Format baselines

baselines/fedbn/docs/multirun_plot.ipynb

@@ -58,12 +58,12 @@
    "source": [
     "def fuse_by_dataset(losses):\n",
     "    \"\"\"Transform per-round history (list of dicts) into\n",
-    "    \n",
+    "\n",
     "    a single dict, with values as lists.\"\"\"\n",
     "    fussed_losses = {}\n",
     "\n",
     "    for _, loss_dict in losses:\n",
-    "        for k,v in loss_dict.items():\n",
+    "        for k, v in loss_dict.items():\n",
     "            if k in fussed_losses:\n",
     "                fussed_losses[k].append(v)\n",
     "            else:\n",
@@ -82,9 +82,14 @@
     "    res_list = []\n",
     "    for results in list(Path(path_multirun).glob(\"**/history.pkl\")):\n",
     "        data, config = read_pickle_and_config(results)\n",
-    "        pre_train_loss = data['history'].metrics_distributed_fit['pre_train_losses']\n",
+    "        pre_train_loss = data[\"history\"].metrics_distributed_fit[\"pre_train_losses\"]\n",
     "        fussed_losses = fuse_by_dataset(pre_train_loss)\n",
-    "        res_list.append({'strategy': config['client']['client_label'], 'train_losses': fussed_losses})\n",
+    "        res_list.append(\n",
+    "            {\n",
+    "                \"strategy\": config[\"client\"][\"client_label\"],\n",
+    "                \"train_losses\": fussed_losses,\n",
+    "            }\n",
+    "        )\n",
     "    return res_list"
    ]
   },
@@ -95,7 +100,7 @@
    "outputs": [],
    "source": [
     "# Here replace with the path of the multi run you just comleted\n",
-    "all_losses = process_multirun_data('../multirun/2023-11-15/22-28-28')"
+    "all_losses = process_multirun_data(\"../multirun/2023-11-15/22-28-28\")"
    ]
   },
   {
@@ -104,30 +109,29 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "\n",
     "def average_by_client_type(all_fused_lossed):\n",
     "    \"\"\"If there are multliple runs for the same strategy add them up,\n",
-    "    \n",
+    "\n",
     "    average them later. This is useful if you run the `--multirun` running\n",
     "    more than one time the same configuration.\"\"\"\n",
-    "    \n",
+    "\n",
     "    # identify how many unique clients were used\n",
     "    to_plot = {}\n",
     "    for res in all_fused_lossed:\n",
-    "        strategy = res['strategy']\n",
+    "        strategy = res[\"strategy\"]\n",
     "        if strategy not in to_plot:\n",
     "            to_plot[strategy] = {}\n",
-    "        \n",
-    "        for dataset, train_loss in res['train_losses'].items():\n",
+    "\n",
+    "        for dataset, train_loss in res[\"train_losses\"].items():\n",
     "            if dataset in to_plot[strategy]:\n",
-    "                to_plot[strategy][dataset]['train_loss'] += np.array(train_loss)\n",
-    "                to_plot[strategy][dataset]['run_count'] += 1\n",
+    "                to_plot[strategy][dataset][\"train_loss\"] += np.array(train_loss)\n",
+    "                to_plot[strategy][dataset][\"run_count\"] += 1\n",
     "            else:\n",
-    "                to_plot[strategy][dataset] = {'train_loss': np.array(train_loss)}\n",
-    "                to_plot[strategy][dataset]['run_count'] = 1\n",
+    "                to_plot[strategy][dataset] = {\"train_loss\": np.array(train_loss)}\n",
+    "                to_plot[strategy][dataset][\"run_count\"] = 1\n",
     "\n",
     "    # print(to_plot)\n",
-    "    return to_plot\n"
+    "    return to_plot"
    ]
   },
   {
@@ -167,25 +171,25 @@
     "print(datasets)\n",
     "\n",
     "num_datasets = len(datasets)\n",
-    "fig, axs = plt.subplots(figsize=(14,4), nrows=1, ncols=num_datasets)\n",
+    "fig, axs = plt.subplots(figsize=(14, 4), nrows=1, ncols=num_datasets)\n",
     "\n",
     "\n",
     "for s_id, (strategy, results) in enumerate(to_plot.items()):\n",
     "    for i, dataset in enumerate(datasets):\n",
-    "        loss = results[dataset]['train_loss']/results[dataset]['run_count']\n",
+    "        loss = results[dataset][\"train_loss\"] / results[dataset][\"run_count\"]\n",
     "        axs[i].plot(range(len(loss)), loss, label=strategy)\n",
-    "        axs[i].set_xlabel('Round')\n",
+    "        axs[i].set_xlabel(\"Round\")\n",
     "        if i == 0:\n",
-    "            axs[i].set_ylabel('Train Loss')\n",
+    "            axs[i].set_ylabel(\"Train Loss\")\n",
     "\n",
     "        axs[i].legend()\n",
     "\n",
-    "        if s_id==0:\n",
+    "        if s_id == 0:\n",
     "            axs[i].grid()\n",
     "            axs[i].set_title(dataset)\n",
-    "            axs[i].set_xticks(np.arange(0,100+1, 25))\n",
+    "            axs[i].set_xticks(np.arange(0, 100 + 1, 25))\n",
     "\n",
-    "saveFig('train_loss.png', fig)"
+    "saveFig(\"train_loss.png\", fig)"
    ]
   },
   {

baselines/hfedxgboost/hfedxgboost/__init__.py

@@ -1 +1 @@
-"""hfedxgboost baseline package."""
+"""Hfedxgboost baseline package."""

baselines/hfedxgboost/hfedxgboost/strategy.py

@@ -9,9 +9,8 @@
 from flwr.common import FitRes, Scalar, ndarrays_to_parameters, parameters_to_ndarrays
 from flwr.common.logger import log
 from flwr.server.client_proxy import ClientProxy
-
-from flwr.server.strategy.aggregate import aggregate
 from flwr.server.strategy import FedAvg
+from flwr.server.strategy.aggregate import aggregate
 
 
 class FedXgbNnAvg(FedAvg):