärke auf fünf Prozent is recognized as
"intent":
"entities":
"intent":
"confidence": 0.86
"name": "GetTemperature"
"raw_text": "wie ist die temperatur"
This seems to me like another bug, no?
I mean I've put text as input not a sound file.
And also with a confidence of 86%?
Here is my full ini file.
[SetVolume]
setze [die] Lautstärke auf (0..100){volume} [Prozent]
stelle [die] Lautstärke auf (0..100){volume} [Prozent]
[GetTime]
wie spät ist es
sag mir die Uhrzeit
[GetTemperature]
wie ist die Temperatur
wie (heiß | kalt) ist es
[GetGarageState]
ist das Garagentor (offen | geschlossen)
[ChangeLightState]
light_name = (Wohnzimmerlampe | Garagenlicht) {name}
light_state = (ein | aus) {state}
schalte (die | das) <light_name> <light_state>
Oh, I've changed the recognition from fuzzywuzzy to fsticuffs and not it works!
But is this expected that it only works with fsticuffs? I didn't find this in the docs.
@timaschew
But is this expected that it only works with fsticuffs? I didn't find this in the docs.
I don't think that this is expected behavior. This might be related to https://github.com/synesthesiam/rhasspy/issues/147#issuecomment-570987225.
I am trying to use the following sentence with number ranges:
However, when I say "set the volume to 20%", it gives me this as intent:
Do I need to activate something to make it work or is this a bug?