Using qkeras layers concurrently with Tensorflow's pruning tools.

[Duchstf]

Hello, very cool project!!

I'm just wondering if it would be possible to train the model using qkeras layers with the pruning tools in Tensorflow's model optimization package. For example, can we have something like this?

tf.keras.Sequential([
    sparsity.prune_low_magnitude(
        l.QConv2D(32, 5, padding='same', activation='relu'),
        input_shape=input_shape,
        **pruning_params)])

Thanks,

Duc.

[zhuangh]

@Duchstf good point, actually we have been looking into this with the team.

[vloncar]

@zhuangh Do you have an ETA for this? If you are not working on it, we can look into it, it seems straightforward to implement. I hope I'm not missing something.

[zhuangh]

No ETA yet. And for sure, if you have any things in mind, feel free to create PRs. We would be happy to review. thanks!

[zhuangh]

Hi @vloncar and @Duchstf

I assigned this to you. Hope it is OK. Feel free to let us know if you have any questions. thanks!

[vloncar]

Hi @zhuangh, custom prunable layers rely on stuff from tensorflow_model_optimization which is not part of the bare tensorflow installation. Should we implement it as a hard dependency or do we go for the monkeypatched version so that this functionality is avaliable only when tensorflow-model-optimization package is installed?

[zhuangh]

we could use tensorflow_model_optimization as dependency when installing qkeras. Let me know if you have any concern.

[zhuangh]

Merged and close this issue!