Aleksey Vlasenko

Thanks for reporting this issue. I did BQ reader benchmarks and tried to optimize it before first release for Wiki dataset with few columns and compared it to GCS. According...

I got a repro, will see what I can do to make it better.

Sorry, forgot to provide an update here. After profiling reader on a benchmark with 100+ columns (see https://github.com/vlasenkoalexey/bigquery_perftest/blob/master/bq_perftest_mult_columns.py) I realized that the bottleneck is batch step: ```python streams_ds = tf.data.Dataset.from_tensor_slices(streams)...

Did you have a chance to try approach suggested in https://github.com/tensorflow/io/issues/1066#issuecomment-757074730 ? And also please confirm that your data is stored in the location close to where you are reading...

BigQuery TensorFlow reader can only read from the table, but you have an option to select which columns you want to read. It is limited to the functionality that BigQuery...

This is limitation of TPU platform. Internally TPUs are interfacing with a dedicated VM that doesn't have TF.IO installed (mostly for security reasons). We are working to change that.

I'll follow up on this and provide an update.

Turns out that there is significant amount of work we need to do internally in order to support BigQuery reader for TPUs. At the same time TPU team is updating...

Google has recently announced Cloud TPU support where your code runs directly on a machine hosting TPUs: https://cloud.google.com/blog/products/compute/introducing-cloud-tpu-vms Therefore BQ reader should work using this setup, feel free to try...

You specify which TensorFlow version to use when you create a TPU VM using `gcloud alpha compute tpus tpu-vm create` version argument. Just find a corresponding version of TF.IO from...