-
Hi, thank you for the nice implementations!
I have a question on running l2 norm application with data normalization.
While data is normalized, it seems like the perturbation \epsilon is not scale…
-
您好！
我想用‘simple_verification.py’中代码计算的lp和Up实现您在CROWN那篇论文中实现的二分查找法，并尝试更改如下：
def test(input,model):
eps = 0
gap_gx = 100
eps_LB = -1
eps_UB = 1
counter = 0
is_pos = True
…
-
Hi there,
I added a BatchNorm2d layer in the model and auto-LiRPA failed to parse the graph. Is it a bug?
```console
$ python simple_training.py --num_epochs 3 --scheduler_opts "start=2,length=…
-
Do you do last layer elision similar to the original IBP? I saw some references to `C_merged`, is it doing something similar?
if yes it'd be great if you could describe that bit of the code a littl…
-
Hi,
Recently I am using your auto_LiRPA to compute the perturbation bound of hidden features. However, I found that the ub and lb are always exactly the same.
```
eps = 0.012
actions = actions.c…
-
Hi,
Thanks so much for sharing the pytorch version CROWN-IBP code. I am really excited about the excellent result (7.02 certified accuracy) on MNIST dataset. However, after a simple try, I have som…
-
I wish something like this could be supported in Armory, so that we don't have to create many configuration files in the development phase.
```bash
$ armory run scenario_configs/baseline.json data…
-
Hi there,
I was trying to reproduce Table 2 in https://openreview.net/pdf?id=Skxuk1rFwB
I was able to produce the same results as in CROWN-IBP on three models, but failed on this one:
```cons…