Error - RuntimeError: a view of a leaf Variable that requires grad is being used in an in-place operation.
Cause - In-place operation was attempted on a tensor that is a "leaf" variable (a tensor that requires gradients and is at the start of the computation graph
Fix - output = [o.detach().clone() for o in output]
explanation - After applying detach() and clone(), the new tensor is not part of the computation graph and does not require gradients. Therefore, it is safe to perform in-place operations on it without affecting the gradient computations or causing errors
N150 - golden
E150,E300 -golden, E150,E300 -silicon
Failed jobs and it's link
customer-pytorch-ssd300-resnet50-ssd300-resnet50-pytorch-gs-e150-golden - Link
customer-pytorch-ssd300-resnet50-ssd300-resnet50-pytorch-gs-e300-golden - Link
customer-pytorch-ssd300-resnet50-ssd300-resnet50-pytorch-gs-e150-silicon - Link
customer-pytorch-ssd300-resnet50-ssd300-resnet50-pytorch-gs-e300-silicon - Link
Error - pipegen error
Fix - override problematic op with low t-stream shape
After fix -
E150_compute.log
E150_silicon.log
E300_compute.log
E300_silicon.log