94247ad6cb
Make num_train_optimization_steps int
2019-05-13 12:38:22 +02:00
49a77ac16f
Clean up a little bit
2019-05-12 00:31:10 +02:00
3bf3f9596f
Fixing the issues reported in https://github.com/huggingface/pytorch-pretrained-BERT/issues/556
...
Reason for issue was that optimzation steps where computed from example size, which is different from actual size of dataloader when an example is chunked into multiple instances.
Solution in this pull request is to compute num_optimization_steps directly from len(data_loader).
2019-05-12 00:13:45 +02:00
00c7fd2b79
Division to num_train_optimizer of global_step in lr_this_step is removed.
2019-05-09 10:57:03 +03:00
fa37b4da77
Merge branch 'master' of https://github.com/huggingface/pytorch-pretrained-BERT
2019-05-09 10:55:24 +03:00
5289b4b9e0
Division to num_train_optimizer of global_step in lr_this_step is removed.
2019-05-09 10:51:38 +03:00
0198399d84
Merge pull request #570 from MottoX/fix-1
...
Create optimizer only when args.do_train is True
2019-05-08 16:07:50 +02:00
18c8aef9d3
Fix documentation typo
2019-05-02 19:23:36 +08:00
74dbba64bc
Prepare optimizer only when args.do_train is True
2019-05-02 19:09:29 +08:00
2dee86319d
Merge pull request #527 from Mathieu-Prouveur/fix_value_training_loss
...
Update example files so that tr_loss is not affected by args.gradient…
2019-04-30 11:12:55 +02:00
87b9ec3843
Fix tr_loss rescaling factor using global_step
2019-04-29 12:58:29 +02:00
ed8fad7390
Update example files so that tr_loss is not affected by args.gradient_accumulation_step
2019-04-24 14:07:00 +02:00
d94c6b0144
fix training schedules in examples to match new API
2019-04-23 11:17:06 +02:00
3d78e226e6
Merge pull request #489 from huggingface/tokenization_serialization
...
Better serialization for Tokenizers and Configuration classes - Also fix #466
2019-04-16 08:49:54 +02:00
3571187ef6
fix saving models in distributed setting examples
2019-04-15 16:43:56 +02:00
1135f2384a
clean up logger in examples for distributed case
2019-04-15 15:22:40 +02:00
60ea6c59d2
added best practices for serialization in README and examples
2019-04-15 15:00:33 +02:00
179a2c2ff6
update example to work with new serialization semantic
2019-04-15 14:33:23 +02:00
c49ce3c722
fix tsv read error in Windows
2019-04-11 15:40:19 -04:00
d07db28f52
Fix typo in example code
...
Modify 'unambigiously' to 'unambiguously'
2019-03-31 01:20:18 +08:00
e5b63fb542
Merge branch 'master' of https://github.com/ananyahjha93/pytorch-pretrained-BERT
...
pull current master to local
2019-03-17 08:30:13 -04:00
8a4e90ff40
corrected folder creation error for MNLI-MM, verified GLUE results
2019-03-17 08:16:50 -04:00
e0bf01d9a9
added hack for mismatched MNLI
2019-03-16 14:10:48 -04:00
4c721c6b6a
added eval time metrics for GLUE tasks
2019-03-15 23:21:24 -04:00
043c8781ef
added code for all glue task processors
2019-03-14 04:24:04 -04:00
22a465a91f
Simplify code, delete redundancy line
...
delete redundancy line `if args.train`, simplify code.
2019-03-13 09:42:06 +08:00
994d86609b
fixing PYTORCH_PRETRAINED_BERT_CACHE use in examples
2019-03-06 10:21:24 +01:00
5c85fc3977
fix typo - logger info
2019-03-06 10:05:21 +01:00
0f96d4b1f7
Run classifier processor for SST-2.
2019-03-05 13:38:28 -06:00
eebc8abbe2
clarify and unify model saving logic in examples
2019-02-11 14:04:19 +01:00
848aae49e1
Merge branch 'master' into python_2
2019-02-06 00:13:20 +01:00
448937c00d
python 2 compatibility
2019-02-06 00:07:46 +01:00
1579c53635
more explicit notation: num_train_step => num_train_optimization_steps
2019-02-05 15:36:33 +01:00
5169069997
make examples consistent, revert error in num_train_steps calculation
2019-01-30 11:47:25 +01:00
9c6a48c8c3
fix learning rate/fp16 and warmup problem for all examples
2019-01-27 14:07:24 +01:00
01ff4f82ba
learning rate problems in run_classifier.py
2019-01-22 23:40:06 +01:00
c9fd350567
remove default when action is store_true in arguments
2019-01-07 13:01:54 +01:00
766c6b2ce3
Merge pull request #159 from jaderabbit/master
...
Allow do_eval to be used without do_train and to use the pretrained model in the output folder
2019-01-07 12:31:06 +01:00
77966a43a4
Merge pull request #156 from rodgzilla/cl_args_doc
...
Adding new pretrained model to the help of the `bert_model` argument.
2019-01-07 12:27:16 +01:00
193e2df8ba
Remove rogue comment
2019-01-03 13:13:06 +02:00
c64de50ea4
nb_tr_steps is not initialized
2019-01-03 12:34:57 +02:00
b96149a19b
Training loss is not initialized if only do_eval is specified
2019-01-03 10:32:10 +02:00
be3b9bcf4d
Allow one to use the pretrained model in evaluation when do_train is not selected
2019-01-03 09:02:33 +02:00
186f75342e
Adding new pretrained model to the help of the `bert_model` argument.
2019-01-02 14:00:59 +01:00
99709ee61d
loading saved model when n_classes != 2
...
Required to for: Assertion `t >= 0 && t < n_classes` failed, if your default number of classes is not 2.
2018-12-20 13:55:47 -08:00
ae88eb88a4
set encoding to 'utf-8' in calls to open
2018-12-14 13:48:58 +01:00
e1eab59aac
no fp16 on evaluation
2018-12-13 14:54:02 +01:00
087798b7fa
fix reloading model for evaluation in examples
2018-12-13 14:48:12 +01:00
0cf88ff084
make examples work without apex
2018-12-13 13:28:00 +01:00
d3fcec1a3e
add saving and loading model in examples
2018-12-13 12:50:44 +01:00
samuelbroscheit