Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,19 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
---
|
| 4 |
+
|
| 5 |
+
reproducing: "Ensemble everything everywhere: Multi-scale aggregation for adversarial robustness" (https://arxiv.org/pdf/2408.05446)
|
| 6 |
+
|
| 7 |
+
hyperparameters:
|
| 8 |
+
|
| 9 |
+
```
|
| 10 |
+
- architecture = based on Torchvision's Resnet152 default implementation
|
| 11 |
+
- criterion = torch.nn.CrossEntropyLoss()
|
| 12 |
+
- optimizer = torch.optim.AdamW
|
| 13 |
+
- scaler = GradScaler
|
| 14 |
+
- datasets = ["cifar10", "cirfar100"]
|
| 15 |
+
|
| 16 |
+
- lr = 0.0001
|
| 17 |
+
- num_epochs = 16 (higher would be even better, but maybe by <1%)
|
| 18 |
+
- crossmax_k = 2 (difference between crossmax_k=2 and crossmax_k=3 is about 1-2%, so it's not a big deal)
|
| 19 |
+
```
|