AutoML预测上传至Google Cloud Functions后报错的问题咨询

I’ve run into this exact issue before—version mismatches between your local google-cloud-automl library and the one installed by default in Cloud Functions are almost always the culprit here. Let’s break down what’s happening and how to fix it:

Why This Error Happens #

Your local code works because you’re using a version of the AutoML client where predict() accepts multiple positional arguments (name, payload, params). But when deployed to Cloud Functions, the default installed library version uses a different method signature—likely expecting only a single PredictRequest object instead of separate arguments. This mismatch causes the "takes 1 to 2 positional arguments but X were given" error.

Step 1: Align Library Versions #

First, make sure Cloud Functions uses the exact same version of google-cloud-automl that you tested locally.

1. Run pip show google-cloud-automl on your local machine to get your version (e.g., 2.10.0).
2. Create or update a requirements.txt file in your Cloud Functions deployment folder with this explicit version:

   google-cloud-automl==2.10.0
   

This ensures Cloud Functions installs the same library you used locally, eliminating signature mismatches.

Step 2: Use the Correct Predict Call for Your Version #

Depending on your library version, adjust the predict() call to match the expected signature:

For Newer Versions (v1.0+) #

The predict() method expects a single PredictRequest object. Update your code like this:

from google.cloud import automl_v1beta1
from google.cloud.automl_v1beta1.types import PredictRequest

def get_sentiment(content):
    """ Returns a google cloud platform payload class containing the sentiment score given by our NLP sentiment analyser.
    :param content: STRING (UTF-8 encoded, ASCII)
    :return: <class 'google.cloud.automl.types.PredictResponse'>
    """
    options = automl_v1beta1.ClientOptions(api_endpoint='automl.googleapis.com')
    prediction_client = automl_v1beta1.PredictionServiceClient(client_options=options)
    name = model_sentiment  # Ensure this is set to your full model resource name (projects/.../models/...)
    
    # Build the payload
    payload = {'text_snippet': {'content': content, 'mime_type': 'text/plain'}}
    
    # Create the request object
    request = PredictRequest(
        name=name,
        payload=payload,
        params={}
    )
    
    # Call predict with the request object
    response = prediction_client.predict(request)
    return response

For Older Versions (Pre-v1.0) #

If you’re stuck on an older library version, use keyword arguments to avoid positional mismatches:

# Replace your existing predict call with this
request = prediction_client.predict(
    name=name,
    payload=payload,
    params=params
)

Note: Using keyword arguments makes the call more robust to signature changes, but aligning versions is still the better long-term fix.

Step 3: Verify Model Resource Name #

Double-check that model_sentiment is set to the full, valid resource name of your AutoML model, formatted like:
projects/YOUR_PROJECT_ID/locations/YOUR_LOCATION/models/YOUR_MODEL_ID
If this variable is hardcoded locally but not properly configured in Cloud Functions (e.g., missing environment variables), that could cause additional issues.

Final Checks #

• Test your code locally with the exact version specified in requirements.txt to ensure it works before deploying.
• Confirm Cloud Functions has the necessary IAM permissions to access your AutoML model (the service account used by Cloud Functions should have the AutoML Predictor role).

内容的提问来源于stack exchange，提问作者Alfred Lindholm