About applying a three-qubit unitary gate onto a circuit

[talentkeychen]

What is your issue?

Here is the issue. I am still new to quimb. I am trying to apply a three-qubit unitary gate onto my circuit. The unitary gate is constructed by writing a 8 by 8 unitary matrix manually called 'mat' and its type is a numpy array, and make it as an operator:

op = qu.core.qu(mat, dtype=complex, sparse=True)

Now when I try to apply it to my circuit using apply apply_gate_raw:

circ.apply_gate_raw(op,[0,1,2])

I get the error message as follows:

---

ValueError Traceback (most recent call last) Input In [27], in <cell line: 4>() 2 op = transform_to_tensor(nhssh_mat,complex) 3 index_list = [0,1,2] ----> 4 add_dissipation_oracle(circ,op,0,1,2)

Input In [25], in add_dissipation_oracle(circ, operator, phy_site1, phy_site2, ancl_site) 8 def add_dissipation_oracle(circ, 9 operator, 10 phy_site1, 11 phy_site2, 12 ancl_site): 13 index_list = [phy_site1,phy_site2,ancl_site] ---> 14 circ.apply_gate_raw(operator,index_list)

File ~\anaconda3\envs\quimb\lib\site-packages\quimb\tensor\circuit.py:822, in Circuit.apply_gate_raw(self, U, where, tags, gate_round, gateopts) 820 tags.add(f'ROUND{gate_round}') 821 opts = {self.gate_opts, gate_opts} --> 822 self.psi.gate(U, where, tags=tags, opts) 823 self.gates.append((id(U), *where))

File ~\anaconda3\envs\quimb\lib\site-packages\quimb\tensor\tensor_1d.py:722, in TensorNetwork1DVector.gate(self, inplace, *args, kwargs) 720 @functools.wraps(gate_TN_1D) 721 def gate(self, *args, inplace=False, *kwargs): --> 722 return gate_TN_1D(self, args, inplace=inplace, kwargs)

File ~\anaconda3\envs\quimb\lib\site-packages\quimb\tensor\tensor_1d.py:219, in gate_TN_1D(tn, G, where, contract, tags, propagate_tags, inplace, cur_orthog, **compress_opts) 216 tags = tags_to_oset(tags) 218 if (ng > 2) and contract in _TWO_BODY_ONLY: --> 219 raise ValueError(f"Can't use contract='{contract}' for >2 sites.") 221 G = maybe_factor_gate_into_tensor(G, dp, ng, where) 223 if contract == 'swap+split' and ng > 1:

ValueError: Can't use contract='auto-split-gate' for >2 sites.

Is there a solution for it? Thank you.

[jcmgray]

Yes for the moment the 'auto-split-gate' method should probably just ignore 3+ qubit gates rather than trying raising (ideally it would find a low rank decomposition).

As a workaround you should be able to supply contract=False when calling apply_gate_raw.

[talentkeychen]

Thank you for your answer! May I update that this solution seems to work well, but only if I supply sparse=False in:

op = qu.core.qu(mat, dtype=complex, sparse=False)

In my previous code shown above, if I supply sparse=True, then I will get the following error:

---

ValueError Traceback (most recent call last) Input In [27], in <cell line: 1>() ----> 1 circ.apply_gate_raw(op,[0,1,3],contract=False)

File ~/opt/anaconda3/envs/quimb/lib/python3.10/site-packages/quimb/tensor/circuit.py:822, in Circuit.apply_gate_raw(self, U, where, tags, gate_round, gateopts) 820 tags.add(f'ROUND{gate_round}') 821 opts = {self.gate_opts, gate_opts} --> 822 self.psi.gate(U, where, tags=tags, opts) 823 self.gates.append((id(U), *where))

File ~/opt/anaconda3/envs/quimb/lib/python3.10/site-packages/quimb/tensor/tensor_1d.py:722, in TensorNetwork1DVector.gate(self, inplace, *args, kwargs) 720 @functools.wraps(gate_TN_1D) 721 def gate(self, *args, inplace=False, *kwargs): --> 722 return gate_TN_1D(self, args, inplace=inplace, kwargs)

File ~/opt/anaconda3/envs/quimb/lib/python3.10/site-packages/quimb/tensor/tensor_1d.py:221, in gate_TN_1D(tn, G, where, contract, tags, propagate_tags, inplace, cur_orthog, compress_opts) 218 if (ng > 2) and contract in _TWO_BODY_ONLY: 219 raise ValueError(f"Can't use contract='{contract}' for >2 sites.") --> 221 G = maybe_factor_gate_into_tensor(G, dp, ng, where) 223 if contract == 'swap+split' and ng > 1: 224 psi.gate_with_auto_swap(G, where, cur_orthog=cur_orthog, 225 inplace=True, compress_opts)

File ~/opt/anaconda3/envs/quimb/lib/python3.10/site-packages/quimb/tensor/tensor_1d.py:94, in maybe_factor_gate_into_tensor(G, dp, ng, where) 92 G = ops.asarray(G) 93 if ng >= 2: ---> 94 G = reshape(G, [dp] 2 ng) 96 elif not shape_matches_nd: 97 raise ValueError( 98 f"Gate with shape {G.shape} doesn't match sites {where}.")

File ~/opt/anaconda3/envs/quimb/lib/python3.10/site-packages/autoray/autoray.py:300, in reshape(x, shape) 298 """Array reshaped.""" 299 try: --> 300 return x.reshape(shape) 301 except AttributeError: 302 return do("reshape", x, shape)

File ~/opt/anaconda3/envs/quimb/lib/python3.10/site-packages/scipy/sparse/base.py:119, in spmatrix.reshape(self, *args, **kwargs) 87 """reshape(self, shape, order='C', copy=False) 88 89 Gives a new shape to a sparse matrix without changing its data. (...) 115 matrices 116 """ 117 # If the shape already matches, don't bother doing an actual reshape 118 # Otherwise, the default is to convert to COO and use its reshape --> 119 shape = check_shape(args, self.shape) 120 order, copy = check_reshape_kwargs(kwargs) 121 if shape == self.shape:

File ~/opt/anaconda3/envs/quimb/lib/python3.10/site-packages/scipy/sparse/sputils.py:322, in check_shape(args, current_shape) 319 raise ValueError('can only specify one unknown dimension') 321 if len(new_shape) != 2: --> 322 raise ValueError('matrix shape must be two-dimensional') 324 return new_shape

ValueError: matrix shape must be two-dimensional

[jcmgray]

Glad it works! Yes, you will need to supply the gates as ndarray like objects, scipy sparse matrices are two-dimensional only, and their purpose somewhat orthogonal to the tensor network functionality.

[jcmgray]

Three-qubit gates should be handled fine by both Circuit and CircuitMPS now, so closing for the moment.