0

I'm trying to load .pbmm and .scorer models available in the Mozilla Deepspeech documentation.

The load_model() function of Tensorflow doesn't support this file type and throws this error:

OSError: Unable to open file (file signature not found).

I ultimately wish to get the hidden layer responses of the deepspeech model for a specific research purpose. It would be great if someone could tell me how to load this model or suggest other speech recognition models that could help solve this purpose of getting hidden layer responses. Thanks in advance.

Rodrigo Laguna
  • 1,796
  • 1
  • 26
  • 46
Kailash Lakshmikanth
  • 1
  • 1
  • Can you add a link to those files, docs and small code snipped with what you have tried? Also should add at least the tensorflow version – Rodrigo Laguna Aug 12 '22 at 11:04
  • This is the code snippet: from google.colab import drive drive.mount('/content/drive') import tensorflow as tf model1_file='/content/drive/MyDrive/deepspeech-0.9.3-models.pbmm' model2_file='/content/drive/MyDrive/deepspeech-0.9.3-models.scorer' loaded_model1 = tf.keras.models.load_model(model1_file) loaded_model2= tf.keras.models.load_model(model2_file) – Kailash Lakshmikanth Aug 12 '22 at 11:16
  • please, update the question with this data to let others help you – Rodrigo Laguna Aug 12 '22 at 14:34

1 Answers1

0

This section of the DeepSpeech PlayBook has more information on model deployment, and contains information on how to convert .pbmm files to other formats.

Kathy Reid
  • 575
  • 4
  • 6