Closed ATriantafyllopoulos closed 1 month ago
@ureichel @maxschmitt does anybody of you have experience how to adapt configs to work with our Python version of opensmile
?
@ATriantafyllopoulos
For me, also the IS10_paraling
works, when adapting it in exactly the same way as you described for IS09_emotion
etc.
@maxschmitt thanks for checking! From python I only get the following error message:
Traceback (most recent call last):
File "/home/trianand/local-work/aibo-retro/export-models/test_smile.py", line 22, in <module>
smile = opensmile.Smile(
File "/home/trianand/local-work/-ModularAITaskTraining/.devenv/state/venv/lib/python3.10/site-packages/audobject/core/decorator.py", line 115, in wrapper
func(self, *args, **kwargs)
File "/home/trianand/local-work/-ModularAITaskTraining/.devenv/state/venv/lib/python3.10/site-packages/audeer/core/utils.py", line 192, in new_func
return func(*args, **kwargs)
File "/home/trianand/local-work/-ModularAITaskTraining/.devenv/state/venv/lib/python3.10/site-packages/opensmile/core/smile.py", line 171, in __init__
self._feature_names(),
File "/home/trianand/local-work/-ModularAITaskTraining/.devenv/state/venv/lib/python3.10/site-packages/opensmile/core/smile.py", line 330, in _feature_names
num_elements = smile.external_sink_get_num_elements(
File "/home/trianand/local-work/-ModularAITaskTraining/.devenv/state/venv/lib/python3.10/site-packages/opensmile/core/lib.py", line 440, in external_sink_get_num_elements
self._check_smile_result(
File "/home/trianand/local-work/-ModularAITaskTraining/.devenv/state/venv/lib/python3.10/site-packages/opensmile/core/lib.py", line 545, in _check_smile_result
raise OpenSmileException(result, message.decode("ascii"))
opensmile.core.lib.OpenSmileException: Code: 4, Message: specified component does not exist
Note that I have opensmile-v2.5.0
and python-v3.10.11
. The C++ opensmile
appears to be v3.27.4
The ERR
messages I get in the original issue only appear in the log
How do you come across version v3.27.4
of openSMILE? I was not aware that a version like this exists (latest official release is 3.0.2
).
Anyway, the fact that -sink
is not found seems to be the main issue for now. Could it be that there is some conflicting installation of openSMILE? (You are working on Ubuntu, right?)
Could you also send me the modified config file for IS10_paraling
to ensure that we are on the same page?
You are right about the C++ version, I have v3.0.2-1-g341aea9
, the other number came from my shell, which shows versions for stuff, but apparently got it wrong.
But that is for the code which I use to reference my configs (i.e. what I cloned locally), the binary packaged with the python wrapper (v2.5.0
) is v3.0.1
. However, you are also using the same setup as I am.
Attaching my config: IS10_paraling.zip
Sorry, I meant the modified config file for the feature set, not the package.
A misclick, sorry! Here's the correct zip
It seems option name is \{\cm[source{?}:include external sink]}
again in line 59
, but it should be \{\cm[sink{?}:include external sink]}
.
After changing that, it is working fine for me.
Ah, I'm an idiot. Of course. Sorry, copy-pasting sometimes get you..it works now!
Currently, the
opensmile-python
wrapper supports only a limited set of the original cofings. I have been trying to use the configs of the ComParE challenges.In general, it seems that one needs to substitute the "old" input/output configs with the newer format (input/output) supported in the
opensmile-wrapper
Essentially, this boils down to substituting:
\{../shared/standard_wave_input.conf.inc}
with\{\cm[source{?}:include external source]}
\{../shared/standard_data_output.conf.inc}
with\{\cm[sink{?}:include external sink]}
This works fine with the following configs:
{IS09_emotion, IS11_speaker_state, IS12_speaker_trait, IS13_ComParE.conf}
.However, when I try it on
IS10_paraling
, I getopensmile.core.lib.OpenSmileException: Code: 4, Message: specified component does not exist
. The onlyERR
messages in thelogfile
are:(ERR) [0] commandlineParser: parse: unknown option '-sink' on commandline!
(ERR) [0] commandlineParser: parse: unknown option '-sinkLevel' on commandline!
However, the other configs also sometimes show errors without failing to initialize the
smile_api
. Is there anything particular I should pay attention to for this config?