ValueError: An error occurred while calling the read_parquet method registered to the cudf backend.
Original Message: An error occurred while calling the read_parquet method registered to the pandas backend.
Original Message: Invalid gcloud credentials
that GCS bucket contains some files at path gcs://anaconda-public-data/nyc-taxi/2015.parquet that are not actually parquet files
cuspatial.haversine_distance() expects to receive 2 cuspatial.GeoSeries objects
TypeError('haversine_distance() takes 2 positional arguments but 4 were given')
Then ran this notebook code (just the LocalCUDACluster parts and below), on a machine with a few 80GB H100s. Confirmed that data was pulled successfully without needing to authenticate with GCP, and that cuspatial.haversine_distance() ran without error and produced plausible-looking results.
Fixes the
cuspatial.haversine_distance()
example using the public NYC taxi data hosted on GSC (GSC console link).Running through the notebook code for that example (the same one that ends up at https://docs.rapids.ai/deployment/nightly/examples/rapids-autoscaling-multi-tenant-kubernetes/notebook/), I encountered 3 issues:
gcs://anaconda-public-data/nyc-taxi/2015.parquet
that are not actually parquet filescuspatial.haversine_distance()
expects to receive 2cuspatial.GeoSeries
objectsLooks like that changed here: https://github.com/rapidsai/cuspatial/pull/924.
This resolves those issues.
How I tested this
Following the instructions from https://docs.rapids.ai/install#install-rapids, ran jupyter lab in a RAPIDS container like this:
Then ran this notebook code (just the
LocalCUDACluster
parts and below), on a machine with a few 80GB H100s. Confirmed that data was pulled successfully without needing to authenticate with GCP, and thatcuspatial.haversine_distance()
ran without error and produced plausible-looking results.