Summary: 31 instances, 21 unique

Text	Count
# TODO: Consider to add a cap position topn + 1 when there is no	1
# TODO Add recall metrics to TF1 in another cl.	1
# TODO: These parameters may be inferred from the call inputs.	1
# TODO: Extend to more generic encoders in addition to BERT.	1
# TODO: integrate PSStrategy to pipeline.	1
# TODO: Add mask argument for metric.compute() call	1
# TODO: Add checks to validate (ragged) shapes of input tensors.	1
# TODO: supports for distributed training and evaluation.	1
TODO: Add the lambda_weight support.	1
# TODO: Add the per-head loss loss/head_name to metrics.	1
# TODO: Ensure only v2 Feature Columns are used.	1
# TODO: try lambda_weights for circle loss.	1
# TODO: Add mask argument for metric.compute() call.	1
# TODO: Be more smart to infer is_valid.	1
# TODO: Remove defaults common in Estimator pipeline and here.	1
# TODO: Support compatibility with TPUs.	1
# TODO: Define these bulky types as globals at the top.	1
# TODO: Add support for circle loss and neural sort losses.	1
# TODO: Figure out a better way to set train_op_fn and optimizer	1
# TODO: Add mask argument for metric.compute() call	11
# TODO: Attach the link to the paper.	1