Skip to content

issues Search Results · repo:keras-team/keras-rs language:Python

Filter by

12 results
 (82 ms)

12 results

inkeras-team/keras-rs (press backspace or delete to remove)

keras-team/keras-rs
Potentially limit `max_ids_per_partition`, `max_unique_ids_per_partition` for stacked tables

Why do we not specify these two params? Should we specify them here? This should fix two issues - right now, the two params are not getting passed down because we are auto-stacking, and also prevent us ...
  • abheesht17
  • Opened 
    19 days ago
  • #166

keras-team/keras-rs
`DistributedEmbedding`: Add a way to record FDO stats and update input config (`max_ids`, `max_unique_ids`, etc.)

We often get warnings while training, similar to: 08:44:12.046114: W jax_tpu_embedding/sparsecore/lib/core/input_preprocessing_util.cc:251] No Coo Buffer Size provided for table cat_14_table_cat_15_table_cat_23_table_cat_24_table_cat_25_table_cat_33_table_cat_34_table_cat_35_table_cat_36_table, ...
  • abheesht17
  • Opened 
    on Aug 20
  • #145

keras-team/keras-rs
Multi-host training does not work for distributed embeddings

Code (sourced from this example): https://gist.github.com/abheesht17/0491a8d34756a208a000ff867ad7cc13 This is the error which crops up: https://gist.github.com/abheesht17/694a07e33d217b23692105e281a91fe4 ...
bug
  • abheesht17
  • 1
  • Opened 
    on Aug 17
  • #143

keras-team/keras-rs
NaNs in output of `DistributedEmbedding.preprocess()`

Reproduced this here: https://colab.research.google.com/gist/abheesht17/57575eaf845c248cbf1e760e1332a97c/distributed_embedding_jax.ipynb. Is this expected and just for padding?
bug
  • abheesht17
  • 2
  • Opened 
    on Aug 14
  • #141

keras-team/keras-rs
🗺️ KerasRS Roadmap 🗺️

- Large-Scale Benchmarks: Publish performance benchmarks and guides for our DistributedEmbedding layer, showcasing its ability to train models with massive embedding tables. - Add a rich set of ...
  • abheesht17
  • 1
  • Opened 
    on Aug 1
  • #137

keras-team/keras-rs
NaN loss when some embedding tables are on SC, and the rest are on default device

Reproduced here: https://colab.research.google.com/gist/abheesht17/0fd87eba48b18ad926428df177ab386c/dist-embedding-bug-mismatch-in-features-when-both-default-device-sc-are-used.ipynb. Did some basic debugging. ...
bug
  • abheesht17
  • 2
  • Opened 
    on Jul 29
  • #134

keras-team/keras-rs
Add support for extra dimensions to `SamplingProbabilityCorrection` and `HardNegativeMining`.

Same as in RemoveAccidentalHits.
  • hertschuh
  • Opened 
    on Mar 21
  • #39

keras-team/keras-rs
Mismanagement of Public and Private Symbols and Violation of Python's Official Interface Conventions

The current implementation of symbol management in the project relies on the namex package, which enforces a structure where the code must reside in src and then creates an api subpackage that references ...
  • AlanBogarin
  • 1
  • Opened 
    on Mar 20
  • #38

keras-team/keras-rs
Add `temperature`, `lambda_weights` to pairwise losses

  • abheesht17
  • 3
  • Opened 
    on Mar 1
  • #36

keras-team/keras-rs
Go over examples

There are a few things we need to look into for examples. - [ ] For the GRU4Rec example, we need to figure out if no loss reduction is the right way to do it (instead of sum_over_batch_size). The ...
  • abheesht17
  • Opened 
    on Feb 25
  • #34
Issue origami icon

Learn how you can use GitHub Issues to plan and track your work.

Save views for sprints, backlogs, teams, or releases. Rank, sort, and filter issues to suit the occasion. The possibilities are endless.Learn more about GitHub Issues
ProTip! 
Restrict your search to the title by using the in:title qualifier.
Issue origami icon

Learn how you can use GitHub Issues to plan and track your work.

Save views for sprints, backlogs, teams, or releases. Rank, sort, and filter issues to suit the occasion. The possibilities are endless.Learn more about GitHub Issues
ProTip! 
Restrict your search to the title by using the in:title qualifier.
Issue search results · GitHub