Reuben Morais
Reuben Morais
@derekdreery we haven't seen this even when using really long audios as input (hours). If you can share a script to reproduce the problem it'd greatly help in fixing it.
@derekdreery on the issue description you said you used some freely available data to reproduce it? Can you share that?
We can't always know the dataset size, for example with the recently introduced WebDataset input. I guess we could save the total after the first epoch and then use it...
bazel-zig-cc uses [platforms](https://docs.bazel.build/versions/main/platforms-intro.html) while TensorFlow is still using `--crosstool_top` and `--cpu`, so we need to wait until TensorFlow upgrades to be able to leverage it.
We don't have pre-built binaries for FreeBSD. Patches are welcome! The CI builds are defined in https://github.com/coqui-ai/STT/blob/main/.github/workflows/build-and-test.yml
Thanks for opening! Did you also make parallel changes to the PathTrie to go with this scoring change here? Could you share them as well so we can have the...
The error message indicates you don't have a working setup of CUDA and/or cuDNN: ``` tensorflow.python.framework.errors_impl.UnknownError: Fail to find the dnn implementation. [[{{node tower_0/cudnn_lstm/cudnn_lstm/CudnnRNNCanonicalToParams}}]] ```
Can you try with the training Docker image to double check it's not related to the CUDA setup?
PRs welcome! I basically never use `make install` so I'm not the best one to fix this :)
> We have also built it ourselves Pull requests would be greatly appreciated :)