search

srvk / eesen

The official repository of the Eesen project
http://arxiv.org/abs/1507.08240
Apache License 2.0
822 stars 343 forks source link

Fix filenames in train_ctc_tf_ml.sh #170

Closed xinjli closed 6 years ago

xinjli commented 6 years ago

Fix some filename issues in tf_clean steps/train_ctc_tf_ml.sh, so that tf babel script can run without crashing

xinjli commented 6 years ago

I see, I was trying to make the script work for babel dataset and did not consider swbd case. Do you mean that I can push the first two changes: change_dev into _cv and _nohup into _tr, and should not touch the remaining part? Thanks!

fmetze commented 6 years ago

We also have the scripts in https://bitbucket.org/mgowayyed/lorelei-audio https://bitbucket.org/mgowayyed/lorelei-audio which was what we used in last year’s lorelei evaluation. Not sure how they are different - this implementation here was originally provided by Siddharth or Ramon?

On Jan 31, 2018, at 12:33 AM, Xinjian Li notifications@github.com wrote:

I see, I was trying to make the script work for babel dataset and did not consider swbd case. Do you mean that I can push the first two changes: change_dev into _cv and _nohup into _tr, and should not touch the remaining part? Thanks!

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/srvk/eesen/pull/170#issuecomment-361828821, or mute the thread https://github.com/notifications/unsubscribe-auth/AEnA8fPtnOJRKTzxOLTZHR9tX-vVIJjNks5tP_ssgaJpZM4RzIUC.

ramonsanabria commented 6 years ago

I don't have access to those repos. I was thinking to do a general script that could accept babel and swbd. But I thought about it and lets merge it. We can change the data preparation scripts for swbd when we want to do multi target.

2018-01-31 10:02 GMT-05:00 Florian Metze notifications@github.com:

We also have the scripts in https://bitbucket.org/mgowayyed/lorelei-audio https://bitbucket.org/mgowayyed/lorelei-audio which was what we used in last year’s lorelei evaluation. Not sure how they are different - this implementation here was originally provided by Siddharth or Ramon?

On Jan 31, 2018, at 12:33 AM, Xinjian Li notifications@github.com wrote:

I see, I was trying to make the script work for babel dataset and did not consider swbd case. Do you mean that I can push the first two changes: change_dev into _cv and _nohup into _tr, and should not touch the remaining part? Thanks!

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub < https://github.com/srvk/eesen/pull/170#issuecomment-361828821>, or mute the thread https://github.com/notifications/unsubscribe-auth/ AEnA8fPtnOJRKTzxOLTZHR9tX-vVIJjNks5tP_ssgaJpZM4RzIUC.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/srvk/eesen/pull/170#issuecomment-361958752, or mute the thread https://github.com/notifications/unsubscribe-auth/AMlwPZUqJObUMMA_WyZsYv9b3viMw-d_ks5tQIBygaJpZM4RzIUC .

xinjli commented 6 years ago

OK, thanks!