diff --git a/experiments/cifar10/augmentation/cifar_augment.yaml b/experiments/cifar10/augmentation/cifar_augment.yaml
index c6365cd1..24d4f70c 100644
--- a/experiments/cifar10/augmentation/cifar_augment.yaml
+++ b/experiments/cifar10/augmentation/cifar_augment.yaml
@@ -1,8 +1,20 @@
 batch_augment:
   pipeline: null
   transforms:
-    RandomVerticalFlip:
+    #RandomVerticalFlip:
+    #  p: 0.5
+    RandomHorizontalFlip:
+      p: 0.5
+    RandomAffine:
+      degrees: [-15, 15]
+      translate: [0.1, 0.1]
+      scale: [0.9, 1.1]
+      shear: [-5, 5]
       p: 0.5
     RandomCrop:
       size: [32,32]
       padding: 4
+    RandomErasing:
+      p: 0.5
+      #scale: [0.!, 0.3]
+      #value: [0.4914, 0.4822, 0.4465]
diff --git a/experiments/cifar10/config.yaml b/experiments/cifar10/config.yaml
index c4749fa3..62640ef5 100644
--- a/experiments/cifar10/config.yaml
+++ b/experiments/cifar10/config.yaml
@@ -12,14 +12,14 @@ defaults:
 
 
 monitor:
-  metric: val_f1_micro
+  metric: val_accuracy
   direction: maximize
 
 module:
-  batch_size: 64
+  batch_size: 512
 
 trainer:
-  max_epochs: 50
+  max_epochs: 30
 
-scheduler:
-  max_lr: 0.1
+optimizer:
+  lr: 0.3
diff --git a/experiments/cifar10/experiment/sweep_lr.yaml b/experiments/cifar10/experiment/sweep_lr.yaml
index d3424177..f9995d03 100644
--- a/experiments/cifar10/experiment/sweep_lr.yaml
+++ b/experiments/cifar10/experiment/sweep_lr.yaml
@@ -6,4 +6,4 @@ hydra:
     subdir: lr=${scheduler.max_lr}
   sweeper:
     params:
-      scheduler.max_lr: 0.0001,0.001,0.01,0.1
+      scheduler.max_lr: 0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9
diff --git a/external/hannah-tvm b/external/hannah-tvm
index 0ede7e36..d2410e8d 160000
--- a/external/hannah-tvm
+++ b/external/hannah-tvm
@@ -1 +1 @@
-Subproject commit 0ede7e36f0cb295e322031dd6e9adad860acca7f
+Subproject commit d2410e8d098a3054bcabe269853188ab13bfb55f
diff --git a/hannah/models/timm.py b/hannah/models/timm.py
index 71b39b3f..280aa98a 100644
--- a/hannah/models/timm.py
+++ b/hannah/models/timm.py
@@ -228,12 +228,16 @@ def __init__(
                 if hasattr(self.encoder, "conv1"):
                     input_conv = self.encoder.conv1
                     out_channels = input_conv.out_channels
-                    new_conv = torch.nn.Conv2d(input_channels, out_channels, 3, 1)
+                    new_conv = torch.nn.Conv2d(
+                        input_channels, out_channels, 3, 1, padding=1
+                    )
                     self.encoder.conv1 = new_conv
                 elif hasattr(self.encoder, "conv_stem"):
                     input_conv = self.encoder.conv_stem
                     out_channels = input_conv.out_channels
-                    new_conv = torch.nn.Conv2d(input_channels, out_channels, 3, 1)
+                    new_conv = torch.nn.Conv2d(
+                        input_channels, out_channels, 3, 1, padding=1
+                    )
                     self.encoder.conv_stem = new_conv
                 else:
                     logger.critical(
@@ -242,6 +246,7 @@ def __init__(
 
                 if hasattr(self.encoder, "maxpool"):
                     self.encoder.maxpool = torch.nn.Identity()
+
         elif stem == "default":
             logger.info("""Using default stem for pulp model""")
         else: