with_detect_anomaly function

Context-manager that enable anomaly detection for the autograd engine.

This does two things:

with_detect_anomaly(code)

Arguments

• code: Code that will be executed in the detect anomaly context.

Details

• Running the forward pass with detection enabled will allow the backward pass to print the traceback of the forward operation that created the failing backward function.
• Any backward computation that generate "nan" value will raise an error.

Warning

This mode should be enabled only for debugging as the different tests will slow down your program execution.

Examples

if (torch_is_installed()) {
x <- torch_randn(2, requires_grad = TRUE)
y <- torch_randn(1)
b <- (x^y)$sum()
y$add_(1)

try({
  b$backward()

  with_detect_anomaly({
    b$backward()
  })
})
}