Unable to fetch graph version

DeepSpeech
#1

Dear Support.

./…/DeepSpeech/deepspeech --model data/cv/output_graph.pbmm --lm data/cv/lm.binary --trie data/cv/trie --audio data/audio_test.wav

There is error
TensorFlow: v1.14.0-21-ge77504a
DeepSpeech: v0.6.1-0-g3df20fe
Unable to fetch graph version: Invalid argument: Session was not created with a graph before Run()!
Could not create model.

While, I have installed tensorflow-gpu==1.15 and it shows in pip3 list as well.
But, I think I am using old file of convert_graph.

I want to ask that can I get the link of convert_graphdef_memmapped_format for this above stated version ? or if I am making any nasty mistake while using?

Please help so that I can properly create the exported model for run.

#2

Can you tell where this model is coming from ? It’s obviously not a 0.6 one.

#3

The model I have self trained with the current deepspeech.

#4

Obivously you have improperly exported it.

#5

Could be. But the export is by the deepspeech run. If there is any way to resolve the error or the mistake I am making , you point out, it would be great help.

#6

That’s wrong, please use 1.14 as documented.

This is also documented how to use taskcluster.py for this.

#7

Hard to help when you don’t tell what you do exactly.

#8

https://deepspeech.readthedocs.io/en/v0.6.1/TRAINING.html#making-a-mmap-able-model-for-inference

#9

Great help. Thank you !!