February 9, 2022

The function to display model architecture has been added !

  • Model architecture can be viewed as a graph format.
  • By selecting each layer in the graph, user can see useful information for compression, such as input and output shape and latency of each layer.
  • Please note that latency can be displayed for models under 250MB.

Structured pruning method for PyTorch framework has been added !

  • Geometric Median Pruning (GM Pruning).

A video user guide is available.

  • How to use NPTK : Click here.
  • Click the link for details (including fine-tuning step): Click here.