Missing imports and undefined variables in the example

[ayyucedemirbas]

Hello, while running the CT_Foundation_Demo.ipynb notebook in this repository, I encountered a series of issues with the get_ct_embeddings function and its usage in subsequent cells. Below is a step-by-step description of the problems I faced and the adjustments I made to keep the code running.

I attempted to run the following code block:

from google.api_core.client_options import ClientOptions
from google.api_core.retry import Retry
from google.cloud import aiplatform
from google.api_core import exceptions
_API_ENDPOINT = 'us-west1-aiplatform.googleapis.com'


def create_lidc_series_url(study_instance_uid, series_instance_uid):
  return ('https://healthcare.googleapis.com/v1/projects/hai-cd3-foundations/'
          'locations/us-central1/datasets/ct3d/dicomStores/lidc-idri/dicomWeb/'
          f'studies/{study_instance_uid}/series/{series_instance_uid}')


_RETRIABLE_TYPES = (
    exceptions.TooManyRequests,  # HTTP 429
    exceptions.InternalServerError,  # HTTP 500
    exceptions.BadGateway,  # HTTP 502
    exceptions.ServiceUnavailable,  # HTTP 503
    exceptions.DeadlineExceeded,  # HTTP 504
)


def _is_retryable(exc):
  return isinstance(exc, _RETRIABLE_TYPES)


retry_policy = Retry(predicate=_is_retryable)


def get_ct_embeddings(mytoken: str, dicom_urls: List[str]) -> List[Any]:
  """Calls the API to collect the embeddings from a given volume.

  Args:
    mytoken: The token to access the DICOM volume and API.
    dicom_urls: The list of Series-level DICOM URL to the CT volumes.

  Returns:
    The list of embeddings or errors generated by the service. Differences in
    Vertex end-point configurations may change the return type. The caller is
    responsible for interpreting this value and extracting the requisite
    data.

  """
  api_client = aiplatform.gapic.PredictionServiceClient(
      client_options=ClientOptions(api_endpoint=_API_ENDPOINT)
  )
  endpoint = api_client.endpoint_path(
      project='hai-cd3-foundations', location='us-central1', endpoint=300
  )

  # Create a single instance to send - you can send up to 5.
  instances = []
  for dicom_url in dicom_urls:
    instances.append({"dicom_path": dicom_url, "bearer_token": mytoken})

  response = api_client.predict(
      endpoint=endpoint, instances=[instance], retry=retry_policy, timeout=600
  )
  assert len(response.predictions) == len(dicom_urls)
  assert len(response.predictions[0]) == 3
  # You can get the model version for this prediction
  # response.predictions[0]['model_version']

  # Check for no errors
  responses = []
  for a_prediction in response.predictions:
    if a_prediction['error_response'] is not None:
      responses.append(a_prediction['error_response'])
    else:
      embeddings = np.array(a_prediction['embedding_result']['embedding'],
                            dtype=np.float32)
      assert embeddings.shape == (1408,), 'Unexpected embeddings shape.'
      responses.append(embeddings)

  return embeddings

However, this resulted in the following error:

NameError: name 'List' is not defined

To resolve this, I added from typing import List. Upon running the code again, I then received: NameError: name 'Any' is not defined

I removed the return type hint -> List[Any] to temporarily bypass this, which resolved the error.

Then, while using get_ct_embeddings in the next cell, first, I tried the following lines:

TOKEN_ = !gcloud beta auth application-default print-access-token
TOKEN = TOKEN_[0]

my_url = create_lidc_series_url(study_uids[VOLUME_TO_SHOW],
                                corresponding_series_uids[VOLUME_TO_SHOW])
get_ct_embeddings(mytoken=TOKEN, dicom_url=my_url)

Here, I received this error message: TypeError: get_ct_embeddings() got an unexpected keyword argument 'dicom_url'

So I removed the dicom_url argument while calling the get_ct_embeddings method, and this step caused the following error message:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-21-9f9c4d2f19aa> in <cell line: 8>()
      6 my_url = create_lidc_series_url(study_uids[VOLUME_TO_SHOW],
      7                                 corresponding_series_uids[VOLUME_TO_SHOW])
----> 8 get_ct_embeddings(mytoken=TOKEN)

TypeError: get_ct_embeddings() missing 1 required positional argument: 'dicom_urls'

At this point, I realized that the argument name must be dicom_urls, not dicom_url. I then updated the call to:

get_ct_embeddings(mytoken=TOKEN, dicom_urls=my_url)

which resolved the problem. However, I then received this new error:

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
<ipython-input-25-9002dd1d675d> in <cell line: 1>()
----> 1 get_ct_embeddings(mytoken=TOKEN, dicom_urls=my_url)

<ipython-input-18-187d7402fd08> in get_ct_embeddings(mytoken, dicom_urls)
     55 
     56   response = api_client.predict(
---> 57       endpoint=endpoint, instances=[instance], retry=retry_policy, timeout=600
     58   )
     59   assert len(response.predictions) == len(dicom_urls)

NameError: name 'instance' is not defined

This error occurred because in the previous cell response = api_client.predict(endpoint=endpoint, instances=[instance], retry=retry_policy, timeout=600) was referenced, but instance was not defined in the code.

Apologies for the length of this issue. I wanted to provide as much detail as possible to make it clear. Thank you so much!