Issue in Decoding

Deep Speech
#1

After training, I got this error which is even after validation.
Pre-processing of the test set was also done. But It gave me errors while decoding.
It also computed the acoustic model.

Traceback (most recent call last):
File “DeepSpeech.py”, line 942, in
tf.app.run(main)
File “/usr/local/lib/python3.6/dist-packages/tensorflow/python/platform/app.py”, line 125, in run
_sys.exit(main(argv))
File “DeepSpeech.py”, line 898, in main
test()
File “DeepSpeech.py”, line 659, in test
evaluate.evaluate(test_data, graph)
File “/home/cle-dl-1/DeepSpeech/DeepSpeech-0.4.1/evaluate.py”, line 179, in evaluate
num_processes=num_processes, scorer=scorer)
File “/usr/local/lib/python3.6/dist-packages/ds_ctcdecoder/init.py”, line 102, in ctc_beam_search_decoder_batch
for beam_results in batch_beam_results
File “/usr/local/lib/python3.6/dist-packages/ds_ctcdecoder/init.py”, line 102, in
for beam_results in batch_beam_results
File “/usr/local/lib/python3.6/dist-packages/ds_ctcdecoder/init.py”, line 101, in
[(res.probability, alphabet.decode(res.tokens)) for res in beam_results]
File “/home/cle-dl-1/DeepSpeech/DeepSpeech-0.4.1/util/text.py”, line 50, in decode
res += self.string_from_label(label)
File “/home/cle-dl-1/DeepSpeech/DeepSpeech-0.4.1/util/text.py”, line 29, in string_from_label
return self._label_to_str[label]
IndexError: list index out of range

errorbest
errorbest.png1920×1080 388 KB

#2

This is mismatching alphabet file: some of your data contains characters not in your alphabet file.

#3

Okay. Thank you! It has been resolved. But the problem now is, it doesn’t give me any output in decoding. I trained on almost 100 hours of data and longest transcription it has is almost 12 seconds.
I got the model, things work fine but it always giving me the blank result. Why is it so ?

#4

There could a lot of reasons, without more details on your training setup it’s hard to tell anything …

#5

@noor_e_emaan11 i am getting the same error can you tell me how did your problem resolved
i did not find any character in dataset which is not in alphabet,

#6

Check your flags.py, alphabet.txt file, put all the characters from train, test and dev even they are replicating.
This error occured randomly with me. I couldn’t specify one reason for it.
try with next iteration after checking flag.py, hopefully, It will resolve.

1 Like
#7

THANKS… my problem is solved, there was no ’ character in my alphabet.txt.

1 Like