Reuben Morais

Results 199 comments of Reuben Morais

After manually patching the code to replace `cpp.tensor-flow` with `cpp.tensorflow` I ran into the same problem but this time with boost.Program_Options: https://storage.googleapis.com/sourcerer-app/library-models/v2/cpp/cpp.boost-program_options.pb

How do I rebuild the protobufs? I ran into this problem with a release version of the app, installed via `curl -s https://sourcerer.io/app/install | bash`, the only reason why I...

And yes, something *is* wrong with the protobuf, it's trying to download a protobuf from https://storage.googleapis.com/sourcerer-app/library-models/v2/cpp/cpp.tensor-flow.pb and getting a 404 error. Instead of giving up, it tries to parse the...

@yaronskaya successfully got past the `cpp.tensorflow` problem, now getting a similar error with `cpp.boost-program_options`: ``` [i] Downloading cpp.boost-program_options classifier. [i] Finished downloading cpp.boost-program_options classifier. [i] Loading cpp.boost-program_options evaluator. [e] Hashing...

The acoustic model output has the alphabet embedded in it (number of predicted classes) so you can't use a different alphabet just on the decoder end. For doing transfer learning...

@dabinat this bug is for using different language models, not acoustic models. The idea is that you can instantiate multiple language model instances and then for every new stream you...

In any case, the rough idea I had for how to implement/expose this is by separating language model instances into their own handle, so `DS_EnableDecoderWithLM` would turn into `DS_CreateDecoderWithLM`, which...

I missed the fact that this was in 1.0, it involves more changes than the ones currently in #2681. It shouldn't be too difficult, but it will complicate the API...

@MatthewWaller a preliminary model can be found here: https://github.com/reuben/DeepSpeech/releases/tag/v0.0.1-alpha

That blacklist doesn't remove `previous_state_{c,h}`, but rather makes the freezing process ignore them, since I want them to be variables (not constants) in the final exported graph. The idea is...