vagrawal / deepsphinx Goto Github PK

Hello,

In the closing response to issue#3, it is written that deepsphinx was written for python 2.
Tensorflow, however, requires python 3 for installing on Windows.

Has anyone tried to install deepsphinx on windows?

May I also add that the backslashes in the Windows path may interfere
with the input line structure.

Thanks,
Yuval

Hello!

Thank you for all the improvements and documentation.

I installed in Ubuntu
sudo python3 setup.py build
sudo python3 setup.py install

then issued the command:

shell: python3 bin/deepsphinx-infer --checkpoint-path data/batch-21937.data-00000-of-00001 --audio-file ~/speech/23-00-25_217c.wav

There was an error - see below:

Traceback (most recent call last):
File "bin/deepsphinx-infer", line 5, in
from deepsphinx.api import Predict
File "/usr/local/lib/python3.5/dist-packages/deepshinx-0.1-py3.5.egg/deepsphinx/api.py", line 7, in
ImportError: No module named 'flags'

Although the module was in the "build"

shell: find . -name flags*
./build/lib/deepsphinx/flags.py
./deepsphinx/flags.py

The same error occurs also with python3 bin/deepsphinx-infer --help and on Windows.

Thanks,
Yuval

I am a newbie exploring attention based models and your work has been of great help in understanding some existing architectures. I would be grateful if you could put up some information/guidelines on how to use lm with a tf model in your code. (like what input format is expected or how to use make_fst to construct one)
Thanks
Best Regards

Hello!

Thanks for making the project compatible with python-3 !

In setup.py, change the script name from 'bin/deepsphinx_train' to
'bin/deepsphinx-train' .

Could you please provide basic examples of script invocations for training and inference?

I ran
python bin\deepsphinx-train --trans-file data\ds-input.txt --nouse-train-lm --job-dir data\train5
and got a division by zero, something was probably wrong in the call.

Thanks,
Yuval

INFO:tensorflow:Getting speaker stats
INFO:tensorflow:Starting training
INFO:tensorflow:Epoch completed, saving
INFO:tensorflow:Evaluation started
INFO:tensorflow:Restoring parameters from data\elta\train55\checkpoints/-0
Traceback (most recent call last):
File "bin\deepsphinx-train", line 270, in
tf.app.run(train)
File "C:\Users\user\AppData\Local\Programs\Python\Python35\lib\sitepackages\tensorflow\python\platform\app.py", line 48, in run_sys.exit(main(_sys.argv[:1] + flags_passthrough))
File "bin\deepsphinx-train", line 153, in train
lm_fst)
File "bin\deepsphinx-train", line 79, in run_eval
tot_wer / tot_ev, tot_cer / tot_ev))
ZeroDivisionError: float division by zero

Hi,
I have checked your repository and willing to do some experiment with it. But as you have mentioned I need to train it first and then use it. If you have a well trained model then please add it to your READ.ME file so that it will be accessible to all.

Else share the training data repository, so that I can do the training with it. I have found something in the tensorflow repository, where the datasets size is 100 GB or 21 GB https://github.com/pannous/tensorflow-speech-recognition#ultimate-goal
I want to confirm whether this is the training data that you used for the training purpose or is there something else. If you have used the training data then how you have utilized it.
Kindly, share your thoughts regarding it.

Hello!

I have trained on the same data from windows and from unix, the data contain 1293 wav files.
I did not create an FST or a language model. A checkpoint directory was created and I used it for
inference.

Surprisingly, when running inference on one of the wav files in the training data - the result was blank.
Could it be related to the lack of language model?

Here is the unix training run, the windows training run gave the same messages (but different files).

Thanks,
     Yuval

shell:~/speech/deepsphinx$ python3 bin/deepsphinx-train --job-dir data --trans-file data/ds-input-unix.txt --nouse-train-lm --batch-size 1293

INFO:tensorflow:Getting speaker stats
INFO:tensorflow:Starting training
INFO:tensorflow:Epoch completed, saving
INFO:tensorflow:Evaluation started
INFO:tensorflow:Restoring parameters from data/checkpoints/batch-0
Traceback (most recent call last):
File "bin/deepsphinx-train", line 248, in
tf.app.run(train)
File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/platform/app.py", line 48, in run
_sys.exit(main(_sys.argv[:1] + flags_passthrough))
File "bin/deepsphinx-train", line 154, in train
lm_fst)
File "bin/deepsphinx-train", line 80, in run_eval
tot_wer / tot_ev, tot_cer / tot_ev))
ZeroDivisionError: float division by zero

Hello Vishal,
I have the following issue while running the program and I am not able to resolve it. Please help me:

>>> from deepsphinx.api import Predict
>>> ds = Predict(Predict.default_flags(),'batch-21937.data-00000-of-00001')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/aims-whm/deepsphinx-master/deepsphinx/api.py", line 24, in __init__
    1.0)
  File "/home/aims-whm/deepsphinx-master/deepsphinx/seq2seq_model.py", line 241, in seq2seq_model
    keep_prob)
  File "/home/aims-whm/deepsphinx-master/deepsphinx/seq2seq_model.py", line 51, in encoding_layer
    dtype=tf.float32)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/ops/rnn.py", line 396, in bidirectional_dynamic_rnn
    seq_dim=time_dim, batch_dim=batch_dim)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/ops/rnn.py", line 389, in _reverse
    seq_dim=seq_dim, batch_dim=batch_dim)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/ops/array_ops.py", line 2355, in reverse_sequence
    name=name)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/ops/gen_array_ops.py", line 2633, in reverse_sequence
    batch_dim=batch_dim, name=name)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/framework/op_def_library.py", line 589, in apply_op
    param_name=input_name)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/framework/op_def_library.py", line 60, in _SatisfiesTypeConstraint
    ", ".join(dtypes.as_dtype(x).name for x in allowed_list)))
TypeError: Value passed to parameter 'seq_lengths' has DataType float64 not in list of allowed values: int32, int64

How big is the model you've trained, can we use him simply