Pbmm file not generating

DeepSpeech
#1

trying to convert output_graph.pb into output_graph.pbmm ( after downloading “convert_graphdef_memmapped_format” from taskcluster.py) getting this error convert_graphdef_memmapped_format: command not found…!
any help…?

#2

Read the doc ? Graph the convert_graphdef_memmapped_format binary ?

#3

I am following the doc and all things are right just stuck on this step.
$ convert_graphdef_memmapped_format --in_graph=output_graph.pb --out_graph=output_graph.pbmm

#4

https://deepspeech.readthedocs.io/en/r0.9/TRAINING.html?highlight=convert_graphdef_memmapped_format#making-a-mmap-able-model-for-inference

You have not read them.

#5

If you searched you would have found convert_graphdef_memmapped_format on https://github.com/mozilla/DeepSpeech/releases/tag/v0.9.3 as well since TaskCluster is not used anymore and v0.9.3 artifacts have expired by now.

#6

For those still facing this little issue, I solved it like that:

  • provided the complete path to convert_graphdef_memmapped_format. How? I made sure I did everything right in the previous step by running the command again:

        python3 util/taskcluster.py --source tensorflow --artifact convert_graphdef_memmapped_format --branch r1.15 --target .

  • saw the console output:

      File already exists: /home/natalia/lab/DeepSpeech/convert_graphdef_memmapped_format

  • instead of “convert_graphdef_memmapped_format --in_graph=o …” provide the COMPLETE path, in my case:

      /home/natalia/lab/DeepSpeech/convert_graphdef_memmapped_format --in_graph=output_graph.pb --out_graph=output_graph.pbmm

Voila.

P.S. and yes, I have read the docs. And I also got stuck

2 Likes