GitLab

Refactor BartModel so that input checks are handled within enc/dec

[hf_api] Get the public list of all the models on huggingface

Support keras JSON/HDF5 serialization of main layers

fix missed BartForMaskedLM renaming

* improved documentation

Add models special tokens to its pretrained configs

* Pass kwargs to configuration

* Setter

* test

[Bart] FP16 Support

Refactoring and bug fixing beam search generate

Signed-off-by: Morgan Funtowicz <morgan@huggingface.co>

Signed-off-by: Morgan Funtowicz <morgan@huggingface.co>

* add empty model cards for every current DeepPavlov model

* fix: replace cyrillic `с` with `c`

* docs: add model cards for current DeepPavlov BERT models

* docs: add links for arXiv preprints

Signed-off-by: Morgan Funtowicz <morgan@huggingface.co>

Be explicit that this is config for the transformers package (as these
layers may coexist with other custom stuff in a Keras model, plus the
Keras container itself is called config, and config["config"] is not
great)

Add explicit error handling for initializer calls that have neither
the `config` nor the `transformers_config` argument, or have both.

Preserve the original initializer function's metadata. See
https://docs.python.org/3/library/functools.html#functools.update_wrapper

Signed-off-by: Morgan Funtowicz <morgan@huggingface.co>

This was the beginnings of an attempt to address the test failure on
this layer, and instead I backed out of making this layer
keras-serializable at all ... so it was a mistake to commit this.

Add beam search to generation tf 2 0

fix: passing config as Layer trainable param