Naturally trained surrogates for the Scissors Effect

.gitattributes

@@ -0,0 +1 @@
+*.pt filter=lfs diff=lfs merge=lfs -text

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Yuhang Jiang, Xiaojing Chen
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -0,0 +1,136 @@
+---
+license: mit
+library_name: pytorch
+tags:
+  - adversarial-robustness
+  - adversarial-examples
+  - transfer-attack
+  - image-classification
+datasets:
+  - cifar10
+  - cifar100
+---
+
+# Scissors Effect: naturally trained surrogates
+
+The five checkpoints used as the *standard* (naturally trained) surrogates in
+**"The Scissors Effect: When Resize-Based Input Diversity Helps or Hurts Transfer
+Attacks"**, published in *Transactions on Machine Learning Research* (2026).
+
+Every other model in that paper comes from the RobustBench zoo, torchvision, or
+`open_clip` and downloads on its own. These five do not exist anywhere else, and
+without them the CIFAR-10 standard side and the whole CIFAR-100 experiment cannot
+be reproduced without retraining. That is the only reason this repository exists.
+
+- **Paper:** [arXiv:2606.22516](https://arxiv.org/abs/2606.22516) · [OpenReview](https://openreview.net/forum?id=b4pCcgJM0M)
+- **Code:** [github.com/Avalon-S/ScissorsEffect](https://github.com/Avalon-S/ScissorsEffect)
+- **Project page:** [avalon-s.github.io/ScissorsEffect](https://avalon-s.github.io/ScissorsEffect/)
+
+## The checkpoints
+
+| File | Dataset | Architecture | Clean accuracy | Size | Train time |
+|---|---|---|---|---|---|
+| `c10_resnet18.pt` | CIFAR-10 | ResNet-18 | 94.92% | 42.7 MB | 31 min |
+| `c10_resnet50.pt` | CIFAR-10 | ResNet-50 | 94.66% | 90.0 MB | 69 min |
+| `c10_vgg16.pt` | CIFAR-10 | VGG-16 | 93.82% | 56.2 MB | 22 min |
+| `c10_densenet121.pt` | CIFAR-10 | DenseNet-121 | 95.39% | 27.1 MB | 111 min |
+| `Standard_WRN28_10.pt` | CIFAR-100 | WRN-28-10 | 81.07% | 139.5 MB | 2.5 h |
+
+Times are for a single RTX 4090. The paper quotes the CIFAR-10 four as reaching
+93.8–95.4% clean accuracy, which is the range above.
+
+The architectures are CIFAR variants: 3x3 stem, no initial max-pool, not ImageNet
+architectures fed upsampled 32x32 images.
+
+## Two storage formats
+
+This is the one thing worth reading before you load anything.
+
+**The four CIFAR-10 checkpoints are saved wrapped.** The normalisation lives
+inside the checkpoint, so the model consumes `[0,1]` images directly. Each is a
+dict with `state_dict`, `arch`, `clean_acc` and the training `recipe`.
+
+**The CIFAR-100 checkpoint is a bare state dict** for RobustBench's
+`WideResNet(depth=28, num_classes=100, widen_factor=10)`. It carries no
+normalisation of its own; the loader applies the CIFAR-100 statistics around it.
+
+Loading either one by hand:
+
+```python
+import torch, importlib.util
+
+# --- CIFAR-10: wrapped, consumes [0,1] ---
+spec = importlib.util.spec_from_file_location(
+    "t", "ScissorsEffect/scripts/train_cifar10_standard.py")
+m = importlib.util.module_from_spec(spec); spec.loader.exec_module(m)
+
+ck = torch.load("c10_resnet18.pt", map_location="cpu", weights_only=False)
+net = m.Normalized(m.BUILDERS[ck["arch"]]())
+net.load_state_dict(ck["state_dict"], strict=True)
+net.eval()                       # net(x) with x in [0,1]
+
+# --- CIFAR-100: bare, normalise outside ---
+from robustbench.model_zoo.architectures.wide_resnet import WideResNet
+
+sd = torch.load("Standard_WRN28_10.pt", map_location="cpu", weights_only=False)
+sd = sd.get("state_dict", sd) if isinstance(sd, dict) else sd
+sd = {(k[7:] if k.startswith("module.") else k): v for k, v in sd.items()}
+wrn = WideResNet(depth=28, num_classes=100, widen_factor=10)
+wrn.load_state_dict(sd, strict=True)
+# mean (0.5071, 0.4865, 0.4409), std (0.2673, 0.2564, 0.2762)
+```
+
+Through the paper's own code you do not need any of this. Put the files where
+`models/loader.py` looks for them and ask for the model by name:
+
+```
+<MODEL_ROOT>/cifar10/standard/c10_resnet18.pt
+<MODEL_ROOT>/cifar10/standard/c10_resnet50.pt
+<MODEL_ROOT>/cifar10/standard/c10_vgg16.pt
+<MODEL_ROOT>/cifar10/standard/c10_densenet121.pt
+<MODEL_ROOT>/cifar100/Linf/Standard_WRN28_10.pt
+```
+
+Loading refuses a partially initialised network and re-asserts the stored clean
+accuracy, so a truncated download raises instead of quietly degrading a result.
+
+## Verify the download
+
+```
+sha256sum -c SHA256SUMS
+```
+
+## How they were trained
+
+One recipe for all five: SGD (lr 0.1, momentum 0.9, weight decay 5e-4),
+MultiStepLR at epochs [100, 150] with gamma 0.1, 200 epochs, batch 128,
+`RandomCrop(32, padding=4)` + `RandomHorizontalFlip`. No label smoothing, no
+mixup or cutmix, no EMA. The run aborts if a model finishes below a
+clean-accuracy floor, so a failed model never reaches the experiments.
+
+Reproduce with the script in the code repository:
+
+```
+python scripts/train_cifar10_standard.py                      # the CIFAR-10 four
+python scripts/train_cifar10_standard.py --dataset cifar100   # the WRN-28-10
+```
+
+`provenance/` holds the records the training runs wrote themselves:
+`cifar10_training_summary.json` (recipe, per-model accuracy and wall-clock),
+`cifar100_training_summary.txt`, and the per-epoch `cifar100_training_log.txt`.
+
+## Citation
+
+```bibtex
+@article{jiang2026scissors,
+  title   = {The Scissors Effect: When Resize-Based Input Diversity Helps or Hurts Transfer Attacks},
+  author  = {Jiang, Yuhang and Chen, Xiaojing},
+  journal = {Transactions on Machine Learning Research},
+  year    = {2026},
+  url     = {https://openreview.net/forum?id=b4pCcgJM0M}
+}
+```
+
+## License
+
+MIT, matching the code repository. The paper is published by TMLR under CC BY 4.0.

SHA256SUMS

@@ -0,0 +1,5 @@
+16560472171d18750c31daec5571708edfc1e6ce050392ae3deb591e15f4fabf  Standard_WRN28_10.pt
+e3499d870c5defeebba589b8dc50515111bde6487f4e2833d727c276a95ad5fd  c10_densenet121.pt
+390c455bd6f05e485576cf6f37fdef4fdf315747098f6544d2d752dc0442aa48  c10_resnet18.pt
+12f88aff5da6b5331708736ac95fc6dcecf987ff1182dfcef4313f74f0eb763c  c10_resnet50.pt
+a5238349fe396f8cf6dc2d81bc5c6ab26c7b73b0e59478fb2646f3a516a7d850  c10_vgg16.pt

Standard_WRN28_10.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16560472171d18750c31daec5571708edfc1e6ce050392ae3deb591e15f4fabf
+size 146274451

c10_densenet121.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3499d870c5defeebba589b8dc50515111bde6487f4e2833d727c276a95ad5fd
+size 28406887

c10_resnet18.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:390c455bd6f05e485576cf6f37fdef4fdf315747098f6544d2d752dc0442aa48
+size 44777041

c10_resnet50.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12f88aff5da6b5331708736ac95fc6dcecf987ff1182dfcef4313f74f0eb763c
+size 94404963

c10_vgg16.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a5238349fe396f8cf6dc2d81bc5c6ab26c7b73b0e59478fb2646f3a516a7d850
+size 58958557

provenance/cifar100_training_log.txt

@@ -0,0 +1,200 @@
+Epoch [1/200] Train Loss: 3.9027, Train Acc: 9.57% | Test Loss: 3.8001, Test Acc: 12.13% | LR: 0.100000 | Time: 45.3s
+Epoch [2/200] Train Loss: 3.1617, Train Acc: 21.53% | Test Loss: 3.1292, Test Acc: 23.34% | LR: 0.100000 | Time: 45.0s
+Epoch [3/200] Train Loss: 2.5378, Train Acc: 33.16% | Test Loss: 2.5066, Test Acc: 35.34% | LR: 0.100000 | Time: 44.8s
+Epoch [4/200] Train Loss: 2.1321, Train Acc: 41.99% | Test Loss: 2.2360, Test Acc: 39.95% | LR: 0.100000 | Time: 44.9s
+Epoch [5/200] Train Loss: 1.8709, Train Acc: 47.96% | Test Loss: 2.1828, Test Acc: 42.93% | LR: 0.100000 | Time: 44.9s
+Epoch [6/200] Train Loss: 1.6979, Train Acc: 52.26% | Test Loss: 2.0140, Test Acc: 45.92% | LR: 0.100000 | Time: 45.1s
+Epoch [7/200] Train Loss: 1.5627, Train Acc: 55.70% | Test Loss: 1.9089, Test Acc: 48.22% | LR: 0.100000 | Time: 44.7s
+Epoch [8/200] Train Loss: 1.4759, Train Acc: 58.06% | Test Loss: 1.8772, Test Acc: 49.85% | LR: 0.100000 | Time: 45.0s
+Epoch [9/200] Train Loss: 1.3977, Train Acc: 59.92% | Test Loss: 1.5909, Test Acc: 54.88% | LR: 0.100000 | Time: 44.9s
+Epoch [10/200] Train Loss: 1.3402, Train Acc: 61.67% | Test Loss: 1.7453, Test Acc: 52.87% | LR: 0.100000 | Time: 44.8s
+Epoch [11/200] Train Loss: 1.2825, Train Acc: 63.08% | Test Loss: 1.6069, Test Acc: 55.49% | LR: 0.100000 | Time: 44.9s
+Epoch [12/200] Train Loss: 1.2409, Train Acc: 64.15% | Test Loss: 1.7075, Test Acc: 54.09% | LR: 0.100000 | Time: 44.9s
+Epoch [13/200] Train Loss: 1.2071, Train Acc: 65.10% | Test Loss: 1.6929, Test Acc: 55.67% | LR: 0.100000 | Time: 44.9s
+Epoch [14/200] Train Loss: 1.1687, Train Acc: 65.92% | Test Loss: 1.8124, Test Acc: 53.43% | LR: 0.100000 | Time: 44.9s
+Epoch [15/200] Train Loss: 1.1389, Train Acc: 66.88% | Test Loss: 1.8655, Test Acc: 52.55% | LR: 0.100000 | Time: 44.9s
+Epoch [16/200] Train Loss: 1.1070, Train Acc: 67.82% | Test Loss: 1.6661, Test Acc: 55.89% | LR: 0.100000 | Time: 44.9s
+Epoch [17/200] Train Loss: 1.0881, Train Acc: 68.32% | Test Loss: 1.7088, Test Acc: 55.00% | LR: 0.100000 | Time: 44.9s
+Epoch [18/200] Train Loss: 1.0621, Train Acc: 68.97% | Test Loss: 1.5663, Test Acc: 57.96% | LR: 0.100000 | Time: 44.9s
+Epoch [19/200] Train Loss: 1.0412, Train Acc: 69.49% | Test Loss: 1.7498, Test Acc: 56.38% | LR: 0.100000 | Time: 44.9s
+Epoch [20/200] Train Loss: 1.0307, Train Acc: 69.67% | Test Loss: 1.5155, Test Acc: 58.64% | LR: 0.100000 | Time: 44.9s
+Epoch [21/200] Train Loss: 1.0113, Train Acc: 70.39% | Test Loss: 1.7142, Test Acc: 55.77% | LR: 0.100000 | Time: 44.9s
+Epoch [22/200] Train Loss: 0.9941, Train Acc: 70.93% | Test Loss: 1.5311, Test Acc: 59.82% | LR: 0.100000 | Time: 45.0s
+Epoch [23/200] Train Loss: 0.9876, Train Acc: 71.02% | Test Loss: 1.6224, Test Acc: 58.29% | LR: 0.100000 | Time: 44.9s
+Epoch [24/200] Train Loss: 0.9735, Train Acc: 71.35% | Test Loss: 1.6523, Test Acc: 57.64% | LR: 0.100000 | Time: 44.9s
+Epoch [25/200] Train Loss: 0.9581, Train Acc: 71.83% | Test Loss: 1.6919, Test Acc: 57.26% | LR: 0.100000 | Time: 44.9s
+Epoch [26/200] Train Loss: 0.9450, Train Acc: 72.13% | Test Loss: 1.5474, Test Acc: 59.01% | LR: 0.100000 | Time: 44.9s
+Epoch [27/200] Train Loss: 0.9370, Train Acc: 72.36% | Test Loss: 1.8919, Test Acc: 54.00% | LR: 0.100000 | Time: 44.8s
+Epoch [28/200] Train Loss: 0.9256, Train Acc: 72.66% | Test Loss: 1.5647, Test Acc: 58.60% | LR: 0.100000 | Time: 44.9s
+Epoch [29/200] Train Loss: 0.9221, Train Acc: 72.96% | Test Loss: 1.6751, Test Acc: 57.73% | LR: 0.100000 | Time: 44.9s
+Epoch [30/200] Train Loss: 0.9061, Train Acc: 73.27% | Test Loss: 1.6911, Test Acc: 57.54% | LR: 0.100000 | Time: 44.9s
+Epoch [31/200] Train Loss: 0.9072, Train Acc: 73.35% | Test Loss: 1.5949, Test Acc: 59.13% | LR: 0.100000 | Time: 44.9s
+Epoch [32/200] Train Loss: 0.9025, Train Acc: 73.42% | Test Loss: 1.5185, Test Acc: 60.23% | LR: 0.100000 | Time: 44.9s
+Epoch [33/200] Train Loss: 0.8947, Train Acc: 73.52% | Test Loss: 1.6397, Test Acc: 58.47% | LR: 0.100000 | Time: 45.0s
+Epoch [34/200] Train Loss: 0.8822, Train Acc: 73.92% | Test Loss: 1.4328, Test Acc: 61.76% | LR: 0.100000 | Time: 45.0s
+Epoch [35/200] Train Loss: 0.8820, Train Acc: 73.88% | Test Loss: 1.5502, Test Acc: 59.72% | LR: 0.100000 | Time: 44.9s
+Epoch [36/200] Train Loss: 0.8679, Train Acc: 74.24% | Test Loss: 1.9002, Test Acc: 53.13% | LR: 0.100000 | Time: 44.9s
+Epoch [37/200] Train Loss: 0.8680, Train Acc: 74.26% | Test Loss: 1.8651, Test Acc: 54.57% | LR: 0.100000 | Time: 44.9s
+Epoch [38/200] Train Loss: 0.8625, Train Acc: 74.67% | Test Loss: 1.5945, Test Acc: 59.54% | LR: 0.100000 | Time: 45.1s
+Epoch [39/200] Train Loss: 0.8509, Train Acc: 74.81% | Test Loss: 1.5634, Test Acc: 59.47% | LR: 0.100000 | Time: 44.7s
+Epoch [40/200] Train Loss: 0.8581, Train Acc: 74.44% | Test Loss: 1.6967, Test Acc: 58.22% | LR: 0.100000 | Time: 45.0s
+Epoch [41/200] Train Loss: 0.8496, Train Acc: 74.84% | Test Loss: 1.7438, Test Acc: 56.23% | LR: 0.100000 | Time: 44.8s
+Epoch [42/200] Train Loss: 0.8347, Train Acc: 75.30% | Test Loss: 1.6046, Test Acc: 58.69% | LR: 0.100000 | Time: 44.9s
+Epoch [43/200] Train Loss: 0.8323, Train Acc: 75.26% | Test Loss: 1.4764, Test Acc: 62.59% | LR: 0.100000 | Time: 44.9s
+Epoch [44/200] Train Loss: 0.8398, Train Acc: 75.13% | Test Loss: 1.6000, Test Acc: 58.60% | LR: 0.100000 | Time: 45.0s
+Epoch [45/200] Train Loss: 0.8358, Train Acc: 75.38% | Test Loss: 1.6417, Test Acc: 57.48% | LR: 0.100000 | Time: 44.7s
+Epoch [46/200] Train Loss: 0.8318, Train Acc: 75.33% | Test Loss: 1.4859, Test Acc: 60.97% | LR: 0.100000 | Time: 45.2s
+Epoch [47/200] Train Loss: 0.8237, Train Acc: 75.64% | Test Loss: 1.5108, Test Acc: 60.73% | LR: 0.100000 | Time: 44.7s
+Epoch [48/200] Train Loss: 0.8345, Train Acc: 75.28% | Test Loss: 1.5540, Test Acc: 60.49% | LR: 0.100000 | Time: 44.8s
+Epoch [49/200] Train Loss: 0.8193, Train Acc: 75.65% | Test Loss: 1.6214, Test Acc: 58.54% | LR: 0.100000 | Time: 44.9s
+Epoch [50/200] Train Loss: 0.8197, Train Acc: 75.89% | Test Loss: 1.4732, Test Acc: 60.97% | LR: 0.100000 | Time: 45.0s
+Epoch [51/200] Train Loss: 0.8180, Train Acc: 75.70% | Test Loss: 1.4832, Test Acc: 61.94% | LR: 0.100000 | Time: 44.9s
+Epoch [52/200] Train Loss: 0.8156, Train Acc: 75.70% | Test Loss: 1.4530, Test Acc: 61.64% | LR: 0.100000 | Time: 44.9s
+Epoch [53/200] Train Loss: 0.8045, Train Acc: 75.97% | Test Loss: 1.5574, Test Acc: 59.76% | LR: 0.100000 | Time: 44.9s
+Epoch [54/200] Train Loss: 0.8183, Train Acc: 75.68% | Test Loss: 1.6482, Test Acc: 59.02% | LR: 0.100000 | Time: 45.1s
+Epoch [55/200] Train Loss: 0.7978, Train Acc: 76.34% | Test Loss: 1.7140, Test Acc: 57.05% | LR: 0.100000 | Time: 44.9s
+Epoch [56/200] Train Loss: 0.7988, Train Acc: 76.39% | Test Loss: 1.5139, Test Acc: 61.08% | LR: 0.100000 | Time: 44.9s
+Epoch [57/200] Train Loss: 0.8053, Train Acc: 75.99% | Test Loss: 1.5584, Test Acc: 59.73% | LR: 0.100000 | Time: 44.9s
+Epoch [58/200] Train Loss: 0.8056, Train Acc: 76.09% | Test Loss: 1.6650, Test Acc: 57.70% | LR: 0.100000 | Time: 44.9s
+Epoch [59/200] Train Loss: 0.8008, Train Acc: 76.03% | Test Loss: 1.4592, Test Acc: 62.11% | LR: 0.100000 | Time: 44.9s
+Epoch [60/200] Train Loss: 0.7940, Train Acc: 76.52% | Test Loss: 1.5460, Test Acc: 60.72% | LR: 0.100000 | Time: 44.9s
+Epoch [61/200] Train Loss: 0.7821, Train Acc: 76.52% | Test Loss: 1.8348, Test Acc: 55.11% | LR: 0.100000 | Time: 44.9s
+Epoch [62/200] Train Loss: 0.7937, Train Acc: 76.36% | Test Loss: 1.6835, Test Acc: 59.28% | LR: 0.100000 | Time: 44.9s
+Epoch [63/200] Train Loss: 0.7908, Train Acc: 76.45% | Test Loss: 1.4228, Test Acc: 62.84% | LR: 0.100000 | Time: 44.9s
+Epoch [64/200] Train Loss: 0.7843, Train Acc: 76.73% | Test Loss: 1.4802, Test Acc: 61.32% | LR: 0.100000 | Time: 44.9s
+Epoch [65/200] Train Loss: 0.7790, Train Acc: 76.91% | Test Loss: 1.5727, Test Acc: 60.03% | LR: 0.100000 | Time: 44.9s
+Epoch [66/200] Train Loss: 0.7826, Train Acc: 76.95% | Test Loss: 1.8275, Test Acc: 56.28% | LR: 0.100000 | Time: 44.9s
+Epoch [67/200] Train Loss: 0.7828, Train Acc: 76.61% | Test Loss: 1.6868, Test Acc: 58.27% | LR: 0.100000 | Time: 44.9s
+Epoch [68/200] Train Loss: 0.7908, Train Acc: 76.37% | Test Loss: 1.5092, Test Acc: 60.84% | LR: 0.100000 | Time: 44.9s
+Epoch [69/200] Train Loss: 0.7740, Train Acc: 76.89% | Test Loss: 1.4356, Test Acc: 61.51% | LR: 0.100000 | Time: 44.9s
+Epoch [70/200] Train Loss: 0.7720, Train Acc: 77.03% | Test Loss: 1.5211, Test Acc: 60.80% | LR: 0.100000 | Time: 44.9s
+Epoch [71/200] Train Loss: 0.7750, Train Acc: 76.99% | Test Loss: 1.4151, Test Acc: 62.25% | LR: 0.100000 | Time: 44.9s
+Epoch [72/200] Train Loss: 0.7690, Train Acc: 76.86% | Test Loss: 1.8465, Test Acc: 54.89% | LR: 0.100000 | Time: 44.9s
+Epoch [73/200] Train Loss: 0.7714, Train Acc: 77.07% | Test Loss: 1.9562, Test Acc: 54.86% | LR: 0.100000 | Time: 44.7s
+Epoch [74/200] Train Loss: 0.7563, Train Acc: 77.49% | Test Loss: 1.5790, Test Acc: 59.72% | LR: 0.100000 | Time: 44.8s
+Epoch [75/200] Train Loss: 0.7757, Train Acc: 76.94% | Test Loss: 1.8245, Test Acc: 56.46% | LR: 0.100000 | Time: 44.9s
+Epoch [76/200] Train Loss: 0.7745, Train Acc: 77.02% | Test Loss: 1.5566, Test Acc: 58.86% | LR: 0.100000 | Time: 45.0s
+Epoch [77/200] Train Loss: 0.7610, Train Acc: 77.42% | Test Loss: 1.7337, Test Acc: 56.23% | LR: 0.100000 | Time: 44.9s
+Epoch [78/200] Train Loss: 0.7700, Train Acc: 77.07% | Test Loss: 1.5514, Test Acc: 60.30% | LR: 0.100000 | Time: 44.9s
+Epoch [79/200] Train Loss: 0.7634, Train Acc: 77.28% | Test Loss: 1.4950, Test Acc: 61.60% | LR: 0.100000 | Time: 44.9s
+Epoch [80/200] Train Loss: 0.7672, Train Acc: 77.29% | Test Loss: 1.4062, Test Acc: 62.34% | LR: 0.100000 | Time: 44.9s
+Epoch [81/200] Train Loss: 0.7700, Train Acc: 77.03% | Test Loss: 1.5979, Test Acc: 60.66% | LR: 0.100000 | Time: 44.9s
+Epoch [82/200] Train Loss: 0.7565, Train Acc: 77.47% | Test Loss: 1.6587, Test Acc: 58.53% | LR: 0.100000 | Time: 44.9s
+Epoch [83/200] Train Loss: 0.7612, Train Acc: 77.17% | Test Loss: 1.4111, Test Acc: 63.18% | LR: 0.100000 | Time: 44.9s
+Epoch [84/200] Train Loss: 0.7532, Train Acc: 77.36% | Test Loss: 1.4371, Test Acc: 61.80% | LR: 0.100000 | Time: 44.9s
+Epoch [85/200] Train Loss: 0.7559, Train Acc: 77.45% | Test Loss: 1.6022, Test Acc: 60.38% | LR: 0.100000 | Time: 44.9s
+Epoch [86/200] Train Loss: 0.7583, Train Acc: 77.36% | Test Loss: 1.5432, Test Acc: 60.98% | LR: 0.100000 | Time: 44.9s
+Epoch [87/200] Train Loss: 0.7435, Train Acc: 77.76% | Test Loss: 1.4442, Test Acc: 62.52% | LR: 0.100000 | Time: 44.9s
+Epoch [88/200] Train Loss: 0.7587, Train Acc: 77.35% | Test Loss: 1.5879, Test Acc: 60.21% | LR: 0.100000 | Time: 44.9s
+Epoch [89/200] Train Loss: 0.7498, Train Acc: 77.58% | Test Loss: 1.9159, Test Acc: 55.91% | LR: 0.100000 | Time: 44.9s
+Epoch [90/200] Train Loss: 0.7652, Train Acc: 77.27% | Test Loss: 1.9317, Test Acc: 54.45% | LR: 0.100000 | Time: 44.9s
+Epoch [91/200] Train Loss: 0.7430, Train Acc: 77.74% | Test Loss: 1.5898, Test Acc: 59.67% | LR: 0.100000 | Time: 44.9s
+Epoch [92/200] Train Loss: 0.7515, Train Acc: 77.68% | Test Loss: 1.7150, Test Acc: 57.93% | LR: 0.100000 | Time: 44.9s
+Epoch [93/200] Train Loss: 0.7510, Train Acc: 77.47% | Test Loss: 1.6669, Test Acc: 58.87% | LR: 0.100000 | Time: 44.9s
+Epoch [94/200] Train Loss: 0.7531, Train Acc: 77.50% | Test Loss: 1.6000, Test Acc: 60.03% | LR: 0.100000 | Time: 44.9s
+Epoch [95/200] Train Loss: 0.7480, Train Acc: 77.75% | Test Loss: 1.8505, Test Acc: 57.02% | LR: 0.100000 | Time: 44.9s
+Epoch [96/200] Train Loss: 0.7493, Train Acc: 77.69% | Test Loss: 1.4097, Test Acc: 63.13% | LR: 0.100000 | Time: 44.4s
+Epoch [97/200] Train Loss: 0.7473, Train Acc: 77.70% | Test Loss: 1.6782, Test Acc: 57.95% | LR: 0.100000 | Time: 45.1s
+Epoch [98/200] Train Loss: 0.7409, Train Acc: 77.94% | Test Loss: 1.6238, Test Acc: 60.37% | LR: 0.100000 | Time: 44.6s
+Epoch [99/200] Train Loss: 0.7403, Train Acc: 78.05% | Test Loss: 1.5009, Test Acc: 62.69% | LR: 0.100000 | Time: 45.0s
+Epoch [100/200] Train Loss: 0.7369, Train Acc: 77.95% | Test Loss: 1.4428, Test Acc: 61.71% | LR: 0.010000 | Time: 44.9s
+Epoch [101/200] Train Loss: 0.3328, Train Acc: 90.41% | Test Loss: 0.7579, Test Acc: 78.75% | LR: 0.010000 | Time: 44.9s
+Epoch [102/200] Train Loss: 0.2021, Train Acc: 94.34% | Test Loss: 0.7470, Test Acc: 79.23% | LR: 0.010000 | Time: 45.0s
+Epoch [103/200] Train Loss: 0.1572, Train Acc: 95.86% | Test Loss: 0.7529, Test Acc: 79.66% | LR: 0.010000 | Time: 45.0s
+Epoch [104/200] Train Loss: 0.1267, Train Acc: 96.84% | Test Loss: 0.7633, Test Acc: 79.46% | LR: 0.010000 | Time: 44.9s
+Epoch [105/200] Train Loss: 0.1038, Train Acc: 97.52% | Test Loss: 0.7766, Test Acc: 79.43% | LR: 0.010000 | Time: 44.8s
+Epoch [106/200] Train Loss: 0.0861, Train Acc: 98.13% | Test Loss: 0.7881, Test Acc: 79.70% | LR: 0.010000 | Time: 44.9s
+Epoch [107/200] Train Loss: 0.0733, Train Acc: 98.48% | Test Loss: 0.7924, Test Acc: 79.27% | LR: 0.010000 | Time: 45.0s
+Epoch [108/200] Train Loss: 0.0628, Train Acc: 98.78% | Test Loss: 0.7896, Test Acc: 79.73% | LR: 0.010000 | Time: 44.9s
+Epoch [109/200] Train Loss: 0.0552, Train Acc: 99.01% | Test Loss: 0.7937, Test Acc: 79.83% | LR: 0.010000 | Time: 44.7s
+Epoch [110/200] Train Loss: 0.0472, Train Acc: 99.27% | Test Loss: 0.8089, Test Acc: 79.72% | LR: 0.010000 | Time: 45.1s
+Epoch [111/200] Train Loss: 0.0421, Train Acc: 99.32% | Test Loss: 0.8139, Test Acc: 79.68% | LR: 0.010000 | Time: 44.9s
+Epoch [112/200] Train Loss: 0.0391, Train Acc: 99.43% | Test Loss: 0.8124, Test Acc: 79.80% | LR: 0.010000 | Time: 44.8s
+Epoch [113/200] Train Loss: 0.0351, Train Acc: 99.52% | Test Loss: 0.8067, Test Acc: 79.91% | LR: 0.010000 | Time: 45.1s
+Epoch [114/200] Train Loss: 0.0310, Train Acc: 99.62% | Test Loss: 0.8262, Test Acc: 79.81% | LR: 0.010000 | Time: 45.0s
+Epoch [115/200] Train Loss: 0.0300, Train Acc: 99.61% | Test Loss: 0.8221, Test Acc: 79.65% | LR: 0.010000 | Time: 44.7s
+Epoch [116/200] Train Loss: 0.0294, Train Acc: 99.61% | Test Loss: 0.8166, Test Acc: 79.80% | LR: 0.010000 | Time: 45.0s
+Epoch [117/200] Train Loss: 0.0272, Train Acc: 99.71% | Test Loss: 0.8259, Test Acc: 79.47% | LR: 0.010000 | Time: 44.9s
+Epoch [118/200] Train Loss: 0.0251, Train Acc: 99.74% | Test Loss: 0.8297, Test Acc: 79.89% | LR: 0.010000 | Time: 44.8s
+Epoch [119/200] Train Loss: 0.0227, Train Acc: 99.80% | Test Loss: 0.8145, Test Acc: 80.22% | LR: 0.010000 | Time: 44.9s
+Epoch [120/200] Train Loss: 0.0254, Train Acc: 99.71% | Test Loss: 0.8241, Test Acc: 80.07% | LR: 0.010000 | Time: 44.9s
+Epoch [121/200] Train Loss: 0.0234, Train Acc: 99.78% | Test Loss: 0.8329, Test Acc: 79.60% | LR: 0.010000 | Time: 44.9s
+Epoch [122/200] Train Loss: 0.0253, Train Acc: 99.70% | Test Loss: 0.8425, Test Acc: 79.44% | LR: 0.010000 | Time: 45.1s
+Epoch [123/200] Train Loss: 0.0234, Train Acc: 99.76% | Test Loss: 0.8405, Test Acc: 79.39% | LR: 0.010000 | Time: 44.9s
+Epoch [124/200] Train Loss: 0.0234, Train Acc: 99.75% | Test Loss: 0.8282, Test Acc: 79.50% | LR: 0.010000 | Time: 44.7s
+Epoch [125/200] Train Loss: 0.0236, Train Acc: 99.74% | Test Loss: 0.8477, Test Acc: 79.29% | LR: 0.010000 | Time: 45.0s
+Epoch [126/200] Train Loss: 0.0293, Train Acc: 99.59% | Test Loss: 0.8431, Test Acc: 79.36% | LR: 0.010000 | Time: 44.9s
+Epoch [127/200] Train Loss: 0.0261, Train Acc: 99.69% | Test Loss: 0.8391, Test Acc: 79.22% | LR: 0.010000 | Time: 44.8s
+Epoch [128/200] Train Loss: 0.0236, Train Acc: 99.76% | Test Loss: 0.8223, Test Acc: 79.98% | LR: 0.010000 | Time: 45.0s
+Epoch [129/200] Train Loss: 0.0281, Train Acc: 99.63% | Test Loss: 0.8501, Test Acc: 78.91% | LR: 0.010000 | Time: 44.9s
+Epoch [130/200] Train Loss: 0.0269, Train Acc: 99.67% | Test Loss: 0.8546, Test Acc: 79.20% | LR: 0.010000 | Time: 45.0s
+Epoch [131/200] Train Loss: 0.0283, Train Acc: 99.66% | Test Loss: 0.8738, Test Acc: 78.55% | LR: 0.010000 | Time: 44.8s
+Epoch [132/200] Train Loss: 0.0296, Train Acc: 99.63% | Test Loss: 0.8865, Test Acc: 78.38% | LR: 0.010000 | Time: 44.9s
+Epoch [133/200] Train Loss: 0.0311, Train Acc: 99.61% | Test Loss: 0.8661, Test Acc: 78.93% | LR: 0.010000 | Time: 44.9s
+Epoch [134/200] Train Loss: 0.0301, Train Acc: 99.59% | Test Loss: 0.8746, Test Acc: 78.97% | LR: 0.010000 | Time: 45.1s
+Epoch [135/200] Train Loss: 0.0382, Train Acc: 99.44% | Test Loss: 0.8479, Test Acc: 78.61% | LR: 0.010000 | Time: 44.9s
+Epoch [136/200] Train Loss: 0.0423, Train Acc: 99.33% | Test Loss: 0.9072, Test Acc: 77.69% | LR: 0.010000 | Time: 44.9s
+Epoch [137/200] Train Loss: 0.0400, Train Acc: 99.33% | Test Loss: 0.8955, Test Acc: 78.07% | LR: 0.010000 | Time: 45.0s
+Epoch [138/200] Train Loss: 0.0435, Train Acc: 99.25% | Test Loss: 0.9308, Test Acc: 77.14% | LR: 0.010000 | Time: 44.8s
+Epoch [139/200] Train Loss: 0.0438, Train Acc: 99.26% | Test Loss: 0.8984, Test Acc: 77.95% | LR: 0.010000 | Time: 44.9s
+Epoch [140/200] Train Loss: 0.0441, Train Acc: 99.26% | Test Loss: 0.9217, Test Acc: 77.19% | LR: 0.010000 | Time: 44.9s
+Epoch [141/200] Train Loss: 0.0453, Train Acc: 99.23% | Test Loss: 0.9352, Test Acc: 77.44% | LR: 0.010000 | Time: 44.9s
+Epoch [142/200] Train Loss: 0.0555, Train Acc: 98.94% | Test Loss: 0.9568, Test Acc: 76.63% | LR: 0.010000 | Time: 44.7s
+Epoch [143/200] Train Loss: 0.0600, Train Acc: 98.79% | Test Loss: 0.9447, Test Acc: 76.35% | LR: 0.010000 | Time: 45.0s
+Epoch [144/200] Train Loss: 0.0625, Train Acc: 98.79% | Test Loss: 0.9626, Test Acc: 76.67% | LR: 0.010000 | Time: 44.8s
+Epoch [145/200] Train Loss: 0.0634, Train Acc: 98.70% | Test Loss: 1.0035, Test Acc: 75.66% | LR: 0.010000 | Time: 44.8s
+Epoch [146/200] Train Loss: 0.0721, Train Acc: 98.45% | Test Loss: 0.9892, Test Acc: 75.96% | LR: 0.010000 | Time: 45.1s
+Epoch [147/200] Train Loss: 0.0739, Train Acc: 98.37% | Test Loss: 1.0425, Test Acc: 74.90% | LR: 0.010000 | Time: 44.9s
+Epoch [148/200] Train Loss: 0.0710, Train Acc: 98.48% | Test Loss: 1.0720, Test Acc: 74.12% | LR: 0.010000 | Time: 44.9s
+Epoch [149/200] Train Loss: 0.0691, Train Acc: 98.52% | Test Loss: 1.0322, Test Acc: 75.64% | LR: 0.010000 | Time: 44.9s
+Epoch [150/200] Train Loss: 0.0709, Train Acc: 98.45% | Test Loss: 1.0078, Test Acc: 75.43% | LR: 0.001000 | Time: 44.8s
+Epoch [151/200] Train Loss: 0.0319, Train Acc: 99.53% | Test Loss: 0.8278, Test Acc: 79.33% | LR: 0.001000 | Time: 44.9s
+Epoch [152/200] Train Loss: 0.0164, Train Acc: 99.86% | Test Loss: 0.8167, Test Acc: 79.95% | LR: 0.001000 | Time: 44.9s
+Epoch [153/200] Train Loss: 0.0134, Train Acc: 99.88% | Test Loss: 0.8175, Test Acc: 79.86% | LR: 0.001000 | Time: 44.9s
+Epoch [154/200] Train Loss: 0.0117, Train Acc: 99.90% | Test Loss: 0.8095, Test Acc: 80.09% | LR: 0.001000 | Time: 45.0s
+Epoch [155/200] Train Loss: 0.0099, Train Acc: 99.94% | Test Loss: 0.8089, Test Acc: 80.18% | LR: 0.001000 | Time: 45.0s
+Epoch [156/200] Train Loss: 0.0094, Train Acc: 99.93% | Test Loss: 0.8105, Test Acc: 80.34% | LR: 0.001000 | Time: 44.9s
+Epoch [157/200] Train Loss: 0.0085, Train Acc: 99.95% | Test Loss: 0.8116, Test Acc: 80.46% | LR: 0.001000 | Time: 44.8s
+Epoch [158/200] Train Loss: 0.0084, Train Acc: 99.95% | Test Loss: 0.8080, Test Acc: 80.63% | LR: 0.001000 | Time: 44.9s
+Epoch [159/200] Train Loss: 0.0079, Train Acc: 99.96% | Test Loss: 0.8108, Test Acc: 80.34% | LR: 0.001000 | Time: 44.9s
+Epoch [160/200] Train Loss: 0.0075, Train Acc: 99.95% | Test Loss: 0.8093, Test Acc: 80.51% | LR: 0.001000 | Time: 44.9s
+Epoch [161/200] Train Loss: 0.0072, Train Acc: 99.96% | Test Loss: 0.8053, Test Acc: 80.45% | LR: 0.001000 | Time: 45.0s
+Epoch [162/200] Train Loss: 0.0068, Train Acc: 99.96% | Test Loss: 0.8102, Test Acc: 80.57% | LR: 0.001000 | Time: 44.9s
+Epoch [163/200] Train Loss: 0.0069, Train Acc: 99.95% | Test Loss: 0.8074, Test Acc: 80.55% | LR: 0.001000 | Time: 44.9s
+Epoch [164/200] Train Loss: 0.0065, Train Acc: 99.97% | Test Loss: 0.8064, Test Acc: 80.45% | LR: 0.001000 | Time: 44.9s
+Epoch [165/200] Train Loss: 0.0064, Train Acc: 99.95% | Test Loss: 0.8052, Test Acc: 80.64% | LR: 0.001000 | Time: 44.9s
+Epoch [166/200] Train Loss: 0.0064, Train Acc: 99.96% | Test Loss: 0.8056, Test Acc: 80.51% | LR: 0.001000 | Time: 44.9s
+Epoch [167/200] Train Loss: 0.0061, Train Acc: 99.97% | Test Loss: 0.8064, Test Acc: 80.46% | LR: 0.001000 | Time: 44.9s
+Epoch [168/200] Train Loss: 0.0061, Train Acc: 99.97% | Test Loss: 0.8035, Test Acc: 80.59% | LR: 0.001000 | Time: 44.8s
+Epoch [169/200] Train Loss: 0.0060, Train Acc: 99.98% | Test Loss: 0.7973, Test Acc: 80.91% | LR: 0.001000 | Time: 44.8s
+Epoch [170/200] Train Loss: 0.0059, Train Acc: 99.98% | Test Loss: 0.8002, Test Acc: 80.65% | LR: 0.001000 | Time: 44.8s
+Epoch [171/200] Train Loss: 0.0058, Train Acc: 99.97% | Test Loss: 0.8065, Test Acc: 80.50% | LR: 0.001000 | Time: 44.8s
+Epoch [172/200] Train Loss: 0.0057, Train Acc: 99.98% | Test Loss: 0.8017, Test Acc: 80.78% | LR: 0.001000 | Time: 45.0s
+Epoch [173/200] Train Loss: 0.0057, Train Acc: 99.96% | Test Loss: 0.7986, Test Acc: 80.72% | LR: 0.001000 | Time: 45.0s
+Epoch [174/200] Train Loss: 0.0057, Train Acc: 99.96% | Test Loss: 0.7983, Test Acc: 80.83% | LR: 0.001000 | Time: 44.9s
+Epoch [175/200] Train Loss: 0.0053, Train Acc: 99.97% | Test Loss: 0.7980, Test Acc: 80.93% | LR: 0.001000 | Time: 44.9s
+Epoch [176/200] Train Loss: 0.0055, Train Acc: 99.97% | Test Loss: 0.7985, Test Acc: 80.82% | LR: 0.001000 | Time: 44.9s
+Epoch [177/200] Train Loss: 0.0055, Train Acc: 99.97% | Test Loss: 0.7972, Test Acc: 80.94% | LR: 0.001000 | Time: 44.9s
+Epoch [178/200] Train Loss: 0.0052, Train Acc: 99.98% | Test Loss: 0.7952, Test Acc: 80.79% | LR: 0.001000 | Time: 44.9s
+Epoch [179/200] Train Loss: 0.0053, Train Acc: 99.97% | Test Loss: 0.8015, Test Acc: 80.79% | LR: 0.001000 | Time: 44.9s
+Epoch [180/200] Train Loss: 0.0053, Train Acc: 99.97% | Test Loss: 0.7948, Test Acc: 80.82% | LR: 0.001000 | Time: 44.9s
+Epoch [181/200] Train Loss: 0.0050, Train Acc: 99.97% | Test Loss: 0.7988, Test Acc: 80.91% | LR: 0.001000 | Time: 44.9s
+Epoch [182/200] Train Loss: 0.0051, Train Acc: 99.97% | Test Loss: 0.7972, Test Acc: 80.99% | LR: 0.001000 | Time: 44.9s
+Epoch [183/200] Train Loss: 0.0053, Train Acc: 99.97% | Test Loss: 0.7954, Test Acc: 80.93% | LR: 0.001000 | Time: 44.9s
+Epoch [184/200] Train Loss: 0.0054, Train Acc: 99.97% | Test Loss: 0.7942, Test Acc: 81.05% | LR: 0.001000 | Time: 44.9s
+Epoch [185/200] Train Loss: 0.0053, Train Acc: 99.97% | Test Loss: 0.8007, Test Acc: 80.70% | LR: 0.001000 | Time: 44.9s
+Epoch [186/200] Train Loss: 0.0054, Train Acc: 99.97% | Test Loss: 0.7921, Test Acc: 80.91% | LR: 0.001000 | Time: 44.9s
+Epoch [187/200] Train Loss: 0.0053, Train Acc: 99.96% | Test Loss: 0.7964, Test Acc: 80.82% | LR: 0.001000 | Time: 44.9s
+Epoch [188/200] Train Loss: 0.0053, Train Acc: 99.98% | Test Loss: 0.7995, Test Acc: 80.89% | LR: 0.001000 | Time: 44.9s
+Epoch [189/200] Train Loss: 0.0054, Train Acc: 99.97% | Test Loss: 0.7967, Test Acc: 80.65% | LR: 0.001000 | Time: 44.9s
+Epoch [190/200] Train Loss: 0.0052, Train Acc: 99.97% | Test Loss: 0.7954, Test Acc: 80.72% | LR: 0.001000 | Time: 44.9s
+Epoch [191/200] Train Loss: 0.0051, Train Acc: 99.97% | Test Loss: 0.7927, Test Acc: 80.82% | LR: 0.001000 | Time: 44.9s
+Epoch [192/200] Train Loss: 0.0049, Train Acc: 99.98% | Test Loss: 0.7932, Test Acc: 80.95% | LR: 0.001000 | Time: 44.9s
+Epoch [193/200] Train Loss: 0.0052, Train Acc: 99.98% | Test Loss: 0.7934, Test Acc: 80.93% | LR: 0.001000 | Time: 44.9s
+Epoch [194/200] Train Loss: 0.0051, Train Acc: 99.98% | Test Loss: 0.7910, Test Acc: 80.85% | LR: 0.001000 | Time: 44.9s
+Epoch [195/200] Train Loss: 0.0051, Train Acc: 99.98% | Test Loss: 0.7888, Test Acc: 80.90% | LR: 0.001000 | Time: 44.9s
+Epoch [196/200] Train Loss: 0.0052, Train Acc: 99.97% | Test Loss: 0.7897, Test Acc: 80.97% | LR: 0.001000 | Time: 44.9s
+Epoch [197/200] Train Loss: 0.0053, Train Acc: 99.97% | Test Loss: 0.7904, Test Acc: 80.95% | LR: 0.001000 | Time: 44.9s
+Epoch [198/200] Train Loss: 0.0052, Train Acc: 99.98% | Test Loss: 0.7842, Test Acc: 80.86% | LR: 0.001000 | Time: 44.9s
+Epoch [199/200] Train Loss: 0.0050, Train Acc: 99.97% | Test Loss: 0.7850, Test Acc: 80.89% | LR: 0.001000 | Time: 44.9s
+Epoch [200/200] Train Loss: 0.0051, Train Acc: 99.97% | Test Loss: 0.7860, Test Acc: 81.07% | LR: 0.001000 | Time: 44.9s

provenance/cifar100_training_summary.txt

@@ -0,0 +1,23 @@
+
+Training Configuration Summary
+==============================
+Model: WideResNet-28-10
+Dataset: CIFAR-100
+Training Type: Standard (Natural)
+
+Hyperparameters:
+- Optimizer: SGD (momentum=0.9, weight_decay=0.0005)
+- Initial LR: 0.1
+- LR Schedule: MultiStepLR at [100, 150] with gamma=0.1
+- Epochs: 200
+- Batch size: 128
+- Augmentation: RandomCrop(32, padding=4) + RandomHorizontalFlip
+- No label smoothing, no mixup/cutmix, no EMA
+
+Results:
+- Final Test Accuracy: 81.07%
+- Training Time: 2.50 hours
+
+Output:
+- Model: /root/robustbench/models/cifar100/Linf/Standard_WRN28_10.pt
+- Log: /root/robustbench/models/cifar100/Linf/training_log.txt

provenance/cifar10_training_summary.json

@@ -0,0 +1,49 @@
+{
+  "recipe": {
+    "models": [
+      "resnet18",
+      "vgg16",
+      "resnet50",
+      "densenet121"
+    ],
+    "data_dir": "/root/autodl-tmp/data",
+    "out_dir": "/root/autodl-tmp/models/cifar10/standard",
+    "epochs": 200,
+    "batch_size": 128,
+    "lr": 0.1,
+    "wd": 0.0005,
+    "milestones": [
+      100,
+      150
+    ],
+    "min_acc": 88.0,
+    "workers": 8,
+    "device": "cuda:0"
+  },
+  "models": {
+    "resnet18": {
+      "clean_acc": 94.92,
+      "best": 94.97,
+      "minutes": 31.17216245730718,
+      "path": "/root/autodl-tmp/models/cifar10/standard/c10_resnet18.pt"
+    },
+    "vgg16": {
+      "clean_acc": 93.82,
+      "best": 93.9,
+      "minutes": 21.685336538155873,
+      "path": "/root/autodl-tmp/models/cifar10/standard/c10_vgg16.pt"
+    },
+    "resnet50": {
+      "clean_acc": 94.66,
+      "best": 94.66,
+      "minutes": 68.99354430437089,
+      "path": "/root/autodl-tmp/models/cifar10/standard/c10_resnet50.pt"
+    },
+    "densenet121": {
+      "clean_acc": 95.39,
+      "best": 95.39,
+      "minutes": 110.7822703520457,
+      "path": "/root/autodl-tmp/models/cifar10/standard/c10_densenet121.pt"
+    }
+  }
+}