Skip to content

Releases: keras-team/keras-hub

v0.11.0

03 May 02:53
@mattdangerw mattdangerw
v0.11.0
4296fd9
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Learn about vigilant mode.
Compare
Choose a tag to compare
v0.11.0

Summary

This release has no major feature updates, but changes the location our source code is help. Source code is split into a src/ and api/ directory with an explicit API surface similar to core Keras.

When adding or removing new API in a PR, use ./shell/api_gen.sh to update the autogenerated api/ files. See our contributing guide.

What's Changed

New Contributors

Full Changelog: v0.10.0...v0.11.0

Contributors

mattdangerw, sampathweb, and 3 other contributors
Assets 2
Loading

v0.10.0

29 Apr 18:16
@SamanehSaadat SamanehSaadat
v0.10.0
bd74d8e
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Learn about vigilant mode.
Compare
Choose a tag to compare
v0.10.0

Summary

  • Added support for Task (CausalLM and Classifier) saving and loading which allows uploading Tasks.
  • Added basic Model Card for Hugging Face upload.
  • Added support for a positions array in our RotaryEmbedding layer.

What's Changed

Full Changelog: v0.9.3...v0.10.0

Contributors

mattdangerw, sampathweb, and 2 other contributors
Assets 2
Loading
0 Join discussion

v0.9.3

10 Apr 21:30
@mattdangerw mattdangerw
v0.9.3
d38494a
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Learn about vigilant mode.
Compare
Choose a tag to compare
v0.9.3

Patch release with fixes for Llama and Mistral saving.

What's Changed

Full Changelog: v0.9.2...v0.9.3

Contributors

mattdangerw
Assets 2
Loading

v0.9.2

09 Apr 03:54
@mattdangerw mattdangerw
v0.9.2
4d10195
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Learn about vigilant mode.
Compare
Choose a tag to compare
v0.9.2

Summary

  • Initial release of CodeGemma.
  • Bump to a Gemma 1.1 version without download issues on Kaggle.

What's Changed

Full Changelog: v0.9.1...v0.9.2

Contributors

mattdangerw and SamanehSaadat
Assets 2
Loading
0 Join discussion

v0.9.1

06 Apr 02:39
@mattdangerw mattdangerw
v0.9.1
c764f98
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Learn about vigilant mode.
Compare
Choose a tag to compare
v0.9.1

Patch fix for bug with stop_token_ids.

What's Changed

Full Changelog: v0.9.0...v0.9.1

Contributors

mattdangerw
Assets 2
Loading
0 Join discussion

v0.9.0

06 Apr 00:42
@mattdangerw mattdangerw
v0.9.0
8731d1d
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Learn about vigilant mode.
Compare
Choose a tag to compare
v0.9.0

The 0.9.0 release adds new models, hub integrations, and general usability improvements.

Summary

  • Added the Gemma 1.1 release.
  • Added the Llama 2, BLOOM and ELECTRA models.
  • Expose new base classes. Allow from_preset() on base classes.
    • keras_nlp.models.Backbone
    • keras_nlp.models.Task
    • keras_nlp.models.Classifier
    • keras_nlp.models.CausalLM
    • keras_nlp.models.Seq2SeqLM
    • keras_nlp.models.MaskedLM
  • Some initial features for uploading to model hubs.
    • backbone.save_to_preset, tokenizer.save_to_preset, keras_nlp.upload_preset.
    • from_preset and upload_preset now work with the Hugging Face Models Hub.
    • More features (task saving, lora saving), and full documentation coming soon.
  • Numerical fixes for the Gemma model at mixed_bfloat16 precision. Thanks unsloth for catching!
# Llama 2. Needs Kaggle consent and login, see https://github.com/Kaggle/kagglehub
causal_lm = keras_nlp.models.LlamaCausalLM.from_preset(
    "llama2_7b_en",
    dtype="bfloat16", # Run at half precision for inference.
)
causal_lm.generate("Keras is a", max_length=128)
# Base class usage.
keras_nlp.models.Classifier.from_preset("bert_base_en", num_classes=2)
keras_nlp.models.Tokenizer.from_preset("gemma_2b_en")
keras_nlp.models.CausalLM.from_preset("gpt2_base_en", dtype="mixed_bfloat16")

What's Changed

New Contributors

Full Changelog: v0.8.2...v0.9.0

Contributors

TheCrazyT, RyanMullins, and 17 other contributors
Assets 2
Loading

v0.8.2

27 Feb 22:46
@mattdangerw mattdangerw
v0.8.2
c9baf2f
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Learn about vigilant mode.
Compare
Choose a tag to compare
v0.8.2

Summary

  • Mistral fixes for dtype and memory usage. #1458

What's Changed

Full Changelog: v0.8.1...v0.8.2.dev0

Contributors

mattdangerw and tirthasheshpatel
Assets 2
Loading
0 Join discussion

v0.8.1

22 Feb 01:24
@mattdangerw mattdangerw
v0.8.1
712f172
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Learn about vigilant mode.
Compare
Choose a tag to compare
v0.8.1

Minor fixes to Kaggle Gemma assets.

What's Changed

Full Changelog: v0.8.0...v0.8.1

Contributors

mattdangerw
Assets 2
Loading

v0.8.0

21 Feb 04:34
@mattdangerw mattdangerw
v0.8.0
cca2050
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Learn about vigilant mode.
Compare
Choose a tag to compare
v0.8.0

The 0.8.0 release focuses on generative LLM features in KerasNLP.

Summary

  • Added the Mistral and Gemma models.
  • Allow passing dtype directly to backbone and task constructors.
  • Add a settable sequence_length property to all preprocessing layers.
  • Added enable_lora() to the backbone class for parameter efficient fine-tuning.
  • Added layer attributes to backbone models for easier access to model internals.
  • Added AlibiBias layer.
# Pass dtype to a model.
causal_lm = keras_nlp.MistralCausalLM.from_preset(
    "mistral_instruct_7b_en",
    dtype="bfloat16"
)
# Settable sequence length property.
causal_lm.preprocessor.sequence_length = 128
# Lora API.
causal_lm.enable_lora(rank=4)
# Easy layer attributes.
for layer in causal_lm.backbone.transformer_layers:
    print(layer.count_params())

What's Changed

New Contributors

Full Changelog: v0.7.0...v0.8.0

Contributors

mattdangerw, sampathweb, and 5 other contributors
Assets 2
Loading
0 Join discussion

v0.7.0

05 Jan 22:29
@mattdangerw mattdangerw
v0.7.0
ce93ca8
This commit was created on GitHub.com and signed with GitHub’s verified signature. The key has expired.
GPG key ID: 4AEE18F83AFDEB23
Expired
Learn about vigilant mode.
Compare
Choose a tag to compare
v0.7.0

This release integrates KerasNLP and Kaggle Models. KerasNLP models will now work in Kaggle offline notebooks and all assets will quickly attach to a notebook rather than needing a slow download.

Summary

KerasNLP pre-trained models are now all made available through Kaggle Models. You can see all models currently available in both KerasCV and KerasNLP here. Individual model pages will include example usage and a file browser to examine all available assets for a model preset.

This change will not affect the existing usage of from_preset(). Statement like keras_nlp.models.BertClassifier.from_preset("bert_base_en") will continue to work and download checkpoints from the Kaggle Models hub.

A note on model saving—for saving support across Keras 2 and Keras 3, we recommend using the new Keras saved model format. You can use model.save('path/to/location.keras') for a full model and model.save_weights('path/to/location.weights.h5') for checkpoints. See the Keras saving guide for more details.

What's Changed

New Contributors

Full Changelog: v0.6.4...v0.7.0

Contributors

Philmod, mattdangerw, and 11 other contributors
Assets 2
Loading
0 Join discussion