Updating validation functions

experiment.py

@@ -194,7 +194,7 @@ def main():
     # Setup Weights and Biases and specify hyperparameters
     test_method = "oe_test" if args.oe_test else "new_class_test"
     if args.baseline:
-        project_name = "TN_Masters_Proj_Baseline_Val_{}".format(test_method)
+        project_name = "TN_Masters_Proj_Fixed_Baseline_Val_{}".format(test_method)
     else:
         project_name = "TN_Masters_Proj_Val_{}_{}_{}".format(args.detection_type, args.loss, test_method)
 
@@ -260,7 +260,7 @@ def main():
             dist_norm=dist_norm
         )
 
-    if args.detection_type == "LS":
+    if args.detection_type == "LS" or args.baseline:
         net = FeatureExtractor(net_type, num_training_classes)
     elif args.detection_type == "KS":
         # Include a kitchen sink class

experiment_single.py

@@ -431,7 +431,10 @@ def main():
     # Setup Weights and Biases and specify hyperparameters
     test_method = "oe_test" if args.oe_test else "new_class_test"
     if args.baseline:
-        project_name = "Single_TN_Masters_Proj_Baseline_Val_{}".format(test_method)
+        if args.margin_baseline:
+            project_name = "Single_TN_Masters_Proj_margin_Baseline_Val_{}".format(test_method)
+        else:
+            project_name = "Single_TN_Masters_Proj_Baseline_Val_{}".format(test_method)
     else:
         project_name = "Single_TN_Masters_Proj_Val_{}_{}_{}".format(args.detection_type, args.loss, test_method)
 

margin_check.ipynb

@@ -2,51 +2,9 @@
  "cells": [
   {
    "cell_type": "code",
-   "execution_count": 1,
+   "execution_count": null,
    "metadata": {},
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "Files already downloaded and verified\n",
-      "Files already downloaded and verified\n",
-      "id batch 0 starting\n",
-      "id batch 1 starting\n",
-      "id batch 2 starting\n",
-      "id batch 3 starting\n",
-      "ood batch 0 starting\n",
-      "ood batch 1 starting\n",
-      "ood batch 2 starting\n",
-      "ood batch 3 starting\n",
-      "ood batch 4 starting\n",
-      "ood batch 5 starting\n",
-      "ood batch 6 starting\n",
-      "ood batch 7 starting\n",
-      "ood batch 8 starting\n",
-      "ood batch 9 starting\n",
-      "ood batch 10 starting\n",
-      "ood batch 11 starting\n",
-      "ood batch 12 starting\n",
-      "ood batch 13 starting\n",
-      "ood batch 14 starting\n",
-      "ood batch 15 starting\n",
-      "ood batch 16 starting\n",
-      "ood batch 17 starting\n",
-      "ood batch 18 starting\n",
-      "ood batch 19 starting\n",
-      "ood batch 20 starting\n",
-      "ood batch 21 starting\n",
-      "ood batch 22 starting\n",
-      "ood batch 23 starting\n",
-      "Test set: Accuracy: 649/1000 (65%)\n",
-      "Test Set: AUROC: 0.5688960000000001\n",
-      "\n",
-      "Accuracy: 64.9\n",
-      "AUROC: 0.5688960000000001\n"
-     ]
-    }
-   ],
+   "outputs": [],
    "source": [
     "import torch\n",
     "device = torch.device(\"cuda:0\")\n",
@@ -308,6 +266,10 @@
     "        pred_sequence.append(id_pred)\n",
     "        target_sequence.append(id_target)\n",
     "\n",
+    "        # FOR DEBUG MAX LOGIT SCORE\n",
+    "        # id_scores, _ = torch.max(id_output, dim=1)\n",
+    "        # id_scores = -1 * id_scores\n",
+    "\n",
     "        # Compute anomaly scores\n",
     "        # Use discriminant function to compute id_scores\n",
     "        # id_distance = torch.abs(id_distance)\n",
@@ -355,6 +317,9 @@
     "        pred_sequence.append(ood_pred)\n",
     "        target_sequence.append(ood_target)\n",
     "\n",
+    "        # FOR DEBUG MAX LOGIT SCORE\n",
+    "        # ood_scores, _ = torch.max(ood_output, dim=1)\n",
+    "        # ood_scores = -1 * ood_scores\n",
     "        ood_scores, _ = torch.max(-1 * ood_distance, dim=1)\n",
     "\n",
     "        # Detaching is important here because it removes these scores from computational graph\n",