By “installation pattern”, I mean the sequence of steps I took to install things and get them to work. In the link above, I go through the installation sequence I tried initially, based on the standard readme.md, vs. the alternative steps I used that worked for me.
Summarizing just the steps that worked for me:
(1) Git clone https://github.com/mozilla/DeepSpeech
(2) Gownload pip wheel manually from here: https://pypi.python.org/pypi/deepspeech/0.1.0
(3) Move wheel file to place I want to keep it. CD to that location. Run pip install <name_of_wheel_file>
(4) Download the language model from the ‘releases’ part of the github project page here: https://github.com/mozilla/DeepSpeech/releases/download/v0.1.0/deepspeech-0.1.0-models.tar.gz.
(5) cd into data folder of DeepSpeech folder you cloned above, move tar.gz model file into there if not already
(6) Unpack that file using tar -xzf <name_of_.tar.gz> Again, I did it in the data folder of deepspeech repo
(7) From data folder, I ran in the command line deepspeech models/output_graph.pb smoke_test/LDC93S1.wav alphabet.txt
- note 1: depending where you unpacked things or where your wav file is located the exact file names and locations might be different.
- note 2: The readme has the pb file called output_models.pb while it is called output_graph.pb when I unpacked it as described above.