Can i use pre trained model with deepspeech py

[JRMeyer]

>>> Kim
[December 18, 2017, 8:32am]

Hi,

Please help me figure out how to use the pre-trained model with
DeepSpeech.py.

I tried to pass the files from pre-trained models to
DeepSpeech.py's corresponding arguments:

> --decoder_library_path slash
> --alphabet_config_path slash
> --lm_binary_path slash
> --lm_trie_path

But I got error saying:

> tensorflow.python.framework.errors_impl.NotFoundError:
> deepspeech-0.1.0-models/models/output_graph.pb: invalid ELF header

I'm not sure how to fix this.

The pre-trained models worked fine with deepspeech.commands sorted-deepspeech-posts sorted-deepspeech-posts~ command line
tool.

My goal is to get the output matrix instead of the text output, is there
a work around for this?

Thank you.

[This is an archived TTS discussion thread from discourse.mozilla.org/t/can-i-use-pre-trained-model-with-deepspeech-py]

[JRMeyer]

>>> kdavis
[December 18, 2017, 9:44am]

To clarify what do you mean by 'output matrix'?

[Archived Post]

[JRMeyer]

>>> reuben
[December 18, 2017, 10:32am]

You can't use the released model with DeepSpeech.py unless you made
significant changes to it. There's no command line parameter that
expects a frozen model to be passed in. From the error, it looks like
maybe you passed --decoder_library_path models/models/output_graph.pb,
when you should be passing the path to libctc_decoder_with_kenlm.so to
it.

[Archived Post]

[JRMeyer]

>>> yv001
[December 18, 2017, 11:25am]

My guess would be that Kim is looking for the probabilities of each
character rather than the most likely character (before the language
model is applied), something like this
https://cdn-images-1.medium.com/max/1000/1*d1ktMdOnFOJRKKyjFP6sqQ.png.

If that's the case, I'd be interested in it too (e.g. as a param of the
commandline deepspeech).

[Archived Post]

[JRMeyer]

>>> Kim
[December 18, 2017, 9:28pm]

Sorry about the confusion. Like yv001 said, by output matrix I meant the
probabilities of each character.

[Archived Post]

[JRMeyer]

>>> kdavis
[December 19, 2017, 10:28am]

This would require writing a custom C++ operator that modifies our
current custom C++ operator for CTC
decoding slash [1 slash ]
and would be a significant amount of work.

[Archived Post]

[JRMeyer]

>>> reuben
[December 19, 2017, 10:42am]

You don't necessarily have to use DeepSpeech.py
to access the logits, you could also write some code to load the release
model using the TensorFlow API, then fetch the 'logits' node. You could
try modifying this code to load from the frozen graph instead of
restoring a checkpoint:
https://github.com/mozilla/DeepSpeech/blob/99d0c311a3e6108ed835fa38e4680ba7b3744fad/DeepSpeech.py#L1748-L1775

The relevant TensorFlow API is tf.import_graph_def.

[Archived Post]

[JRMeyer]

>>> yv001
[January 9, 2018, 10:16am]

A debugging script that shows most likely characters and their
probabilities can be found here: slash
https://github.com/pvanickova/DeepSpeech/blob/master/bin/show_inferred_characters.py

It loads a frozen graph, runs inference on given wav, model and
alphabet, softmaxes the logits to get the probabilities and displays
most likely characters and their probabilities, '-' is used for blank.
Each character prediction is on a new line so the predicted text can be
read in columns.

E.g. 'cat' string could have these character predictions: slash
c k - (0.999957) (1.62438e-05) (6.99057e-06) slash
a - (0.999998) (1.05044e-06) (4.43978e-07) slash
t d - (0.999999) (4.27885e-07) (1.09088e-07)

This is how the script is run:

python3 ./bin/show_inferred_characters.py --input-file '../data/my.wav' --model-file ../data/models/output_graph.pb --alphabet-file ../data/models/alphabet.txt --predicted-character-count 3

If it looks useful enough for others, give me a shout and I'll create a
pull request.



[Output matrix from neural net



[Simple way to get at raw probabilities/logits via python
bindings?



[As of now, is Deep Speech viable for real-world
applications?

[Archived Post]

[JRMeyer]

>>> yv001
[December 27, 2019, 3:12pm]

The script above has been updated to work with 0.6.0 frozen deepspeech
pb models.

[Archived Post]