GitLab

PiperOrigin-RevId: 328206696

PiperOrigin-RevId: 328197125

Keras model serialization causes a lot of problems.

PiperOrigin-RevId: 328162551

PiperOrigin-RevId: 328085299

PiperOrigin-RevId: 327984543

PiperOrigin-RevId: 327983381

PiperOrigin-RevId: 327919609

PiperOrigin-RevId: 327892795

PiperOrigin-RevId: 327838493

PiperOrigin-RevId: 327830072

PiperOrigin-RevId: 327569853

PiperOrigin-RevId: 327518640

Polish Seq2SeqTransformer: (1) consolidate args; (2) add tests for distribution strategy and decoding path. (3) fix bugs

PiperOrigin-RevId: 327455733

PiperOrigin-RevId: 327363070

PiperOrigin-RevId: 327339794

PiperOrigin-RevId: 327325736

PiperOrigin-RevId: 327257703

PiperOrigin-RevId: 327179271

PiperOrigin-RevId: 327160430

PiperOrigin-RevId: 327149314

PiperOrigin-RevId: 327139660

PiperOrigin-RevId: 327094300

PiperOrigin-RevId: 326947209

PiperOrigin-RevId: 326935344

PiperOrigin-RevId: 326739356

Testing: Trained Bert2Bert model with MLIR bridge enabled and then NHNet based on Bert2Bert checkpoint, also with MLIR bridge enabled. Exported this to a saved model, ran inference, and confirmed model performance, accuracy, and prediction parity with the model trained without the MLIR bridge enabled.
PiperOrigin-RevId: 326734770

PiperOrigin-RevId: 326722720

i.e., the checkpoint will be saved using
tf.train.Checkpoint(FLAGS.checkpoint_model_name=model)

PiperOrigin-RevId: 326672697

PiperOrigin-RevId: 326541553

PiperOrigin-RevId: 326496940

PiperOrigin-RevId: 326286926

Fix QQP preprocessing for test set. Test set does not contain qid1 or qid2 columns, so the current try/except block causes the entire test set to be ignored.

PiperOrigin-RevId: 326259054

Project name seems that have changed to project ID. Just changed the project names to project Id.

Add ops for extracting labels from sentences (get_*_labels() functions). This CL includes the ops for extracting labels for:
- BERT's next sentence prediction task
- ALBERT's sentence order prediction task

PiperOrigin-RevId: 326087418