TensorFlow Recommenders is a library for building recommender system models using TensorFlow.
APACHE-2.0 License
Retrieval
task now accepts a list of factorized metrics instead of atfrs.experimental.optimizers.ClippyAdagrad
: a new optimizer based ontf.keras.optimizers.Adagrad
that is able to improve training stability.tfrs.metrics.FactorizedTopK
now accepts sample weights which are used toPublished by maciejkula about 2 years ago
Published by maciejkula over 2 years ago
Release fixes.
Published by maciejkula over 2 years ago
A number of changes to make factorized top-K metric computation more accurate
and less prone to user error.
tfrs.layers.embedding.TPUEmbedding
now supports input features with
dynamic shape. batch_size
argument is deprecated and no longer required.
tfrs.layers.embedding.TPUEmbedding
now supports running on different
versions of TPU.
Pinned TensorFlow to >= 2.9.0 which works with Scann 1.2.7.
tfrs.tasks.Ranking.call
now accepts a compute_batch_metrics
argument to
allow switching off batch metric computation. Following this change,
'compute_metrics'argument does not impact computation of batch metrics.
tfrs.metrics.FactorizedTopK
requires the candidate ids for positiveexact
method to broadcast its ability to return exact ormetrics
constructor parameter for tfrs.metrics.FactorizedTopK
.FactorizedTopK
only makes sense with top-k metrics, and this changek
constructor argument in tfrs.metrics.FactorizedTopK
withks
: a list of k
values at which to compute the top k metric.tfrs.metrics.FactorizedTopK
metric can now compute candidate-id basedtrue_candidate_ids
argument in its call
method.Retrieval
task now also accepts a loss_metrics
argument.Published by maciejkula about 3 years ago
TopK
layer indexing API changed. Indexing with datasets is now done viaindex_from_dataset
method. This change reduces the possibility ofPublished by maciejkula over 3 years ago
tfrs.experimental.models.Ranking
batch_metrics
to tfrs.tasks.Retrieval
for measuring how good thetfrs.experimental.layers.embedding.PartialTPUEmbedding
layer, whichtfrs.layers.embedding.TPUEmbedding
for large embedding lookups andtf.keras.layers.Embedding
for smaller embedding lookups.Published by maciejkula over 3 years ago
factorized_top_k
layers will now raise (to prevent issues similar toPublished by maciejkula over 3 years ago
tfrs.experimental.models.Ranking
, an experimental pre-built model fortfrs.layers.loss.SamplingProbablityCorrection
that logitstfrs.experimental.optimizers.CompositeOptimizer
: an optimizer thattfrs.layers.dcn.Cross
and DotInteraction
layers have been moved totfrs.layers.feature_interaction
package.Published by maciejkula almost 4 years ago
TopK
layers now come with a query_with_exclusions
method, allowingTPUEmbedding
Keras layer for accelerating embedding lookups for largefactorized_top_k.Streaming
layer now accepts a query model, like other
factorized_top_k
layers.
Updated ScaNN to 1.2.0, which requires TensorFlow 2.4.x. When not using
ScaNN, any TF >= 2.3 is still supported.
Published by maciejkula almost 4 years ago
Published by maciejkula almost 4 years ago
Published by maciejkula almost 4 years ago
Published by maciejkula about 4 years ago
tfrs.layers.corpus.DatasetTopk
has been removed,tfrs.layers.corpus.DatasetIndexedTopK
renamed totfrs.layers.factorized_top_k.Streaming
, tfrs.layers.ann.BruteForce
tfrs.layers.factorized_top_k.BruteForce
. All top-k retrievalBruteForce
, Streaming
) now follow a common interface.tfrs.tasks.Ranking.call
now accepts a compute_metrics
argument to allowtfrs.tasks.Ranking
now accepts label and prediction metrics.tfrs.tasks.Retrieval
.Dataset
parallelism enabled by default in DatasetTopK
andDatasetIndexedTopK
layers, bringing over 2x speed-ups to evaluationsevaluate_metrics
argument to tfrs.tasks.Retrieval.call
renamed tocompute_metrics
.Published by maciejkula about 4 years ago
Published by maciejkula about 4 years ago
Published by maciejkula about 4 years ago