search

keunwoochoi / music-auto_tagging-keras

Music auto-tagging models and trained weights in keras/theano
MIT License
616 stars 142 forks source link

problem with the model generation #31

Closed emmanuelbernieri closed 7 years ago

emmanuelbernieri commented 7 years ago

when I run the example_tagging an error occurs in the MusicTaggerCNN:

ValueError: Dimension 0 in both shapes must be equal, but are 96 and 1366 for 'Assign' (op: 'Assign') with input shapes: [96], [1366].

How can we solve this problem ?

Kind regards

Emmanuel

keunwoochoi commented 7 years ago

Could you check #22?

emmanuelbernieri commented 7 years ago

Ok I see,

In that case could you please explain to me on the compact cnn. How do you transform a .wav or .mp3 file into the 1100103.clip.npy ?

Kind regards

keunwoochoi commented 7 years ago

Just pushed an example: https://github.com/keunwoochoi/music-auto_tagging-keras/blob/master/compact_cnn/prepare_audio.py . Basically it's decoding, resampling, downmixing, and trimming.

emmanuelbernieri commented 7 years ago

it's perfectly working 👍

thank you very much