Hi 
Using the below command:
DeepSpeech % python3.7 util/taskcluster.py --source tensorflow --artifact convert_graphdef_memmapped_format --branch r1.15 --target .
Downloading https://community-tc.services.mozilla.com/api/index/v1/task/project.deepspeech.tensorflow.pip.r1.15.osx/artifacts/public/convert_graphdef_memmapped_format ...
I get the below error:
urllib.error.HTTPError: HTTP Error 404: Not Found
macOS Big Sur (v 11.2.1)
Can anyone help me with this issue?
I think DeepSpeech 0.9.3 is based on Python 3.6, which might be incompatible. I’ve found the pre-requisites to be very specific for DeepSpeech, and fiddly to get right.
There are Docker containers described in the Playbook, which might be easier to get going: https://mozilla.github.io/deepspeech-playbook/
And if you just want to execute some commands, you can try Google Colab. This is my notebook for doing some training, you could pick it apart and just use the bits you need.
I fear the artifact just expired …
@pkisiel Unfortunately, between the moment we worked on that r1.15 and the time it expired (and now) there has been some changes to the CI that makes it mostly impossible to rebuild. So you would need to re-build it yourself, sadly.
See the task group I re-run a few months ago to re-generate the binary for training-supported platforms (i.e., linux) https://community-tc.services.mozilla.com/tasks/groups/T8rU9_XlRIGe1KhZOONmZg
It should be just about following TensorFlow r1.15 build setup and then running a bazel build -c opt //tensorflow/contrib/util:convert_graphdef_memmapped_format
If you use the DeepSpeech Playbook Environment instructions to pull down a Docker image, then there is a worked example of creating a memmapped file in the Deployment section