Adding dmm config and modulation to sequence

[a-corni]

In #565:

• we introduced Traps to define RegisterLayout and WeightMap.
• Write unit tests that don't depend on changes in Sequence.

In #576:

• We added the possibility to define slug in define_detuning_map of Register, MappableRegister, RegisterLayout

Here:

• Configure DMM using config_det_map (impossible to declare a DMM using declare_channel
• Configure SLM using config_slm_mask:
  • if in xy: it works as before
  • if in ising: it configures a DMM and either modulates the DMM from the first pulse found on the first declared Global channel having one, or (if no such pulse has been defined) the DMM will be modulated from the first pulse added to a Global Channel.
  • If in neither of them, the dmm to use for SLM configuration is stored and behaves in xy/ising if next channel is in xy/ising.
  • Modification of find_slm_mask_times: it does not stop on a detuned_delay pulse.
• add_dmm_detuning: Add the waveform of the detuning to a DMM. Same as using add to add a Pulse zero amplitude and negative detuning to a DMM. If the DMM is configured as an SLM Mask, this is only possible if a Pulse has been defined on a Global channel before.

In #576:

• The dmm ids are switched when using 'switch_device' (useful when going from a virtual device to a real one). This is also performed for the dmm id used in config_slm_mask.
• Parametrized sequences are now taken into account. The SLM Mask and config detuning map are seen as operations. They are stored to be performed at build. This has an impact on the way declared channels work since we now go into the to_build_calls to find the declared dmms.
• Tests for compatibility between mappable register and detuning maps

In #568 and #576:

• Update serializer and deserializer for devices with DMM channels
• Write UTs for DMM channel serialization
• Update serializer and deserializer for the sequence using detuning map modulation
• Write UTs for sequence serialization and deserialization with detuning map modulation
• Update to the final sequence JSON schema
• Stores Config_det_map as an operation in the schema.

In #565, we develop a Special handling of samples from a DMM channel:

• We add a DMMSamples subclass of ChannelSamples, that stores the DetuningMap and the qubits of the register.
• Note: it is not possible to sample a Sequence with a Mappable Register and a configured DMM.

In #569:

• Fixes on the drawing of Register, RegisterLayout, DetuningMap:
  • The minimal height is extended and a label is added, "atom" for Register and "trap" for the others.
  • Display the DetuningMaps as traps (like a RegisterLayout, because it is not enforced to have an atom at the coordinates at which there is a weight). Label 'traps' is added to plot (label "atoms" added for general registers).
  • labels in DetuningMaps could only take integers, now it can handle QubitId.
• Addition to Sequence:
  • Add possibility to draw detuning maps (2D or 3D). Can be drawn with/without register.
  • Add possibility to draw local amplitude and detuning per group of qubits, that are represented on the register.
• Tutorials:
  • Creates a tutorial explaining how to create a DetuningMap, a DMM channel, how to modulate a detuning map in a Schedule and how to display amplitude and detuning per qubits.
  • Modifies the tutorial on state preparation with the SLM Mask to show difference between XY and Ising mode.
  • Updates some notebooks to take into account changes on Register

[HGSilveri]

I still have to review the changes in Sequence by already have some comments

[HGSilveri]

Things are a bit confusing with this latest commit

[HGSilveri]

• Note 1: If in ising, I am wondering if I should not be simply using find_slm_mask_times/ an equivalent strategy.

I think this is a good idea. Instead of manually searching for the first pulse, we can call find_slm_mask_times, see if the list of times comes out empty or not and take the first non-empty list we get. Just remember that we forgot so far to filter out detuned delays in find_slm_mask_times(), so this should be changed.

[a-corni]

• Note 1: If in ising, I am wondering if I should not be simply using find_slm_mask_times/ an equivalent strategy.

I think this is a good idea. Instead of manually searching for the first pulse, we can call find_slm_mask_times, see if the list of times comes out empty or not and take the first non-empty list we get. Just remember that we forgot so far to filter out detuned delays in find_slm_mask_times(), so this should be changed.

I think the latter should be addressed in another PR, I have tried to modify this and saw an unexpected error that did not seem so trivial (see Note 2).

[HGSilveri]

• Note 1: If in ising, I am wondering if I should not be simply using find_slm_mask_times/ an equivalent strategy.

I think this is a good idea. Instead of manually searching for the first pulse, we can call find_slm_mask_times, see if the list of times comes out empty or not and take the first non-empty list we get. Just remember that we forgot so far to filter out detuned delays in find_slm_mask_times(), so this should be changed.

I think the latter should be addressed in another PR, I have tried to modify this and saw an unexpected error that did not seem so trivial (see Note 2).

It's just a bad test because the first pulse is empty, so when you start ignoring it the qubits are no longer masked and you see the interaction terms appearing. If you put some amplitude on the rise pulse, this should go away

[a-corni]

• Note 1: If in ising, I am wondering if I should not be simply using find_slm_mask_times/ an equivalent strategy.

I think this is a good idea. Instead of manually searching for the first pulse, we can call find_slm_mask_times, see if the list of times comes out empty or not and take the first non-empty list we get. Just remember that we forgot so far to filter out detuned delays in find_slm_mask_times(), so this should be changed.

I think the latter should be addressed in another PR, I have tried to modify this and saw an unexpected error that did not seem so trivial (see Note 2).

It's just a bad test because the first pulse is empty, so when you start ignoring it the qubits are no longer masked and you see the interaction terms appearing. If you put some amplitude on the rise pulse, this should go away

So I have been adding an amplitude to the first pulse but the test is failing anyway, there are off-diagonal elements in the hamiltonian that depends on the amplitude of the pulse. I can't figure out where it is coming from (I have looked at the qobj_list in get_hamiltonian() but nothing easy came up)

[HGSilveri]

So I have been adding an amplitude to the first pulse but the test is failing anyway, there are off-diagonal elements in the hamiltonian that depends on the amplitude of the pulse. I can't figure out where it is coming from (I have looked at the qobj_list in get_hamiltonian() but nothing easy came up)

I can look into this, don't waste too much time on it

[HGSilveri]

So I have been adding an amplitude to the first pulse but the test is failing anyway, there are off-diagonal elements in the hamiltonian that depends on the amplitude of the pulse. I can't figure out where it is coming from (I have looked at the qobj_list in get_hamiltonian() but nothing easy came up)

The off-diagonal terms appear precisely because we changed the Rabi frequency of the first pulse. Makes sense, since now the only unmasked atom (atom3) does feel something at t=0.

[a-corni]

So I have been adding an amplitude to the first pulse but the test is failing anyway, there are off-diagonal elements in the hamiltonian that depends on the amplitude of the pulse. I can't figure out where it is coming from (I have looked at the qobj_list in get_hamiltonian() but nothing easy came up)

The off-diagonal terms appear precisely because we changed the Rabi frequency of the first pulse. Makes sense, since now the only unmasked atom (atom3) does feel something at t=0.

Ookay "atom3" is the only atom feeling something because "atom1" is masked, so "atom0", "atom2" and "atom3" can feel something but due to state preparation error "atom0" and "atom_2" are not present (hence don't feel anything), is that right ?

[HGSilveri]

Ookay "atom3" is the only atom feeling something because "atom1" is masked, so "atom0", "atom2" and "atom3" can feel something but due to state preparation error "atom0" and "atom_2" are not present (hence don't feel anything), is that right ?

Yup, that's it :)

[HGSilveri]

Here's a review of the changes before the merge of #568, clearly nothing stands out

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[HGSilveri]

🎉 🎉 🎉