New sampler implementation #388
Conversation
itertools.pairwise is >=3.10.
Necessary to avoid circular imports
noise models were made void with the change from QubitSamples to ChannelSamples
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
Now we need to pass the duration of the parent sequence as a parameter to embed the samples in an array of the right length
|
Last commits restore the tests and update them to the new architecture. |
|
Thank you @lvignoli ! I'll take over from here and circle back to you if I have any questions |
|
@lvignoli the changes are done. I'm noticing there's a tutorial that I'm guessing should be untracked? |
|
Thank you @HGSilveri! The tutorial was simply to check that the sampler was working as intended, so you can delete it safely. Besides I think @arthurfaria would like to write a notebook tutorial once this is reviewed and merge, see #349. |
Sounds good. Perhaps I will start with that one. Is there any other channel we can communicate like slack or something? |
Indeed, we have a Slack channel that has been dormant but that we can use. I sent you an invite. |
The following sampler implementation is channel-centric—by opposing to the former one which was qubit centric.
Samples dataclasses are defined in
pulser-core/samples.py. Samples are extracted at the channel level from the_ChannelSchedule.get_samples()method. For the regular user, there is only one function to know, thesample(seq)one exposed by thepulser-core/sampler/module.Improvements:
ChannelDrawContentfromChannelSamples.There is much more info needed to draw, its a wider scope than the time series, so I am unsure about the right design.
Sequenceclass to get samples?