Skip to content
Snippets Groups Projects

Repository graph

You can move around the graph by using the arrow keys.
Select Git revision
  • main default
  • old_main
2 results
Created with Raphaël 2.2.013Mar928Feb222019316Dec14Oct98713Sep14Aug131222Jul130Jun2921Apr201918Add deepcopy / use implicit devicemainmainBetter configuration dumpingFixed train/eval mode confusionFixed channel/filter confusion: channel -> filterFixed pytorch-lightning version requirementsFixed an issue with example_input_arrayRemoved unused files, outdated req, and updated readme.mdSome reformattingImproved StructuredPruningCallbackAdded EagleEye pruningReorganized structure pruning directoryAdded functionality to prepare for more pruning methodsAdded test for all image classification net architectures in torchvisionOrganize supported operators betterFixed failed tests due to pytorch version changeImproved CFG visual and type annotationFixed issues on last pruning epochSlightly extended DefaultShapeHandlerSupports loading directly from pruned checkpointAdded lr rewinding and checkpoint reloadingImproved checkpointing and accuracy logging_run_pruning -> prune_module_ (so it can be used directly)Fixed a problem about onnx_output_namesFixed misuse of pl_module.resultImplemented pruning on ConvTranspose2DFixed pruning not going all the way till last channelFixed sample_input on custom-collated dataloaderFixed last prune step not savedAdded capability to prune the conv layer before linear layerFixed a bug when dealing with layers that prunes 0 channelsAdded switch on ONNX model exportRemove a portion of filters from every layerFixed a problemImproved pruning loggingAdded protection ratio for layer to prevent all channels being prunedReformatted filesAdded test for callback (end to end) and fixed some problemsFixed and improved setup.cfgThis only works with lightning 1.3.xUpdated env requirement