Commit Graph

266 Commits

Author SHA1 Message Date
Guillaume B
12a1e0258c Addition of integration tests for Albert 2020-06-22 18:54:51 +02:00
Guillaume B
2897027ab0 Addition of AlbertForQuestionAnswering and AlbertForMultipleChoices 2020-06-22 18:29:16 +02:00
Guillaume B
8b2fda9f54 Addition of AlbertForMaskedLM 2020-06-21 11:50:15 +02:00
Guillaume B
d3e91204ae Addition of AlbertMLMHead 2020-06-21 10:50:58 +02:00
Guillaume B
f4afd35ed6 albert config implementation, download scripts 2020-06-16 19:07:37 +02:00
Guillaume B
bd2906fcfc Examples clean-up 2020-06-06 10:37:53 +02:00
Guillaume B
db3d126a2b non optional LayerState fields 2020-06-05 20:49:26 +02:00
Guillaume B
e0772f42ae Updated version 2020-06-05 09:45:12 +02:00
Guillaume B
40a6a148a0 Migration of translation pipeline to be non mutable 2020-06-04 23:05:35 +02:00
Guillaume B
3781040055 Migration of BART to non mutable generator (2) 2020-06-04 22:12:11 +02:00
Guillaume B
69e7000aeb Definition of Cache enum and implementation for GPT2 2020-06-04 13:00:10 +02:00
Guillaume B
3c010a077b Updated generation code for GPT and GPT2 to allow non mutable models 2020-06-03 19:08:40 +02:00
Guillaume B
829965d68b Model resources update, documentation 2020-05-25 22:01:53 +02:00
Guillaume B
ff0277925e Creation of translation pipeline, added documentation, created unit tests 2020-05-25 16:20:23 +02:00
Guillaume B
67ab4fc772 Updated copyright notices 2020-05-24 21:20:21 +02:00
Guillaume B
d6b5abe947 Split MarianForConditionalGeneration to own module 2020-05-24 20:15:07 +02:00
Guillaume B
cccad54194 Updates to BART for Marian compatibility 2020-05-24 17:00:04 +02:00
Guillaume B
d0fc3ff40d Addition of multi-label classification prediction method for sequence classification pipeline
Version update
2020-05-20 17:51:29 +02:00
Guillaume B
0bbb47d1db Added options for label consolidation for sub tokens 2020-05-11 16:40:25 +02:00
Guillaume B
6d61074f7f Updated consolidation and documentation 2020-05-10 12:27:42 +02:00
Guillaume B
1a9e315edf Added sub-token consolidation 2020-05-10 12:05:13 +02:00
Guillaume B
9d3a944051 Updated token classification pipeline to use next tokenization features, reference to original text 2020-05-10 11:38:28 +02:00
Guillaume B
053413fcbc Addition of generic classification model (following generic token classification pattern) 2020-05-07 17:26:20 +02:00
Guillaume B
83e43ffcd5 Updated resource list 2020-05-03 14:59:13 +02:00
Guillaume B
c6f5cdd859 Updated documentation 2020-05-03 13:37:18 +02:00
Guillaume B
029d4bd47c Addition of Electra resources 2020-05-03 10:05:32 +02:00
Guillaume B
5a1c1ae7a0 Addition of ElectraDiscriminator 2020-05-03 09:46:59 +02:00
Guillaume B
2c25dc1650 Addition of ElectraForTokenClassification 2020-05-01 10:00:47 +02:00
Guillaume B
4334fa1758 Addition of ElectraForMaskedLM 2020-05-01 08:47:52 +02:00
Guillaume B
5bec2548c1 Addition of Electra generator and discriminator heads 2020-04-29 18:59:37 +02:00
Guillaume B
45eeb7ae5b ElectraModel implementation, weights loaded, forward pass 2020-04-29 18:30:36 +02:00
Guillaume B
b2308d741d Addition of Electra Config and Embeddings 2020-04-28 18:51:04 +02:00
Guillaume B
194a447b06 Updated documentation 2020-04-26 11:49:40 +02:00
Guillaume B
cb254995fd Addition of dependencies download example 2020-04-26 10:51:29 +02:00
Guillaume B
0a85d5ba21 Updated NER Pipeline 2020-04-26 09:13:13 +02:00
Guillaume B
2aeace4a6e Updated Question Answering Pipeline 2020-04-26 09:08:04 +02:00
Guillaume B
69b819ce5e Updated Summarization and Sentiment pipelines 2020-04-26 08:57:20 +02:00
Guillaume B
c65943ab2f Updated model weight download 2020-04-24 17:04:28 +02:00
Guillaume B
5f0a36c1f9 Merge remote-tracking branch 'remotes/origin/master' into benchmarks_generation
# Conflicts:
#	examples/summarization.rs
2020-04-06 20:29:32 +02:00
Guillaume B
2d7f4c6743 updated summarization example 2020-04-06 15:36:12 +02:00
Guillaume B
35fcd0b6a3 updated summarization example 2020-04-06 14:45:52 +02:00
Guillaume B
793118bc94 - reset cache for BART model
- sentiment classifier optimization
- question answering pipeline optimization
2020-04-06 13:26:50 +02:00
Guillaume B
dc727e0baa Updated documentation 2020-04-05 11:40:08 +02:00
Guillaume B
6c7353de6e Addition of summarization pipeline and integration tests 2020-04-05 10:21:45 +02:00
Guillaume B
0a8832a012 End-to-end summarization 2020-04-04 17:11:37 +02:00
Guillaume B
fa87fce96e Summarization without beam search 2020-04-04 10:31:18 +02:00
Guillaume B
dbf6841610 Optimization of generation pipeline 2020-04-04 09:40:13 +02:00
Guillaume B
6aa6a4b2a2 Encoder input for summarization validated 2020-04-03 17:11:09 +02:00
Guillaume B
b8fe6a41e5 Updated language generation API to work with mutable models 2020-04-03 14:48:08 +02:00
Guillaume B
55db579baa ConditionalGeneration and SequenceClassification implemented 2020-04-02 16:16:05 +02:00
Guillaume B
fe529b2fec BartModel forward pass validated 2020-04-02 14:48:28 +02:00
Guillaume B
642874b22b BartModel implemented and weights loaded 2020-04-02 11:16:50 +02:00
Guillaume B
44a83ad4ab initial draft for cached attention 2020-03-29 19:31:12 +02:00
Guillaume B
efd2ed0509 BART config & initial input preparation 2020-03-29 12:01:13 +02:00
Guillaume B
eea1667f26 Added documentation for pipelines 2020-03-27 16:17:17 +01:00
Guillaume B
496f7ba0bb Completed documentation for DistilBERT 2020-03-24 19:59:00 +01:00
Guillaume B
6c6e5526ec Completed documentation for BERT 2020-03-22 15:39:22 +01:00
Guillaume B
e1f3b743da BERT documentation (ongoing) 2020-03-22 11:56:58 +01:00
Guillaume B
54b10c54a7 Updated lib structure and visibility 2020-03-22 09:54:49 +01:00
Guillaume B
f17b0d7da8 Lib clean-up and doc landing page 2020-03-21 16:53:15 +01:00
Guillaume B
24335a5883 Lib clean-up (default binary deleted, re-exports removed) 2020-03-21 11:35:41 +01:00
Guillaume B
56bf0f7691 Improved error message for examples 2020-03-19 18:28:56 +01:00
Guillaume B
344e27c927 Performance optimization for language generation 2020-03-17 20:06:45 +01:00
Guillaume B
64a01e3f73 End-to-end generation completed, updated README, bumped version 2020-03-16 16:17:23 +01:00
Guillaume B
5539e9360a Beam population loop 2020-03-16 12:38:49 +01:00
Guillaume B
2dbdb2c2be beam search preparation 2020-03-16 09:46:17 +01:00
Guillaume B
2621e0dc5d Added tests for GPT2 and OpenAI GPT generation 2020-03-15 21:09:46 +01:00
Guillaume B
e572cdb61a End to end generation without beam search 2020-03-15 20:54:21 +01:00
Guillaume B
6e61a54a0f Update of input ids, stop if eos_id and update attention mask 2020-03-15 20:04:36 +01:00
Guillaume B
c2ff52c217 Addition of top_k top_p sampling 2020-03-12 22:30:31 +01:00
Guillaume B
b6e186b7b0 Initial implementation of banned tokens detection 2020-03-12 18:49:45 +01:00
Guillaume B
61721ee2b8 Updated generation based on latest Transformers update (preparation) 2020-03-11 19:31:53 +01:00
Guillaume B
de86e57ee9 Updated generation based on latest Transformers update 2020-03-11 19:17:53 +01:00
Guillaume B
dd3415bbc1 Added repetition penalty 2020-03-10 20:58:18 +01:00
Guillaume B
c7727b0b97 Model forward pass for generation & GPT2 fix 2020-03-10 18:50:22 +01:00
Guillaume B
8f044b5bcc Updated generation preparation 2020-03-09 20:35:29 +01:00
Guillaume B
2d63fc3d18 Added tests for GPT2 2020-03-08 17:55:47 +01:00
Guillaume B
c98b689297 GPT2 download 2020-03-08 17:50:57 +01:00
Guillaume B
1923af0ac7 initial structure for generation 2020-03-08 17:08:18 +01:00
Guillaume B
596e0efc0e Added error message for SQuAD dataset environment variable 2020-03-07 15:49:09 +01:00
Guillaume B
123d90c40d Updated SQuAD example 2020-03-07 15:46:27 +01:00
Guillaume B
996fdb159f Updated QA for improved batching 2020-03-07 15:21:02 +01:00
Guillaume B
20bd0108c3 addition of SQuAD processor 2020-03-07 11:23:05 +01:00
Guillaume B
999f6e52aa initial commit for batch processing of QA inputs 2020-03-06 21:40:32 +01:00
Guillaume B
89b486fb12 Added integration test for OpenAI GPT 2020-03-01 11:29:57 +01:00
Guillaume B
d820ea8eea End-to-end OpenAI GPT model implemented 2020-03-01 11:26:34 +01:00
Guillaume B
ad1dbfcbc2 Initial commit for GPT model 2020-03-01 10:59:48 +01:00
Guillaume B
7a9cf64c20 Completed GPT2 example and integration test 2020-02-29 16:07:47 +01:00
Guillaume B
d038626d37 Updated GPT2 implementation, validated 2020-02-29 14:13:08 +01:00
Guillaume B
a276e65cf4 GPT2 Model implemented 2020-02-29 11:39:19 +01:00
Guillaume B
c12c0c479e Calculation of GPT2 embeddings 2020-02-29 10:56:56 +01:00
Guillaume B
811b0ad483 Implementation of Transformer Block 2020-02-28 21:09:42 +01:00
Guillaume B
e15dfe127d Implementation of Attention 2020-02-28 20:28:33 +01:00
Guillaume B
b0e84fc2b9 initial commit for GPT2 2020-02-27 18:55:17 +01:00
Guillaume B
aa819284df updated license headers 2020-02-25 18:39:08 +01:00
Guillaume B
fe4f7c689d Updated README, bumped version 2020-02-25 18:37:02 +01:00
Guillaume B
afa606152b End to end question answering pipeline implemented 2020-02-25 18:32:26 +01:00
Guillaume B
ae11677e2d Updated position embeddings, features stacking and sinusoidal embeddings config overwrite 2020-02-25 16:02:59 +01:00
Guillaume B
e89cfd5ce6 Example text update 2020-02-25 11:40:49 +01:00
Guillaume B
c11ce51abd minor refactoring 2020-02-25 11:40:16 +01:00
Guillaume B
384ea30612 Creation of QaFeatures initial commit 2020-02-25 11:27:59 +01:00
Guillaume B
1a7da783a9 Created QaExample and parsing from question and context pair 2020-02-24 16:24:43 +01:00
Guillaume B
ed82d934e6 Created QaExample and parsing from question and context pair 2020-02-24 16:23:44 +01:00
Guillaume B
a748a2bdc4 Added test for NER and example for sentiment 2020-02-24 14:27:23 +01:00
Guillaume B
6a3bfee4a3 Added pre-trained NER model to pipelines 2020-02-24 14:19:14 +01:00
Guillaume B
448638c338 Updated tokenizer dependency 2020-02-21 21:38:09 +01:00
Guillaume B
78cacdaf2e Added integration test for RoBERTa MaskedLM 2020-02-19 20:55:50 +01:00
Guillaume B
ee64db897e Initial commit for RoBERTa 2020-02-19 20:52:27 +01:00
Guillaume B
a449684f49 Completed BertForMaskedLM and added integration test 2020-02-17 21:44:12 +01:00
Guillaume B
f3d4736c13 Attention mask tested & validated 2020-02-17 19:11:01 +01:00
Guillaume B
765d2c88de BertModel implemented (incl. loading weights) 2020-02-17 19:02:47 +01:00
Guillaume B
8526a6b9bc Added BertSelfAttention 2020-02-16 17:19:12 +01:00
Guillaume B
20ec99e159 Added BertEmbeddings 2020-02-16 15:16:54 +01:00
Guillaume B
85d17e2736 Initial commit for BERT implementation 2020-02-16 14:29:35 +01:00
Guillaume B
a7264ff0f2 Updated tokenization dependency (no longer requires rust nightly) 2020-02-15 11:37:50 +01:00
Guillaume B
d6716a921d Addition of masked LM for DistilBERT (incl. example) 2020-02-13 22:19:03 +01:00