search

louiskirsch / speechT

An opensource speech-to-text software written in tensorflow
Apache License 2.0
157 stars 36 forks source link

Problem while testing the code #2

Closed arpit601 closed 7 years ago

arpit601 commented 7 years ago

File "/home/arpit_agrawal/test/speechT/speecht/evaluation.py", line 143, in run_epoch decoded_ids = next(decoded_path) StopIteration I am getting this stopIteration error

arpit601 commented 7 years ago

??? @timediv need your help urgently for this issue....

louiskirsch commented 7 years ago

Is this happening with your adapted code or the master version?

arpit601 commented 7 years ago

master version

arpit601 commented 7 years ago

I haven't changed anything in evaluation files. I changed only corpus.py and preprocessing.py and I was able to train the model .

arpit601 commented 7 years ago

I think its generating only one decoded path from greeedy_search_decoder
there is some problem with this part of the code too.. for decoded_path in decoded_ids_paths:
decoded_ids = next(decoded_path.values)

arpit601 commented 7 years ago

@timediv Did you find any error or is it running for you ?

louiskirsch commented 7 years ago

I never had this issue and don't know what of your changes could have caused this right now. It basically must mean there is no decoded path in the result from the greedy decoding. I'll let you know if I can think of something.

arpit601 commented 7 years ago

@timediv copying a part of your evaluation.py for decoded_path in decoded_ids_paths: decoded_ids = next(decoded_path) I think there is an issue with next(decoded_path) . There is no point running next(decoded_path) when that is the same variable iterating in the outer loop. Is it next(decoded_path) or smthng else ? Also right now ,decoded_path has one path that it is giving.

ghost commented 7 years ago

@timediv even I am facing the same issue

louiskirsch commented 7 years ago

Okay that's odd. I will look into this.

arpit601 commented 7 years ago

@timediv Hey Louis !!! Thanks for giving a prompt reply. I have been able to solve this problem . This error occurred due to insufficient training .

louiskirsch commented 7 years ago

@arpit601 Interesting. Can you tell me how I can reproduce the problem. For how long have you trained so that the problem occurred, for how long for the problem to disappear?

louiskirsch commented 7 years ago

@arpit601 and @pdcoded I added evaluation details and downloadable pretrained weights to the README!