Want to train on new dataset over the Pretrained model provided by Deepspeech

DeepSpeech
#21

Sir I am getting the following error :

(deepspeech) ritish@ritish-VirtualBox:~/v_0.6.1_DeepSpeech$ git clone --branch v0.6.1 https://github.com/mozilla/DeepSpeech.git
Cloning into ‘DeepSpeech’…
remote: Enumerating objects: 38, done.
remote: Counting objects: 100% (38/38), done.
remote: Compressing objects: 100% (35/35), done.
remote: Total 17903 (delta 14), reused 14 (delta 3), pack-reused 17865
Receiving objects: 100% (17903/17903), 47.51 MiB | 771.00 KiB/s, done.
Resolving deltas: 100% (12157/12157), done.
Note: checking out ‘3df20fee52fda47d08e3726fd0da86dbb414e9d8’.

You are in ‘detached HEAD’ state. You can look around, make experimental
changes and commit them, and you can discard any commits you make in this
state without impacting any branches by performing another checkout.

If you want to create a new branch to retain commits you create, you may
do so (now or later) by using -b with the checkout command again. Example:

git checkout -b

#22

The cloning of the repot looks good, check with this that you are on 0.6.1

git describe --tags

The standard github link is always for the newest revision, the 0.6.1. documentation is here and repo here.

Not able to install ds_ctcdecoder required for training my own dataset
#23

If I understood well, you mean We can use checkpoints of specific version ( assume 0.X.X ) with that exact branch codes (0.X.X ).

So I can’t use my 0.7.3 (or any other) checkpoints in new cloned Deep speech version 0.8.1?

#24

In fact you can, because as mentionned in the release notes, 0.7 models are compatible with 0.8 (we made no change to the model, and it is the same tensorflow r1.15 release beign used for training)

2 Likes