Error: Can't parse trie file, invalid header. Try updating your trie file

DeepSpeech
#1

Hey ,
Please I m trying to excute my deepspeech model after generating theses binaries : lm.binary, trie file
but I am getting this error.

Error: Can’t parse trie file, invalid header. Try updating your trie file.
terminate called after throwing an instance of ‘int’
Fatal Python error: Aborted

For lm.binary I followed this tutorial :

TUTORIAL : How I trained a specific french model to control my robot

For trie, I used this command line :

generate_trie alphabet.txt lm.binary trie

1 Like
#2

Hello.
Have a look into deepspeech dir.
You ll see new command line.
In Readme or

#3

Check the version of your generate_trie and the version of your python bindings, they need to match

#5

Thx for your reply
What is the new command line ?

#6

Thx for your reply
please How can I know the version of generate_trie

#7

From where did you download it should already give a good hint.

#8

I installed Mozilla’s Tensorflow r1.13 branch
Then I used Bazel to build the deepspeech library and the generate_trie binary
I downloaded DeepSpeech version 0.6.0-alpha.2
and I installed the deepspeech-gpu v0.5.1-0-g4b29b78

#9

I guess I have to use the tensorflow r1.14 branch . Am I right ?

#10

First, you don’t need to rebuild. Second you cannot mix v0.6 and v0.5.

Latest v0.6.0-alpha.3 does require r1.14.

#11

I can not find the deepspeech-gpu version 0.6

#12

and if you try harder, maybe ?

#13

Oh, we don’t have 0.6.0a3 here!

#14

There’s a bug when uploading on Pypi :confused:

And also on Github … https://github.com/mozilla/DeepSpeech/releases/tag/v0.6.0-alpha.3

#15

lol Ok you right,
I did what you told me I did not rebuild and I used just the pre-built binaries for the deepspeech by executing this command :

python3 util/taskcluster.py --target .

I installed the deepspeech-gpu v0.6.0-alpha.1
and I am using Deepspeech version 0.6.0-alpha.3
also I have tensorflow 1.14
but I can not train my model,
here is the error I get :

tensorflow.python.framework.errors_impl.InvalidArgumentError: Cannot assign a device for operation tower_0/DeserializeSparse: Could not satisfy explicit device specification ‘/device:GPU:0’ because no supported kernel for GPU devices is available

#16

I just saw ur reply…
and Yeah I noticed that

#17

We’re preparing alpha4 that should fix it. pip install deepspeech-gpu==0.6.0a4 should work soon now.

Well, again, mismatching. Please let us breathe and time to fix the issue you reported.

1 Like
#18

@yasine.nifa It’s on github and pypi now, so you should be able to use 0.6.0a4 everywhere. Thanks for noticing the issue !

1 Like
#19

Thx !! I will try it right now :slight_smile:

#20

I am getting this error :

tensorflow.python.framework.errors_impl.InvalidArgumentError: Cannot assign a device for operation tower_0/DeserializeSparse: Could not satisfy explicit device specification '/device:GPU:0' because no supported kernel for GPU devices is available.
Colocation Debug Info:
Colocation group had the following types and supported devices: 
Root Member(assigned_device_name_index_=-1 requested_device_name_='/device:GPU:0' assigned_device_name_='' resource_device_name_='' supported_device_types_=[CPU] possible_devices_=[]
DeserializeSparse: CPU 

Colocation members, user-requested devices, and framework assigned devices, if any:
  tower_0/DeserializeSparse (DeserializeSparse) /device:GPU:0

Op: DeserializeSparse
Node attrs: dtype=DT_INT32, Tserialized=DT_VARIANT
Registered kernels:
  device='CPU'; Tserialized in [DT_VARIANT]
  device='CPU'; Tserialized in [DT_STRING]

	 [[{{node tower_0/DeserializeSparse}}]]

Is it due to cuda version ?
by the way I followed all what you did in “https://github.com/mozilla/DeepSpeech”

Here is all the command I executed :

virtualenv -p python3 $HOME/tmp/deepspeech-venv/
source $HOME/tmp/deepspeech-venv/bin/activate
pip install deepspeech-gpu==0.6.0a4 
cd DeepSpeech
pip3 install -r requirements.txt
pip3 install $(python3 util/taskcluster.py --decoder)
pip3 uninstall tensorflow
pip3 install 'tensorflow-gpu==1.14.0'

And I have cuda10.0. Here is the output of nvcc -V :

nvcc: NVIDIA ® Cuda compiler driver
Copyright © 2005-2018 NVIDIA Corporation
Built on Sat_Aug_25_21:08:01_CDT_2018
Cuda compilation tools, release 10.0, V10.0.130

Then when I execute the command to train my model, I get the above error.

#21

That looks weird, maybe a regression on GPUs from switching to TensorFlow r1.14 :confused:

@reuben can you check on your side ?