From 7e118e054f89bceb0cc0e134690b33c2a7cec910 Mon Sep 17 00:00:00 2001 From: "yichu.xu" Date: Sun, 14 Nov 2021 14:15:29 +0800 Subject: [PATCH 1/2] Update main.py Fix for vgg --- main.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/main.py b/main.py index 29c32a9..c74fd5f 100644 --- a/main.py +++ b/main.py @@ -1288,7 +1288,7 @@ def BBP_MAP(nets_list, model_meta_data, layer_type, net_dataidx_map, type_of_this_layer = layer_type[2 * layer_index - 2] type_of_prev_layer = layer_type[2 * layer_index - 2 - 2] - first_fc_identifier = (('fc' in type_of_this_layer or 'classifier' in type_of_this_layer) and ('conv' in type_of_prev_layer or 'features' in type_of_this_layer)) + first_fc_identifier = (('fc' in type_of_this_layer or 'classifier' in type_of_this_layer) and ('conv' in type_of_prev_layer or 'features' in type_of_prev_layer)) if first_fc_identifier: first_fc_index = layer_index @@ -1658,4 +1658,4 @@ def fedma_comm(batch_weights, model_meta_data, layer_type, net_dataidx_map, test_dl_global, assignments_list, comm_round=args.comm_round, - device=device) \ No newline at end of file + device=device) From 21f4d32338fd2563ebd97c737e3b9f4f470029d9 Mon Sep 17 00:00:00 2001 From: "yichu.xu" Date: Sun, 5 Dec 2021 19:39:09 +0800 Subject: [PATCH 2/2] add repobeats --- README.md | 2 ++ 1 file changed, 2 insertions(+) diff --git a/README.md b/README.md index 199316b..5672680 100644 --- a/README.md +++ b/README.md @@ -1,6 +1,8 @@ ## Federated Learning with Matched Averaging This is the code accompanying the ICLR 2020 paper "Federated Learning with Matched Averaging " Paper link: [https://openreview.net/forum?id=BkluqlSFDS] +![Alt](https://repobeats.axiom.co/api/embed/705163eb0ad960618a2f6fedbb1947992ba5a212.svg "Repobeats analytics image") + ### Overview --- FedMA algorithm is designed for federated learning of modern neural network architectures e.g. convolutional neural networks (CNNs) and LSTMs. FedMA constructs the shared global model in a layer-wise manner by matching and averaging hidden elements (i.e. channels for convolution layers; hidden states for LSTM; neurons for fully connected layers) with similar feature extraction signatures.