Skip to main content

With S3

info

For detailed instructions on setting up your S3 bucket, refer to the Bucket Management documentation.

We can now start issuing problems to the TitanQ Solver. Below are some examples with expected results that can be run from your favorite IDE and Python environments locally.

Note that the following Python modules need to be installed:

  • boto3 (for AWS access)
  • numpy (for data generation exchange)

Generate a weights matrix and a bias vector​

Here are some example input weights matrix and bias vectors:

import numpy as np
weights = np.array([[ 0, -1, -1, -1,  1, -1,  1, -1, -1,  1],
                    [-1,  0,  1, -1, -1,  1, -1, -1, -1,  1],
                    [-1,  1,  0, -1,  1, -1,  1, -1, -1, -1],
                    [-1, -1, -1,  0, -1,  1,  1, -1,  1, -1],
                    [ 1, -1,  1, -1,  0, -1, -1,  1, -1,  1],
                    [-1,  1, -1,  1, -1,  0, -1, -1,  1,  1],
                    [ 1, -1,  1,  1, -1, -1,  0,  1,  1,  1],
                    [-1, -1, -1, -1,  1, -1,  1,  0,  1,  1],
                    [-1, -1, -1,  1, -1,  1,  1,  1,  0,  1],
                    [ 1,  1, -1, -1,  1,  1,  1,  1,  1,  0]], dtype=np.float32)
bias = np.zeros(len(weights), dtype=np.float32)
np.save('weights.npy', weights)
np.save('bias.npy', bias)

Upload the two files to the source bucket.​

From the S3 dashboard on the web console, navigate to your chosen bucket, then click Upload:

s3_upload_files.png

or programmatically, using access keys belonging to a user having write access to the source bucket (see section Expanding on the Example)

import boto3

s3 = boto3.client('s3',
    aws_access_key_id=SUPERVISOR_ACCESS_KEY_ID,
    aws_secret_access_key=SUPERVISOR_SECRET_ACCESS_KEY)
s3_client.upload_file(
    'YOUR_LOCAL_WEIGHTS_FILE.npy',
    'SOURCE_BUCKET_NAME',
    'weights.npy')
s3_client.upload_file(
    'YOUR_LOCAL_BIAS_FILE.npy',
    'SOURCE_BUCKET_NAME',
    'bias.npy')

Issuing a solve request​

Submitting a request to the TitanQ API involves preparing a JSON body and some HTTP Header parameters.

note

In the example below, make sure to fill in...

  • your API key (API_KEY)
  • your read user credentials (READ_USER_*)
  • your source bucket name (SOURCE_BUCKET_NAME)
  • your write user credentials (WRITE_USER_*)
  • your destination bucket name (DEST_BUCKET_NAME)
  • your supervisor user credentials (SUPERVISOR_*)
note

These problems were originally meant as bipolar problems. Here, we solve them as binary. To use them in the original bipolar-native setting, please use the bipolar-to-binary converter provided in the TitanQ SDK.

import io
import json
import numpy as np
import requests
import time
import zipfile
import boto3


def main():
  hdr = {
    'content-type': 'application/json',
    'authorization': 'API_KEY'
  }

  titanq_req_body = {
    'input': {
      'bias_file_name': 'bias.npy',
      'weights_file_name': 'weights.npy',
      's3': {
        'bucket_name': 'SOURCE_BUCKET_NAME',
        'access_key_id': 'READ_USER_ACCESS_KEY_ID',
        'secret_access_key': 'READ_USER_SECRET_ACCESS_KEY'
      }
    },
    'output': {
      's3': {
        'bucket_name': 'DEST_BUCKET_NAME',
        'access_key_id': 'WRITE_USER_ACCESS_KEY',
        'secret_access_key': 'WRITE_USER_SECRET_ACCESS_KEY'
      },
      'result_archive_file_name': 'titanq/quickstart_result.zip'
    },
    'parameters': {
      'beta': (1/(np.linspace(0.5, 100, 8, dtype=np.float32))).tolist(),
      'coupling_mult': 0.5,
      'num_chains': 8,
      'num_engines': 4,
      'timeout_in_secs': 0.1,
      'variable_types': 'b'*10,
    }
  }

  try:
    resp = requests.post(
      'https://titanq.infinityq.io/v1/solve',
      headers=hdr,
      data=json.dumps(titanq_req_body))

    if 200 <= resp.status_code < 300:
      # Parse the request's confirmation
      titanq_response_body = json.loads(resp.content)
      message = titanq_response_body['message']
      status = titanq_response_body['status']
      computation_id = titanq_response_body['computation_id']

      # Wait for computation to complete.
      # Guesstimate, assuming no delay in queue, etc.
      time.sleep(titanq_req_body['parameters']['timeout_in_secs'] + 3)

      # using an AWS user with read access to the destination bucket
      s3_client = boto3.client('s3',
        aws_access_key_id='SUPERVISOR_ACCESS_KEY_ID',
        aws_secret_access_key='SUPERVISOR_SECRET_ACCESS_KEY_ID')

      # fetch the result archive file
      remote_object = s3_client.get_object(
        Bucket=titanq_req_body['output']['s3']['bucket_name'],
        Key=titanq_req_body['output']['result_archive_file_name']
      )

      # in memory buffer
      buff = io.BytesIO(remote_object['Body'].read())
      
      # unzip to the current directory
      with zipfile.ZipFile(buff, 'r') as zip_file:
        zip_file.extractall(".")

      # Inspect results
      with open('metrics.json', 'r') as metrics:
        metrics = json.load(metrics)

        print("-" * 15, "+", "-" * 22, sep="")
        print("Ising energy   | Expected ising energy")
        print("-" * 15, "+", "-" * 22, sep="")
        for ising_energy in metrics['ising_energy']:
            print(f"{ising_energy: <14f} | -35")

    else:
      print('Request yielded HTTP {}'.format(resp.status))
  except Exception as e:
    print('Exception ', e)


if __name__ == '__main__':
  main()