XGBoost-1.7 container not working with feature attribution "explain"

[daniel-furman]

Environment details

https://cloud.google.com/vertex-ai/docs/predictions/pre-built-containers#xgboost

XGBoost 1.7 not working with .explain. Running the same deployment code with 1.6, however, does work. Since the "End of availability" date for <=1.6 is fast approaching (Nov 2024), I am wondering if this will be patched and/or if the availability date for 1.6 could be extended.

Steps to reproduce

1. Train an XGBoost classifier
2. Deploy it with feature explanations turned on to 1.7/1.6
3. 1.6 will work, 1.7 will err with this message:

InvalidArgument: 400 {"error": "b'{\"detail\":\"Unsupported content type of the request: application/x-www-form-urlencoded.\\\\nCurrently supported content-type in DefaultSerializer: \\\\\"application/json\\\\\".\"}'"}

Code example to reproduce error

from google.cloud import aiplatform
from google.cloud.aiplatform_v1.types import SampledShapleyAttribution
from google.cloud.aiplatform_v1.types.explanation import ExplanationParameters

exp_metadata = aiplatform.explain.ExplanationMetadata(
    inputs={"features": {}},
    outputs={"score": {}},
)

exp_parameters = ExplanationParameters(
    sampled_shapley_attribution=SampledShapleyAttribution(path_count=5)
)

# Register model
model = aiplatform.Model.upload(
    display_name=model_display_name, 
    artifact_uri=f'gs://{bucket_name}/model',
    serving_container_image_uri='us-docker.pkg.dev/vertex-ai/prediction/xgboost-cpu.1-7:latest',  # swap 1-7 for 1-6 to test difference
    explanation_metadata=exp_metadata,
    explanation_parameters=exp_parameters,
)

# Deploy model to endpoint
endpoint = aiplatform.Endpoint.create(
    display_name=display_name
)
model.deploy(
    endpoint=endpoint,
    deployed_model_display_name=deployed_model_display_name,
    machine_type='n1-standard-2',
    min_replica_count=1,
    max_replica_count=2,
)