search

HawkAaron / E2E-ASR

PyTorch Implementations for End-to-End Automatic Speech Recognition
126 stars 27 forks source link

feature_transform makes 69 dimension #5

Closed adonisues closed 5 years ago

adonisues commented 5 years ago

After "run.sh" and "feature_transform.sh", 69 dim features are given.

It makes error at 54 line in train_rnnt.py .

thank you

HawkAaron commented 5 years ago

@adonisues run.sh would give you 41 dim log fbank feature, could you please run feat-to-dim scp:data/train/feats.scp - to check the feature dimension ?

gccyxy commented 5 years ago

After "run.sh" and "feature_transform.sh", 69 dim features are given.

It makes error at 54 line in train_rnnt.py .

thank you I have met the same question, can you teach me how to solve it?

gccyxy commented 5 years ago

@HawkAaron I have met the same question, But i do not know what you mean ? can you explain it in detail?

gccyxy commented 5 years ago

@HawkAaron My shell is as followes:

[root@node02 s5]# feat-to-dim scp:data/train/feats.scp - feat-to-dim scp:data/train/feats.scp - 23

what can i do in the next?

HawkAaron commented 5 years ago

Please use this fbank.conf to extract acoustic feature if you use kaldi.

If not, please follow other recipes to do feature extraction, and then change the input dimension if necessary.