Open
Description
cbloq_to_quimb is supposed to take a composite bloq (result of flattening a circuit) and quickly turn it into a tensor network, which may be expensive to contract. It hangs when a bloq decomposes to AddIntoPhaseGrad.
AddIntoPhaseGrad tries to make a big unitary as part of my_tensors. It's not really supposed to do this.
I tried modifying the _my_tensors_from_gate helper to give up if the number of qubits is too high; but it still hangs computing cirq.unitary(..) of the AddIntoPhaseGrad.