recommenders

recommenders copied to clipboard

[Idea] Learn from Positives and Negatives Using Two Retrieval Models

3

In a recommendation system with retrieval and ranking stages, the retrieval model typically learns only from positive examples in the training dataset. Consequently, it fails to learn which items users...

rlcauvin

Training took too long time with pre-trained embedding BERT layer

1

Hello everyone, We are trying to integrate pre-trained BERT embedding into our TFRS model. Our model is based on the same definition as https://www.tensorflow.org/recommenders/examples/basic_retrieval. ```python class RetrievalUserModel(tf.keras.Model): def __init__(self: Self,...

beubeu13220

Not possible to run the basic retrieval example

8

Version mismatches in tensorflow , tensorflow_datasets, tensorflow_recommenders, scann and python. Impossible to find a combination that works. You get error messages about something wrong with builder.py in protobof. You upgrade...

pksinghus

It is an honor to have the opportunity to read your paper. In your paper, you mentioned that the gradient clipping application you proposed has a good effect on Adam....

sugarcandy

The movie retrieval tutorial/notebook is broken due to a version conflict of the protobuf dependencies. https://github.com/tensorflow/recommenders/blob/main/docs/examples/basic_retrieval.ipynb Running the pip installs in the notebook (as well as in offline environments) yields...

marco-ve

What is the benefit of having two separate class one for user and one for query?

1

Hello I see in this [tutorial](https://github.com/tensorflow/recommenders/blob/main/docs/examples/deep_recommenders.ipynb), there are separate classes for user model and query model. The same for the movie model and candidate model. Can you please explain what...

abdollahpouri

[Question] Support for other loss functions.

3

This question is very broad and theoretical so apologies for that! I would love to learn more about various loss functions which could be implemented in `tfrs` - when looking...

ydennisy

[Question] expected throughput of cloud tpu on embedding lookup?

3

Hi, I read this blog recently https://cloud.google.com/blog/topics/developers-practitioners/building-large-scale-recommenders-using-cloud-tpus, very interested in it and wondering the raw performance of TPUEmbedding lookup performance.(we can quite easily get the perf data of tf.nn.(safe)embedding_lookup(_sparse) etc....

pmixer

How to generate BruteForce predictions with potential side features in candidate stack

3

Hi, I'm following the tutorial "Building deep retrieval models" and seem to have encountered a slightly different, though related issue than either [this post](https://github.com/tensorflow/recommenders/issues/226) or this [other one](https://github.com/tensorflow/recommenders/issues/162) on this...

kingjosephm

How to make predictions by retrieval models with multiple features?

8

` class UserModel(tf.keras.Model): def __init__(self,use_timestamps,use_distance): super().__init__() self._use_timestamps = use_timestamps self._use_distance = use_distance self.user_embedding = tf.keras.Sequential([ tf.keras.layers.StringLookup( vocabulary=unique_user_ids, mask_token=None), tf.keras.layers.Embedding(len(unique_user_ids) + 1, 32) ]) max_tokens = 10_000 self.Preference1_embedding = tf.keras.Sequential([ tf.keras.layers.StringLookup(...

zeroonesfas