Fix (ptq/benchmark): better dataframe creation

Xilinx · Dec 4, 2023 · e36faa4 · e36faa4
1 parent 652ca8e
commit e36faa4
Showing 1 changed file with 12 additions and 12 deletions.
diff --git a/src/brevitas_examples/imagenet_classification/ptq/benchmark/ptq_benchmark_torchvision.py b/src/brevitas_examples/imagenet_classification/ptq/benchmark/ptq_benchmark_torchvision.py
@@ -270,18 +270,16 @@ def ptq_torchvision_models(args):
     acc_diff = np.around(top1 - fp_accuracy, decimals=3)
     acc_ratio = np.around(top1 / fp_accuracy, decimals=3)
 
-    options_names = [k.replace('_', ' ').capitalize() for k in config_namespace.__dict__.keys()]
-    torchvision_df = pd.DataFrame(
-        columns=options_names + [
-            'Top 1% floating point accuracy',
-            'Top 1% quant accuracy',
-            'Floating point accuracy - quant accuracy',
-            'Quant accuracy / floating point accuracy',
-            'Calibration size',
-            'Calibration batch size',
-            'Torch version',
-            'Brevitas version'])
-    torchvision_df.at[0, :] = [v for _, v in config_namespace.__dict__.items()] + [
+    column_names = [k.replace('_', ' ').capitalize() for k in config_namespace.__dict__.keys()] + [
+        'Top 1% floating point accuracy',
+        'Top 1% quant accuracy',
+        'Floating point accuracy - quant accuracy',
+        'Quant accuracy / floating point accuracy',
+        'Calibration size',
+        'Calibration batch size',
+        'Torch version',
+        'Brevitas version']
+    values = [v for _, v in config_namespace.__dict__.items()] + [
         fp_accuracy,
         top1,
         acc_diff,
@@ -290,6 +288,8 @@ def ptq_torchvision_models(args):
         args.batch_size_calibration,
         torch_version,
         brevitas_version]
+    torchvision_df = pd.DataFrame([values], columns=column_names)
+
     folder = './multirun/' + str(args.idx)
     os.makedirs(folder, exist_ok=True)
     torchvision_df.to_csv(os.path.join(folder, 'RESULTS_TORCHVISION.csv'), index=False)