No matching distribution found for tensorflow==1.15.2 (from deepspeech-training==0.7.3)

DeepSpeech
#1

I’ve been stumped at this error for quite a few days now. I’m following this guide: https://deepspeech.readthedocs.io/en/v0.7.3/TRAINING.html

When I do :

pip3 install --upgrade --force-reinstall -e .

It gives me the error:

ERROR: Could not find a version that satisfies the requirement tensorflow==1.15.2 (from deepspeech-training==0.7.3) (from versions: 2.2.0rc1, 2.2.0rc2, 2.2.0rc3, 2.2.0rc4, 2.2.0)
ERROR: No matching distribution found for tensorflow==1.15.2 (from deepspeech-training==0.7.3)

I tried doing this:

sudo apt-get install python3-dev

But it still gives me the same error. Somehow it works on google colab but I don’t want to work on colab because of my internet issues.

What am I missing? I’m using anaconda navigator python 3.6 environment, on Pop OS 20.04. If you need any other information, please tell me because I’m quite new to all this.

1 Like
#2

Same problem here. I am also using Ubuntu 20.04 and will try again with 18.04 if I don’t fine any other solution.

1 Like
#3

You just need to make sure you’re using Python 3.5, 3.6 or 3.7. TensorFlow 1.15 does not support Python 3.8: https://pypi.org/project/tensorflow/1.15.2/#files

1 Like
#4

Let me know if you get this to work on 18.04

#5

The environment I’m using is python 3.6.

#6

Please share verbose pip install. Make sure you try with vanilla virtualenv and not anaconda.

1 Like
#7

Ok I tried with a vanilla python3.6 virtualenv and the installation went fine. However:

./bin/run-ldc93s1.sh

Returns:

ValueError: Scorer initialization failed with error code 8198
swig/python detected a memory leak of type ‘Alphabet *’, no destructor found.

Does this indicate a problem with my installation or should I go ahead?

#8

There are two things in this message. The second is not important, the first is. Please ensure you have poperly setup git-lfs and that there’s a correct scorer file available. Error message says otherwise.

1 Like
#9

I’ve setup git-lfs again, just to be sure and did git clone. But now it gets stuck after:

Cloning into ‘DeepSpeech’…
remote: Enumerating objects: 115, done.
remote: Counting objects: 100% (115/115), done.
remote: Compressing objects: 100% (73/73), done.
remote: Total 18750 (delta 62), reused 60 (delta 36), pack-reused 18635
Receiving objects: 100% (18750/18750), 47.78 MiB | 96.00 KiB/s, done.
Resolving deltas: 100% (12750/12750), done.

I waited for like 15 minutes before Ctrl+C. And it gives:

^Cwarning: Clone succeeded, but checkout failed.
You can inspect what was checked out with ‘git status’
and retry with ‘git restore --source=HEAD :/’
Exiting because of “interrupt” signal.

I try git restore but it gets stuck again. Nevermind, I run the rest of the commands as usual and it goes smoothly until I run that script again and get the same error. Something’s wrong with cloning, I presume. Should I manually download the zip?

PS I should add that a friend had the identical problem when I asked him to try installing DeepSpeech for verification. He was on ubuntu 19.04. His git clone got stuck too.

#10

Could it be just your network that is slow ? 96kB/s the files that git-lfs downloads are ~900MB.

#11

Well, we don’t get much better in Pakistan. Any alternatives?

#12

I’m afraid there is no good solution here, except being patient.

1 Like
#13

Alright I’ll try git clone again and wait for … some time. Thank you for your help.

Just for confirmation, the scorer required by the script is downloaded with git clone and doesn’t need to be downloaded separately?

#14

git clone will rely on git-lfs yes, so you have no extra step after

1 Like
#15

What you could do is to clone everything on a Google colab and repackage everything without the scorer. This way you could get the code to train now and get the scorer later.

#16

I’m not sure if I understood properly. I cloned DeepSpeech repo on google colab and then zipped the folder and downloaded it. Now upon running this command locally on my computer:

pip3 install --upgrade --force-reinstall -e .

I get the error:

ERROR: Could not find a version that satisfies the requirement ds_ctcdecoder==training/deepspeech_training/VERSION (from deepspeech-training===training-deepspeech-training-VERSION) (from versions: 0.6.1, 0.7.0, 0.7.1, 0.7.3)
ERROR: No matching distribution found for ds_ctcdecoder==training/deepspeech_training/VERSION (from deepspeech-training===training-deepspeech-training-VERSION)

#17

I thought your problem was the large download. This way you can reduce it to a fraction and test on Google whether everything is working so you can replicate it on your system.

#18

I’m pretty sure the download size isn’t the issue. It’s probably a git problem. It gets stuck after resolving deltas. On Colab, I get the same error as above after running the pip install -e . command. In any case, I cloned the repo again, ctrl+c when it got stuck, tried running the ldc93s1 script which failed just like before. But other scripts in the bin/ folder are working so I’m assuming it’s a script problem.

#19

I have no idea what you do, but it runs perfectly on colab:

!curl -s https://packagecloud.io/install/repositories/github/git-lfs/script.deb.sh | sudo bash
!apt-get install git-lfs
!git clone --branch v0.7.3 https://github.com/mozilla/DeepSpeech
%cd DeepSpeech
!pwd
!pip3 install --upgrade pip==20.0.2 wheel==0.34.2 setuptools==46.1.3
!pip3 install --upgrade --force-reinstall -e .
!python3 util/taskcluster.py --target .

Check it and let us know what you did differently

1 Like
#20

I was doing only one thing differently: I didn’t specify the --branch argument in git clone. That has made all the difference. It installs correctly. The util/taskcluster.py throws a lot of http errors though.