I'm starting to use shap values with pytorch. For this, I just tried to run an easy example but I get an error. The code is:
import numpy as np
import torch
from torch import nn, optim
from torch.nn import functional as F
from torchvision import datasets, transforms
import shap
batch_size = 128
num_epochs = 2
device = torch.device("cpu")
class Net(nn.Module):
def __init__(self):
super().__init__()
self.conv_layers = nn.Sequential(
nn.Conv2d(1, 10, kernel_size=5),
nn.MaxPool2d(2),
nn.ReLU(),
nn.Conv2d(10, 20, kernel_size=5),
nn.Dropout(),
nn.MaxPool2d(2),
nn.ReLU(),
)
self.fc_layers = nn.Sequential(
nn.Linear(320, 50),
nn.ReLU(),
nn.Dropout(),
nn.Linear(50, 10),
nn.Softmax(dim=1),
)
def forward(self, x):
x = self.conv_layers(x)
x = x.view(-1, 320)
x = self.fc_layers(x)
return x
def train(model, device, train_loader, optimizer, epoch):
model.train()
for batch_idx, (data, target) in enumerate(train_loader):
data, target = data.to(device), target.to(device)
optimizer.zero_grad()
output = model(data)
loss = F.nll_loss(output.log(), target)
loss.backward()
optimizer.step()
if batch_idx % 100 == 0:
print(
f"Train Epoch: {epoch} [{batch_idx * len(data)}/{len(train_loader.dataset)}"
f" ({100.0 * batch_idx / len(train_loader):.0f}%)]"
f"\tLoss: {loss.item():.6f}"
)
def test(model, device, test_loader):
model.eval()
test_loss = 0
correct = 0
with torch.no_grad():
for data, target in test_loader:
data, target = data.to(device), target.to(device)
output = model(data)
test_loss += F.nll_loss(output.log(), target).item() # sum up batch loss
pred = output.max(1, keepdim=True)[
1
] # get the index of the max log-probability
correct += pred.eq(target.view_as(pred)).sum().item()
test_loss /= len(test_loader.dataset)
print(
f"\nTest set: Average loss: {test_loss:.4f},"
f" Accuracy: {correct}/{len(test_loader.dataset)}"
f" ({100.0 * correct / len(test_loader.dataset):.0f}%)\n"
)
train_loader = torch.utils.data.DataLoader(
datasets.MNIST(
"mnist_data",
train=True,
download=True,
transform=transforms.Compose([transforms.ToTensor()]),
),
batch_size=batch_size,
shuffle=True,
)
test_loader = torch.utils.data.DataLoader(
datasets.MNIST(
"mnist_data", train=False, transform=transforms.Compose([transforms.ToTensor()])
),
batch_size=batch_size,
shuffle=True,
)
model = Net().to(device)
optimizer = optim.SGD(model.parameters(), lr=0.01, momentum=0.5)
for epoch in range(1, num_epochs + 1):
train(model, device, train_loader, optimizer, epoch)
test(model, device, test_loader)
# since shuffle=True, this is a random sample of test data
batch = next(iter(test_loader))
images, _ = batch
#images = images.view(-1, 1, 28, 28)
background = images[:100]
test_images = images[100:110]
e = shap.DeepExplainer(model, background)
shap_values = e.shap_values(test_images)
shap_numpy = [np.swapaxes(np.swapaxes(s, 1, -1), 1, 2) for s in shap_values]
test_numpy = np.swapaxes(np.swapaxes(test_images.numpy(), 1, -1), 1, 2)
# plot the feature attributions
shap.image_plot(shap_numpy, -test_numpy)
The error is:
Traceback (most recent call last):
Cell In[5], line 5 shap.image_plot(shap_numpy, -test_numpy)
File ~\anaconda3\lib\site-packages\shap\plots_image.py:154 in image if len(shap_values[0][row].shape) == 2:
IndexError: index 1 is out of bounds for axis 0 with size 1
As I've said, I am new with all this and I don't know how to fix this error.
The problem is in shap_numpy elements shape.
If you check shap_numpy[0].shape you will get (1, 28, 10, 28), while shap.image_plot expect N images (i.e. shape = (N, width, length, channels)). Since you are working with 28x28 grayscale images, the desired shape is (10,28,28,1).
Try to change
shap_numpy = [np.swapaxes(np.swapaxes(s, 1, -1), 1, 2) for s in shap_values]
To
shap_numpy = [np.swapaxes(s, 0, -1) for s in shap_values]