Abstract representation support for new DMM features #568
Conversation
HGSilveri
commented
Aug 2, 2023
•
HGSilveri
commented
- Update serializer and deserializer for devices with DMM channels
- Write UTs for DMM channel serialization
- Update to the preliminary device JSON schema (should be updated before merging)
- Update serializer and deserializer for the sequence using detuning map modulation (not tested so there could be bugs)
- Write UTs for sequence serialization and deserialization with detuning map modulation
- Update to the final sequence JSON schema
| for dmm_name, ser_det_map in obj["dmm_channels"]: | ||
| splitted_dmm_name = dmm_name.split("_") # dmm_id_num | ||
| dmm_id = "_".join(splitted_dmm_name[0:2]) | ||
| call_num = ( | ||
| 0 if len(splitted_dmm_name) <= 2 else int(splitted_dmm_name[2]) |
There was a problem hiding this comment.
I think this is overly reliant on the conventions in the code. If we have to, we should rather include the information on both the dmm_name and the dmm_id
There was a problem hiding this comment.
Okay I also felt that we should include the information on both the dmm_name and the dmm_id. It's just that since we decided to label the dmm_name as dmm_id_{number of time dmm has been declared} I figured I had all the information dmm_name...
| if dmm_id not in to_config_det_map: | ||
| to_config_det_map[dmm_id] = {} | ||
| to_config_det_map[dmm_id][call_num] = ser_det_map | ||
|
|
||
| for dmm_id, ser_det_maps in to_config_det_map.items(): | ||
| slm_mask_dmm = None | ||
| call_nums = list(ser_det_maps.keys()) | ||
| # If an SLM Mask has been defined, an iteration will be taken | ||
| for i, call_num in enumerate(call_nums): | ||
| if call_nums != i: | ||
| slm_mask_dmm = i | ||
| # Have to wait for the SLM Mask to be configured to finish | ||
| # configuring the last Detuning Maps | ||
| for call_num, ser_det_map in ser_det_maps.items(): | ||
| if slm_mask_dmm is not None and call_num >= slm_mask_dmm: | ||
| break | ||
| det_map = _deserialize_det_map(ser_det_map) | ||
| seq.config_detuning_map(detuning_map=det_map, dmm_id=dmm_id) |
There was a problem hiding this comment.
I think I understand the need for all this complexity, but I'm hoping we can find a better solution
There was a problem hiding this comment.
Well now I am thinking that everything would be solved by taking "config_detuning_map" as an operation instead of putting in "detuning_map". What do you think of this ? Eventually if we are attached to having the information regarding the configured detuning map in the schema I can have both.
Other idea: I pass the slm_mask_dmm, this provides the number of times the dmm used for the SLM can be configured here, and the last configurations are performed after the SLM is configured (all I am doing here is very much to identify what is the dmm_id used for the SLM and when should the configuration of the SLM happen compared to the configuration of this dmm_id)
There was a problem hiding this comment.
I think you're right, we should probably deal with it like a regular operation
| self, detuning_weights: Mapping[QubitId, float] | ||
| self, | ||
| detuning_weights: Mapping[QubitId, float], | ||
| slug: str | None = None, |
There was a problem hiding this comment.
The slug is missing a description in the docstring (here and in other define_detuning_map methods.
| _Call( | ||
| "config_detuning_map", (dmm_name, dmm_id, detuning_map), {} |
There was a problem hiding this comment.
This is quite problematic because _Call is not used only for serialization to the abstract representation. dmm_name is not an argument to config_detuning_map, so I think this will cause problems (for example, when building a sequence from scratch from the stored calls).
Also, the fact that nothing goes wrong suggests that we should be more comprehensive in the unit tests.
There was a problem hiding this comment.
It's true that I don't think we have ever tested building a parametrized sequence having a config_detuning_map... And I guess this will fail indeed. I will start by solving this issue.
* Adding dmm config and modulation to sequence * Reverting changes for _SLMMask * Setting max_amp of dmm to zero * Fixing detection of virtual channels in Device * Reverting changes in serialization * Reverting SLM, modifying available_channels * deleting dmm of schedule if slm of xy, Fixing type * Add insert_slm_mask * fixing available_channels with slm * Special handling of samples from a DMM channel (#565) * Add `DetuningMap.get_qubit_weight_map()` * Define `_DMMSchedule.get_samples()` * Extract DMMSamples from a Sequence * Creating the `Traps` class * Eliminating the effects of the SLM mask * Use COORD_PRECISION in comparison * `Traps.coords` -> `Traps.sorted_coords` * Subclass `Traps` in `WeightMap` * Preserve effects of SLM mask in XY mode * Explicitly ask for labels in DetuningMap.draw() * Including slug in WeightMap * UTs for DetuningMap * Isort * Specialize pulse validation in DMM * Revamp test_dmm UTs * Formatting * Add support for legacy serializer * add in_ising * Add SLM pulse, ising, modif validate_channel * Add waiting_for_first_pulse, fixing mask time * Covering all code, modif after review * Add descr sequence, _config_detuning_map * Finishing UTs, fixing typos * Abstract representation support for new DMM features (#568) * Support for DMM channel serialization * WIP: Prepare for incoming updates to JSON schema * Add descr sequence, _config_detuning_map * Finishing UTs, fixing typos * de-/serializing det maps, adding schema, slug * Dev tests * Fixing doc --------- Co-authored-by: a_corni <antoine.cornillot@pasqal.com> * Testing detuning of unmasked qubits, del print * Enabling use of add with slm * Replacing Modulate_det_map by add_dmm_detuning * Updating schema * finalizing replacing of modulate_detp_map * finalizing replacing of modulate_det_map * Rephrasing doc, add OpAddDmmDet * Handle parametrized sequence, switching device and DMM, and serialize config_det_map as an operation (#576) * Add slug description * Handle DMM & parametrized/switch_device/MapReg * Modifying schema, fixing type * Moving to a store decoration * Storing config_det_map * Taking into account review comments * Delete seq_tools * Testing MappableRegister, strict conversion * Replace modulate_det_map by add_dmm_detuning * Adding DMM notebook and Modifying SLM Mask (#569) * adding dmm, draw to DetuningMap * Testing avoiding circular import * Refactoring to avoid circular imports * Fix broken UTs * Import sorting * Fixing plot DetuningMap * Serialization/Deserialization of DMM in device * Fixing typos * Testing DMM and DetuningMap * Fixing typo * adding xfails, fixing type * Remove DMM from devices and finish UTs * Take into account review comments * Add annotations * Error in Global and Local * Defining _DMMSchedule * Fixing nits * Fixing typo * Creating DMM notebook, modif SLM notebook * Taking into account review comments * Fixing type * Fix labels in reg_drawer, draw det_maps in seq * Fixing doc build, add draw_detuning_maps to docs * Separating register/det_maps drawing from channels * introducing get_qubit_data * Add drawing of quantities per qubit * Fixing for local pulses * Adding bounds to label * Drawing legend for local targetting * FIX sampling rate, QutipEmulator * Updating figures * Updating SLM Mask & local addressability notebook * Adding explanation text * Fixing type * Replace modulate_det_map by add_dmm_detuning * Replace modulate_det_map for add_dmm_detuning * Fixing tests * Revert changes to Bayesian optimization notebook * Fixing docs * clearing outputs * Deleting scaling in favor extending to draw reg * Placing legend in lower part of drawing * Updating figures * Relaunching bayesian optimisation * Taking into account review comments * Printing sequence with DMM * Adding subscript and indent * Defining __str__ for DMM --------- Co-authored-by: HGSilveri <henrique.silverio@tecnico.ulisboa.pt> --------- Co-authored-by: Henrique Silvério <henrique.silverio@tecnico.ulisboa.pt>
