Deepspeech: Error: Trie file version mismatch (4 instead of expected 3). Update your trie file.

[alchemi5t]

• Have I written custom code (as opposed to running examples on an unmodified clone of the repository): No
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Ubuntu 18.04
• TensorFlow installed from (our builds, or upstream TensorFlow): pip installed pre-compiled binary for 1.13.1
• TensorFlow version (use command below):'1.13.1'
• Python version: 3.6.8
• Bazel version (if compiling from source):
• GCC/Compiler version (if compiling from source):
• CUDA/cuDNN version:10
• GPU model and memory:Titan V
• Exact command to reproduce: CUDA_VISIBLE_DEVICES=0 python DeepSpeech.py --train_files ../et/clips/train.csv --dev_files ../et/clips/dev.csv --test_files ../et/clips/test.csv

I have git-lfs and have cloned the repo with it. I am on the master branch and the trie and lm have been cloned with git-lfs. How would i go about fixing this?

[kdavis-mozilla]

Does your master include the commit 0ea5804?

[alchemi5t]

@kdavis-mozilla yes. Let me explain what all i've done.

1) git lfs clone
2) have all the pre-reqs
3) preprocessed the datasets
4) run "CUDA_VISIBLE_DEVICES=0 python DeepSpeech.py --train_files ../et/clips/train.csv --dev_files ../et/clips/dev.csv --test_files ../et/clips/test.csv"

[lissyx]

@kdavis-mozilla yes. Let me explain what all i've done.

1. git lfs clone

2. have all the pre-reqs

3. preprocessed the datasets

4. run "CUDA_VISIBLE_DEVICES=0 python DeepSpeech.py --train_files ../et/clips/train.csv --dev_files ../et/clips/dev.csv --test_files ../et/clips/test.csv"

Can you check ?

$ sha1sum data/lm/lm.binary data/lm/trie 
098c71d7f720ff1842650881e2e3ebb9d18bd7d3  data/lm/lm.binary
5f2c59596a977779f27869768e544c82e840290f  data/lm/trie

[alchemi5t]

@kdavis-mozilla
$ sha1sum lm.binary trie

098c71d7f720ff1842650881e2e3ebb9d18bd7d3 lm.binary
5f2c59596a977779f27869768e544c82e840290f trie

[lissyx]

@alchemi5t What's your ds_ctcdecoder version? pip list

[alchemi5t]

@lissyx ds-ctcdecoder==0.6.0a0

[lissyx]

@lissyx ds-ctcdecoder==0.6.0a0

Ok, can you pick the wheel matching your python version on https://tools.taskcluster.net/index/project.deepspeech.deepspeech.native_client.master/cpu-ctc and retry ?

[lissyx]

pip install --upgrade <URL> with <URL> being one of those package on https://tools.taskcluster.net/index/project.deepspeech.deepspeech.native_client.master/cpu-ctc

[alchemi5t]

@lissyx got it. on it. will reply as soon as i am done testing.

[alchemi5t]

HOLY $*!! @lissyx that actually worked! Thanks a ton, mate!

[lissyx]

Thanks, so we need to release a new alpha then.

[alchemi5t]

yep!
pip install --upgrade https://index.taskcluster.net/v1/task/project.deepspeech.deepspeech.native_client.master.cpu-ctc/artifacts/public/ds_ctcdecoder-0.6.0a0-cp36-cp36m-manylinux1_x86_64.whl

for python 3.6 if anyone needs it.

[lissyx]

yep!
pip install --upgrade https://index.taskcluster.net/v1/task/project.deepspeech.deepspeech.native_client.master.cpu-ctc/artifacts/public/ds_ctcdecoder-0.6.0a0-cp36-cp36m-manylinux1_x86_64.whl

for python 3.6 if anyone needs it.

Thanks, I just pushed the v0.6.0-alpha.1 tag so it should be properly handled as soon as the builds finished.

[lissyx]

Builds finished, it should be transparent now following the docs. Thanks @alchemi5t for reporting.

[lock[bot]]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.