Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Feature] Provide SavedModels of pretrained models #33

Closed
JeffreyCA opened this issue Nov 6, 2019 · 5 comments
Closed

[Feature] Provide SavedModels of pretrained models #33

JeffreyCA opened this issue Nov 6, 2019 · 5 comments
Labels
enhancement New feature or request feature New feature for Spleeter

Comments

@JeffreyCA
Copy link

Can you provide SavedModels of pretrained models (with the saved_model.pbfile)?

@JeffreyCA JeffreyCA added enhancement New feature or request feature New feature for Spleeter labels Nov 6, 2019
@bboettcher3
Copy link

Related to this, do you have any plans to convert the model to a TensorflowJS model? Having a Javascript counterpart would be pretty useful for embedding Spleeter in other applications.

@JeffreyCA
Copy link
Author

Update: SavedModels are already currrently being outputted to /tmp/serving after running a spleeter separation.

You can change this output directory via the DEFAULT_EXPORT_DIRECTORY constant in estimator.py.

@Faylixe
Copy link
Collaborator

Faylixe commented Nov 14, 2019

Related to this, do you have any plans to convert the model to a TensorflowJS model? Having a Javascript counterpart would be pretty useful for embedding Spleeter in other applications.

This will not happening from our side, the only migration we would consider at some point is porting library with TensorFlow 2. Plus I had checked a little bit before releasing and it would be very complicated as TensorFlow JS does not provide the range of functionality Spleeter is currently using in Python (estimator API to be more specific)

But if you are motivated feel free to explore, we would love having other language adoption from contributors.

Update: SavedModels are already currrently being outputted to /tmp/serving after running a spleeter separation.

You can change this output directory via the DEFAULT_EXPORT_DIRECTORY constant in estimator.py.

I will close the issue since you seem to have figured out how but we will also upload those model on the release soon.

@Faylixe Faylixe closed this as completed Nov 14, 2019
@shoegazerstella
Copy link

@JeffreyCA did you succeed in properly convert the SavedModes into tf.js?
I used the method explained here but had these warnings during conversion:

WARNING:tensorflow:Didn't find expected Conv2D or DepthwiseConv2dNative input to 'batch_normalization_8/cond/FusedBatchNorm'
WARNING:tensorflow:Didn't find expected Conv2D or DepthwiseConv2dNative input to 'batch_normalization_20/cond/FusedBatchNorm_1'
WARNING:tensorflow:Didn't find expected Conv2D or DepthwiseConv2dNative input to 'batch_normalization_20/cond/FusedBatchNorm'
WARNING:tensorflow:Didn't find expected Conv2D or DepthwiseConv2dNative input to 'batch_normalization_9/cond/FusedBatchNorm_1'
WARNING:tensorflow:Didn't find expected Conv2D or DepthwiseConv2dNative input to 'batch_normalization_9/cond/FusedBatchNorm'
WARNING:tensorflow:Didn't find expected Conv2D or DepthwiseConv2dNative input to 'batch_normalization_21/cond/FusedBatchNorm_1'
WARNING:tensorflow:Didn't find expected Conv2D or DepthwiseConv2dNative input to 'batch_normalization_21/cond/FusedBatchNorm'
WARNING:tensorflow:Didn't find expected Conv2D or DepthwiseConv2dNative input to 'batch_normalization_10/cond/FusedBatchNorm_1'
WARNING:tensorflow:Didn't find expected Conv2D or DepthwiseConv2dNative input to 'batch_normalization_10/cond/FusedBatchNorm'
WARNING:tensorflow:Didn't find expected Conv2D or DepthwiseConv2dNative input to 'batch_normalization_22/cond/FusedBatchNorm_1'
WARNING:tensorflow:Didn't find expected Conv2D or DepthwiseConv2dNative input to 'batch_normalization_22/cond/FusedBatchNorm'
WARNING:tensorflow:Didn't find expected Conv2D or DepthwiseConv2dNative input to 'batch_normalization_11/cond/FusedBatchNorm_1'
WARNING:tensorflow:Didn't find expected Conv2D or DepthwiseConv2dNative input to 'batch_normalization_11/cond/FusedBatchNorm'
WARNING:tensorflow:Didn't find expected Conv2D or DepthwiseConv2dNative input to 'batch_normalization_23/cond/FusedBatchNorm_1'
WARNING:tensorflow:Didn't find expected Conv2D or DepthwiseConv2dNative input to 'batch_normalization_23/cond/FusedBatchNorm'

@JeffreyCA
Copy link
Author

No I did not try to do that

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request feature New feature for Spleeter
Projects
None yet
Development

No branches or pull requests

4 participants