From b2f2ff67f5ee0f1fe14a0454a20d4f6d09aa951f Mon Sep 17 00:00:00 2001 From: Neta Zmora <31280975+nzmora@users.noreply.github.com> Date: Mon, 22 Oct 2018 21:13:46 +0300 Subject: [PATCH] Update README with a new "What's new in..." feature --- README.md | 23 +++++++++++++++++++++++ 1 file changed, 23 insertions(+) diff --git a/README.md b/README.md index 1499d12..0ac783f 100755 --- a/README.md +++ b/README.md @@ -35,6 +35,29 @@ Network compression can reduce the memory footprint of a neural network, increase its inference speed and save energy. Distiller provides a [PyTorch](http://pytorch.org/) environment for prototyping and analyzing compression algorithms, such as sparsity-inducing methods and low-precision arithmetic. +<details><summary><b>What's New in October?</b></summary> +<p> +We've added collection of activation statistics! + +Activation statistics can be leveraged to make pruning and quantization decisions, and so +we added support to collect these data. +Two types of activation statistics are supported: summary statistics, and detailed records +per activation. +Currently we support the following summaries: +- Average activation sparsity, per layer +- Average L1-norm for each activation channel, per layer +- Average sparsity for each activation channel, per layer + +For the detailed records we collect some statistics per activation and store it in a record. +Using this collection method generates more detailed data, but consumes more time, so +Beware. + +* You can collect activation data for the different training phases: training/validation/test. +* You can access the data directly from each module that you chose to collect stats for. +* You can also create an Excel workbook with the stats. +</p> +</details> + ## Table of Contents * [Feature set](#feature-set) -- GitLab