Can the model definition of DeepSpeech be found in the Github repository?

Hyno · April 1, 2020, 1:48pm

For a research project, there is a desire to experiment with the DeepSpeech model by adding custom layers / manipulate weights and measure performance differences. So far it has not been managed to find the necessary code to do this in the Github repository.

Can the code that defines the model and the weights of the trained model be found in the Github repository, preferably in Python? If so, where are they located.

Thanks in advance!

lissyx · April 1, 2020, 2:17pm

Did you really search ?

https://deepspeech.readthedocs.io/en/latest/DeepSpeech.html

github.com/mozilla/DeepSpeech

training/deepspeech_training/train.py

0cc815f1f


      
          def create_model(batch_x, seq_length, dropout, reuse=False, batch_size=None, previous_state=None, overlap=True, rnn_impl=rnn_impl_lstmblockfusedcell):
              layers = {}
          
              # Input shape: [batch_size, n_steps, n_input + 2*n_input*n_context]
              if not batch_size:
                  batch_size = tf.shape(input=batch_x)[0]
          
              # Create overlapping feature windows if needed
              if overlap:
                  batch_x = create_overlapping_windows(batch_x)
          
              # Reshaping `batch_x` to a tensor with shape `[n_steps*batch_size, n_input + 2*n_input*n_context]`.
              # This is done to prepare the batch for input into the first layer which expects a tensor of rank `2`.
          
              # Permute n_steps and batch_size
              batch_x = tf.transpose(a=batch_x, perm=[1, 0, 2, 3])
              # Reshape to prepare input for first layer
              batch_x = tf.reshape(batch_x, [-1, Config.n_input + 2*Config.n_input*Config.n_context]) # (n_steps*batch_size, n_input + 2*n_input*n_context)
              layers['input_reshaped'] = batch_x

This file has been truncated. show original

Hyno · April 14, 2020, 7:35pm

Thanks for your reply , @lissyx!

Topic		Replies	Views
Load model, get layer_5 and add extra layers DeepSpeech	4	407	December 3, 2020
Details of DeepSpeech architect DeepSpeech	2	913	April 28, 2019
DeepSpeech 2 based models with CNN and multiple RNN layers with BatchNorm DeepSpeech	0	564	July 27, 2020
DeepSpeech Training own English model for call center speech recognition DeepSpeech	22	3254	October 8, 2019
Adding an additional layer to DeepSpeech model DeepSpeech	7	487	March 24, 2020

Can the model definition of DeepSpeech be found in the Github repository?

Related topics