AwsQuantumJob for QuEra has incomplete software stack about amazon-braket-examples HOT 17 CLOSED

from amazon-braket-examples.

Looking into reproing this issue

from amazon-braket-examples.

Just running the hybrid_submit.py has been stucked at "Initializing Braket Job" for me so far. Still Investigating.

from amazon-braket-examples.

indeed, now it never passe the . . . ., this is what I see.
But at the time of filing this ticket the image would start after ~15 dots, then I'd get the crash.
FYI, I run my test when Aquila was up (just 10 hours per week) - perhaps image does not want to start if
deviceArn='arn:aws:braket:us-east-1::device/qpu/quera/Aquila' and Aquila is down?

Initializing Braket Job: arn:aws:braket:us-east-1:765483381942:job/braket-job-default-1689116339095
..........................................................................................................................................................................................................................................................................

from amazon-braket-examples.

1. Which quera_ahs_utils module are you referring to? If you're talking about those that were introduced into the Braket SDK in amazon-braket/amazon-braket-sdk-python#508, the functions were moved to appropriate classes and renamed; for example, get_drive became DrivingField.from_lists.
2. Would you be able to start a job locally, and in your script, print out the Braket SDK version (!pip show amazon-braket-sdk)? The AHS utils were introduced in 1.37.0, so it would help to see what version you're getting.

from amazon-braket-examples.

from amazon-braket-examples.

2.1) here are more details about it
$ pip3 list|grep amazon
amazon-braket-default-simulator 1.18.2
amazon-braket-schemas 1.18.0
amazon-braket-sdk 1.49.0
$pip3 list|grep ahs
quera-ahs-utils 0.4.3

I think this is your environment rather than the job environment; the library versions may be different between the two. As for the jobs themselves, the long initialization may be due to the job being stuck in a job queue; can you verify whether your job arn:aws:braket:us-east-1:765483381942:job/braket-job-default-1689116339095 has completed in the Braket console?

from amazon-braket-examples.

from amazon-braket-examples.

yes, this is my local environment. This is all what I can do now. As I mentioned ,I can't start a hybrid job for 5-10 minutes any more. I do not have patience to wait longer. This 1 time when the hybrid job started for me I could observe the stages: The initial .... stopped and then some image on AWS started, there was a message about it. Next the image accepted my QuEra python code and then I could see it submitted the Aquila job to QuEra, which I could observe via AWS web service as QUEUED. I could also kill this job via AWS web and then on my terminal I'd see the results form killed job returned by the image running on AWS. But I can't do it any more. The AWS image never reports as working. It is all in your court now. I can do nothing until you manage to start the AWS image. Can you start it?

In your code snippet, you have the line

wait_until_complete=True

which causes the ......... to appear and block your Python process. If you run

job = AwsQuantumJob.create(
    deviceArn,
    source_module="hybrid_task.py",
    entry_point="hybrid_task:start_here",
)

the job will still be created, and you can access its status with

job.state()

The wait is unfortunately out of our control; it depends on the number of customers who also have jobs queued up.

Do you see  as described https://docs.aws.amazon.com/braket/latest/developerguide/braket-jobs-first.html This message should show up even if my QuEra task is wrong - it reports the image has started Jan

I don't see the image you posted, but if it's a message output (like the Test job started!!!!! in your script) will not appear past the ...... while the job is in the QUEUED state.

from amazon-braket-examples.

To diagnose versioning issues, would you be able to run this script

import os
import numpy as np
from pprint import pprint
from decimal import Decimal

from braket import _schemas, _sdk, default_simulator
from braket.ahs import AtomArrangement, DrivingField, Hamiltonian, AnalogHamiltonianSimulation
from braket.devices import LocalSimulator
from braket.aws import AwsDevice
from braket.tracking import Tracker

#...!...!....................
def ahs_problem():
    a = 6.1e-6  # meters
    N_atoms = 11

    register = AtomArrangement()
    for i in range(N_atoms):   register.add([0.0, i*a])

    print('\nM:dump atom coordinates')
    
    xL=register.coordinate_list(0)
    yL=register.coordinate_list(1)
    
    for x,y in zip(xL,yL):  print(x,y)

    H = Hamiltonian()
    
    omega_max = 2e6 * 2 * np.pi
    detuning_min = -9e6 * 2 * np.pi
    detuning_max = 7e6 * 2 * np.pi
    
    time_max = Decimal('4')/1000000 # 4 us
    time_ramp =time_max/4

    # rounding is required for Aquila
    omega_max=int(omega_max/400)*400
    detuning_min=int(detuning_min*5)/5
    detuning_max=int(detuning_max*5)/5
    
    time_points = [0, time_ramp, time_max - time_ramp, time_max]
    omega_values = [0, omega_max, omega_max, 0]
    detuning_values = [detuning_min, detuning_min, detuning_max, detuning_max]
    phase_values = [0, 0, 0, 0]

    drive = DrivingField.from_lists(time_points, omega_values, detuning_values, phase_values)
    H += drive

    print('\nM:dump drive Amplitude :\n',drive.amplitude.time_series.times(), drive.amplitude.time_series.values())

    program = AnalogHamiltonianSimulation(  register=register, hamiltonian=H)

    return program

#...!...!....................
def full_task(device,ahs_prog,shots):
    task = device.run(ahs_prog, shots=shots)
    results=task.result()  # this forces wait for task completion/terminationcost tracking
    print('status=%s:'%task.state())
    
    if task.state()=='COMPLETED': 
        rawBitstr=results.get_counts()
        print('COMPLETED, rawBitstr:');  pprint(rawBitstr)
       
#...!...!....................
def start_here():    
    print("Test job started!!!!!")
    print("Versions:")  # See the versions of the Braket libraries in your image
    print(f"schemas: {_schemas.__version__}")
    print(f"SDK: {_sdk.__version__}")
    print(f"default simulator: {default_simulator.__version__}")
    # Use the device declared in the job script
    device = LocalSimulator("braket_ahs")
    print('M:device is braket_ahs')
    ahs_prog=ahs_problem()
    shots=100

    print("M:Test job completed!!!!!")

if __name__ == "__main__":
    start_here()

and run said script locally (assuming the same filename as your script):

from braket.jobs.local import LocalQuantumJob

job = LocalQuantumJob.create(
    device="local:amazon/braket_ahs,
    source_module="hybrid_task.py",
    entry_point="hybrid_task:start_here",
)

This will try to run your script on the local simulator, as well as print out the versions of the libraries in the container image.

from amazon-braket-examples.

from amazon-braket-examples.

from amazon-braket-examples.

Hi,
is there any resolution for this problem?
Thanks
Jan

from amazon-braket-examples.

Hi Jan,

We're very sorry for the delay in getting back to you. You should be able to run your script now, with modifications. First, there is no need to use quera_ahs_utils.drive.get_drive anymore; in fact, you can replace the import block

from braket.ahs.atom_arrangement import AtomArrangement
from braket.ahs.hamiltonian import Hamiltonian
from braket.ahs.analog_hamiltonian_simulation import AnalogHamiltonianSimulation
from quera_ahs_utils.drive import get_drive

with just

from braket.ahs import AnalogHamiltonianSimulation, AtomArrangement, DrivingField, Hamiltonian

and then replace

drive = get_drive(time_points, omega_values, detuning_values, phase_values)

with

drive = DrivingField.from_lists(time_points, omega_values, detuning_values, phase_values)

With these modifications, I can now run the job, and it gets to the

M:executing

print statement, which means that the DrivingField.from_lists is successful. Since from lists was introduced in the same change that added get_counts to AnalogHamiltonianSimulationQuantumTaskResult, this gives me some confidence that the rest of the job should run too.

I believe the issue was fixed with amazon-braket/amazon-braket-sdk-python#643, so pulling in the latest Amazon Braket SDK should do the trick.

As for the example I gave you, Docker is a requirement for running local jobs, so that might be the reason it didn't work.

from amazon-braket-examples.

from amazon-braket-examples.

There isn't a specific example of running a hybrid job with the QuEra Aquila (although a contribution is always welcome); as such, one path forward could be to adapt one of the QuEra example notebooks into a hybrid job. However, the easiest demonstration of success would be to run the script you provided to completion (with the modifications I listed); I'll run the job, and if it completes, then we're golden!

from amazon-braket-examples.

Thank you for catching this issue!

from amazon-braket-examples.