Opensource Mobilenet V1 Image Classification Training Script about model-optimization HOT 5 OPEN

Hi, does the prune_low_magnitude function work with the depthwise_conv2d layers in mobilenet? I can only seem to get pruning results for just the normal conv2d layer.

from model-optimization.

A late update that instead of opensourcing it under this Github repo, we'll be integrating with TensorFlow Official Models. We'll update when it's ready to look at and use.

from model-optimization.

@faportillo : pruning supports the depthwise_conv2d layers, but doesn't actually prune them by default (see https://github.com/tensorflow/model-optimization/blob/master/tensorflow_model_optimization/python/core/sparsity/keras/prune_registry.py#L44).

In the future, for a feature request like this, please make comments only regarding the progress / scope / needs of the feature request. A question on how pruning handles mobilenet in general is better suited in a separate thread. Thanks!

from model-optimization.

Hi,
Currently trying to prune Mobilenet V2 model for classification. Can you point to the integrated training script that you used to get the results ?

from model-optimization.

Hi @liyunlu0618 , can you update recent status? Feel free to close this

from model-optimization.