SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity
3,429 Commits 5 Branches 239 Tags
cbf8f5d32bdbfaef2d31daba0bfdf14fe2640d0b
Commit Graph

342 Commits

Author SHA1 Message Date
Julien Chaumond
cbf8f5d32b [model upload] Support for organizations 2020-03-09 17:33:57 -04:00
Lysandre Debut
5164ea91a7 Skipping outputs (#3116) 
* Minimal example

* Proposal 2

* Proposal 2 for fast tokenizers

* Typings

* Docs

* Revert "Docs" for easier review

This reverts commit eaf0f97062e809887704a542144c537f769d5223.

* Remove unnecessary assignments

* Tests

* Fix faulty type

* Remove prints

* return_outputs -> model_input_names

* Revert "Revert "Docs" for easier review"

This reverts commit 6fdc69408102bf695797f2dfddbb6350c6b9e722.

* code quality
 2020-03-09 13:48:58 -04:00
Patrick von Platen
847d370301 fix typo 2020-03-09 16:18:29 +01:00
Patrick von Platen
3e624c64ca fix repetition penalty mask in tf 2020-03-09 14:55:11 +01:00
Lysandre Debut
0ae91c80aa Change back pipeline signatures (#3105) 
* Change back pipeline signatures

* String types for non-imported objects
 2020-03-06 17:26:18 -05:00
Sam Shleifer
ed37f9fa4f [Bart] _prepare_decoder_inputs should use large negative (#3158) 2020-03-06 16:06:36 -05:00
Thomas Wolf
0416d437fb Merge pull request #3148 from patrickvonplaten/refactoring_beam_search_for_tf_2 
refactored beam search according to torch implementation
 2020-03-06 22:01:46 +01:00
Funtowicz Morgan
db9279dedb Fix QA models binding for Flaubert, XLNet and XLM. (#3100) 
Signed-off-by: Morgan Funtowicz <morgan@huggingface.co>

Format & quality

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

Again.

Signed-off-by: Morgan Funtowicz <morgan@huggingface.co>
 2020-03-06 13:04:29 -05:00
Sam Shleifer
e58b3ec5df add imports to examples (#3160) 2020-03-06 11:15:33 -05:00
Thomas Wolf
6ffe03a0a1 Merge pull request #3137 from tomhosking/bart-refactor 
Refactor BartModel so that input checks are handled within enc/dec
 2020-03-06 13:06:34 +01:00
Thomas Wolf
3e5da38dae Merge pull request #3132 from huggingface/hf_api_model_list 
[hf_api] Get the public list of all the models on huggingface
 2020-03-06 13:05:52 +01:00
Thomas Wolf
9499a3778e Merge pull request #3103 from gthb/keras-serialization 
Support keras JSON/HDF5 serialization of main layers
 2020-03-06 12:59:13 +01:00
patrickvonplaten
9362eb4a07 refactored beam search according to torch implementation 2020-03-06 00:46:29 +01:00
Sam Shleifer
857e0a0d3b Rename BartForMaskedLM -> BartForConditionalGeneration (#3114) 
* improved documentation
 2020-03-05 17:41:18 -05:00
Lysandre Debut
146c521235 Merge branch 'master' into add_models_special_tokens_to_specific_configs 2020-03-05 17:24:42 -05:00
Lysandre Debut
b623ddc000 Pass kwargs to configuration (#3147) 
* Pass kwargs to configuration

* Setter

* test
 2020-03-05 17:16:57 -05:00
Lysandre Debut
0001d05686 Correct missing keys + test (#3143) 2020-03-05 17:01:54 -05:00
Thomas Wolf
1741d740f2 Merge pull request #3145 from sshleifer/bartfp16 
[Bart] FP16 Support
 2020-03-05 22:14:35 +01:00
sshleifer
14d40584b2 remove newline 2020-03-05 13:06:35 -05:00
sshleifer
1360dacaa3 cleanup deltas 2020-03-05 12:57:42 -05:00
sshleifer
810079de1f no ipdb 2020-03-05 12:48:14 -05:00
sshleifer
c36fdc88d4 tests pass 2020-03-05 12:33:08 -05:00
Tom Hosking
31acb8dc52 Remove rogue .DS_Store 2020-03-05 13:51:30 +00:00
Tom Hosking
06a6cb6f36 Refactor BartModel so that input checks are handled within BartEncoder and BartDecoder 2020-03-05 13:45:41 +00:00
Patrick von Platen
e33ed12c3b uncomment expression 2020-03-05 13:41:04 +01:00
Patrick von Platen
4220fd52b9 remove ipdb 2020-03-05 13:36:21 +01:00
Patrick von Platen
c47394b0c9 refactoring and bug fixing beam search generate 2020-03-05 13:12:50 +01:00
Gunnlaugur Thor Briem
4c91a3af94 Document keras_serializable decorator 2020-03-05 11:48:10 +00:00
Gunnlaugur Thor Briem
4be01e5cbf Use name transformers_config in Keras serialization 
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.
 2020-03-05 11:47:35 +00:00
Gunnlaugur Thor Briem
a355f4f0fc Add functools.wraps for wrapper initializer 
Preserve the original initializer function's metadata. See
https://docs.python.org/3/library/functools.html#functools.update_wrapper
 2020-03-05 11:18:50 +00:00
Gunnlaugur Thor Briem
d262a5d48e fix: remove unused import 2020-03-05 11:05:29 +00:00
Julien Chaumond
3f067f4409 [hf_api] slightly more doc 2020-03-04 23:55:46 -05:00
Julien Chaumond
f564f93c84 [hf_api] Get the public list of all the models on huggingface 2020-03-04 23:33:09 -05:00
Lysandre
07a79db505 Fix failing doc samples 2020-03-04 19:11:31 -05:00
Gunnlaugur Thor Briem
4f338ed407 Explicit config_class instead of module inspection 2020-03-04 23:45:29 +00:00
Gunnlaugur Thor Briem
6fe1cc0874 fix: clean up inadvertent change in tf_t5 
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.
 2020-03-04 23:24:15 +00:00
Thomas Wolf
bdd3d0c76d Merge pull request #3118 from patrickvonplaten/add_beam_search_to_generation_tf_2_0 
Add beam search to generation tf 2 0
 2020-03-04 23:28:00 +01:00
Thomas Wolf
3b7f95a506 Merge pull request #3115 from gthb/fix-bogus-param-to-layer-init 
fix: passing config as Layer trainable param
 2020-03-04 21:59:09 +01:00
Patrick von Platen
006097f8ad rename variables named 'word' to 'token' in generate fn (#3119) 
* fix conflits

* fixed naming bug

* make style
 2020-03-04 12:01:17 -05:00
Gunnlaugur Thor Briem
18f4b9274f fix: work with Tensorflow < 2.1.0 
tf.keras.utils.register_keras_serializable was added in TF 2.1.0, so
don't rely on it being there; just decorate the class with it if it
exists.
 2020-03-04 16:57:29 +00:00
Patrick von Platen
7a89a3e493 correct beam search sampling 2020-03-04 17:27:47 +01:00
Patrick von Platen
c4c4c9998a make GPT2 and CTRL shape consistent between torch and TF 2020-03-04 17:27:47 +01:00
patrickvonplaten
2529b2d37e set redorder past sort dimension to its default 2020-03-04 17:27:47 +01:00
patrickvonplaten
61fef6e957 added beam_search generation for tf 2.0 2020-03-04 17:27:47 +01:00
Patrick von Platen
6701fb7859 fix beam_search behavior when sampling (#3106) 
* fix beam_search behavior when sampling

* delete print

* make correct style
 2020-03-04 09:30:51 -05:00
Gunnlaugur Thor Briem
b1116fd673 fix: passing config as Layer trainable param 
Lurking bugs discovered while working on other stuff.
 2020-03-03 23:05:40 +00:00
Gunnlaugur Thor Briem
96c4990165 fix unused imports and style 2020-03-03 22:57:05 +00:00
Gunnlaugur Thor Briem
470753bcf5 Put @keras_serializable only on layers it works on 
And only run the test on TF*MainLayer classes so marked.
 2020-03-03 22:44:45 +00:00
Gunnlaugur Thor Briem
0c716ede8c Use class decorator instead of superclass 
When supplied by Keras deserialization, the config parameter to initializers
will be a dict. So intercept it and convert to PretrainedConfig object (and
store in instance attribute for get_config to get at it) before passing to the
actual initializer. To accomplish this, and repeat as little code as possible,
use a class decorator on TF*MainLayer classes.
 2020-03-03 22:31:42 +00:00
Sam Shleifer
e9e6efdc45 BartForSequenceClassification: fix num_labels, add test (#3110) 2020-03-03 15:54:29 -05:00