Fix bug causing slices with depth > 1 to be forward-prop'd #62
Conversation
|
This is a bit tangential, but in adjusting the unit tests I realized this line should maybe be |
Pull Request Test Coverage Report for Build 13503277786Details
💛 - Coveralls |
I agree, it should be |
| @@ -164,7 +164,8 @@ def handle_timeout(signum, frame): | |||
| # PERF: we will likely need to parallelize this loop | |||
| for i in range(num_observables): | |||
| non_trivial_slice = False | |||
| for op_idx, op_node in enumerate(circuit_to_dag(slice_).topological_op_nodes()): | |||
| op_nodes = reversed(list(circuit_to_dag(slice_).topological_op_nodes())) | |||
There was a problem hiding this comment.
Must we explicitly convert the iterator to a list? Or can we do the following instead?
| op_nodes = reversed(list(circuit_to_dag(slice_).topological_op_nodes())) | |
| op_nodes = reversed(circuit_to_dag(slice_).topological_op_nodes()) |
| @@ -164,7 +164,8 @@ def handle_timeout(signum, frame): | |||
| # PERF: we will likely need to parallelize this loop | |||
| for i in range(num_observables): | |||
| non_trivial_slice = False | |||
| for op_idx, op_node in enumerate(circuit_to_dag(slice_).topological_op_nodes()): | |||
| op_nodes = reversed(list(circuit_to_dag(slice_).topological_op_nodes())) | |||
| for op_idx, op_node in enumerate(op_nodes): | |||
There was a problem hiding this comment.
Since this op_idx now indexes the gates in the layer in reverse, I wonder whether we should update the log message where this gets used. Just to be super clear about this.
| --- | ||
| fixes: | ||
| - | | ||
| Fixed a bug in :func:`qiskit_addon_obp.backpropagate` which caused slices with depth > 1 to be incorrectly forward-propagated. |
There was a problem hiding this comment.
Can we be more explicit and state that gates within a slice will from now on be backpropagated properly, that is, in reverse order.
Fixes #60