search

netpi / compound-word-transformer-tensorflow

AI 音乐 - compound-word-transformer，用 Tensorflow 实现
https://eurychen.me/post/music/ai-compose-music/
138 stars 24 forks source link

WARNING:tensorflow:Detecting that an object or model or tf.train.Checkpoint is being deleted with unrestored values #3

Open dedededefo opened 2 years ago

dedededefo commented 2 years ago

Hello! I'm here again! When I used the trained music to generate music alone, this warning appeared when loading the model. Although there are only warnings and music can be generated, the music quality is very poor. Is it because of these warnings. Or is it because my model is not well trained? Warning such like this--> (WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer's state 'v' for (root).enc_layers.11.mha.wv.kernel) (LQ}WGTS%)GV Y}8IQW%4WA

netpi commented 2 years ago

@dedededefo The picture is about weights-loaded warnning. Poor quality midi is a model-trained problem.

xuboot commented 1 year ago

@netpi Hello, I would like to know which of the current projects MusicTransformer-Pytorch and Optimus-VIRTUOSO, compound-word-transformer-tensorflow is better in music generation or where you find the sample dataset. I would like to test or train your project to see the effect, I am a beginner, I don't know much at the moment, I would like to ask the big guys to guide me, thanks.