Does torchscript optimize models automatically to execute in parallel?

jit
#1

Suppose I have a model like this

class Model(nn.Module):
  def __init__(self):
    self.part1 = nn.Linear(10, 10)
    self.part2 = nn.Linear(10, 10)
  def forward(self, x):
    x1 = self.part1(x)
    x2 = self.part2(x)
   return x1 + x2

In eager mode, I’m sure it’s not executed in parallel but when I torch.jit.script the model, does torchscript automatically turn them into to async?
If so does the optimization depend on device being cpu or cuda?

Torch.jit.fork makes code slower
#2

No, the ops are executed serially, but there may be intra-op parallelism.

1 Like