28ba345ecc
Fixing unused weight_decay argument
...
Currently the L2 regularization is hard-coded to "0.01", even though there is a --weight_decay flag implemented (that is unused). I'm making this flag control the weight decay used for fine-tuning in this script.
2019-08-04 12:31:46 -04:00
44dd941efb
link to `swift-coreml-transformers`
2019-08-01 09:50:30 -04:00
f2a3eb987e
Fix small typos
2019-07-31 11:05:06 -04:00
97091acb8c
Small spelling fix
2019-07-31 10:37:56 -04:00
769bb643ce
Fixing a broken link.
2019-07-31 10:22:41 -04:00
6b763d04a9
Merge pull request #911 from huggingface/small_fixes
...
Small fixes
2019-07-26 21:36:21 +02:00
7b6e474c9a
fix #901
2019-07-26 21:26:44 +02:00
632d711411
fix #908
2019-07-26 21:14:37 +02:00
c054b5ee64
Merge pull request #896 from zijunsun/master
...
fix multi-gpu training bug when using fp16
2019-07-26 19:31:02 +02:00
f0aeb7a814
multi-gpu training also should be after apex fp16（squad）
2019-07-26 15:23:29 +08:00
46cc9dd2b5
Merge pull request #899 from sukuya/master
...
Fixed import to use torchscript flag.
2019-07-25 15:03:21 +02:00
6219ad7216
Merge pull request #888 from rococode/patch-1
...
Update docs for parameter rename
2019-07-25 15:01:22 +02:00
0b6122e96a
Merge pull request #882 from Liangtaiwan/squad_v1_bug
...
fix squad v1 error (na_prob_file should be None)
2019-07-25 14:59:59 +02:00
c244562cae
Merge pull request #893 from joelgrus/patch-2
...
make save_pretrained do the right thing with added tokens
2019-07-25 14:58:48 +02:00
e1e2ab3482
Merge pull request #1 from sukuya/sukuya-patch-1
...
Update torchscript.rst
2019-07-25 16:53:11 +08:00
35c52f2f3c
Update torchscript.rst
...
Import fixed to pytorch_transformers else torchscript flag can't be used.
2019-07-25 16:51:11 +08:00
adb3ef6368
multi-gpu training also should be after apex fp16
2019-07-25 13:09:10 +08:00
ae152cec09
make save_pretrained work with added tokens
...
right now it's dumping the *decoder* when it should be dumping the *encoder*. this fixes that.
2019-07-24 16:54:48 -07:00
66b15f73f0
Update docs for parameter rename
...
OpenAIGPTLMHeadModel now accepts `labels` instead of `lm_labels`
2019-07-24 11:27:08 -07:00
a7fce6d917
fix squad v1 error (na_prob_file should be None)
2019-07-24 16:11:36 +08:00
067923d326
Merge pull request #873 from huggingface/identity_replacement
...
Add nn.Identity replacement for old PyTorch
2019-07-23 18:16:35 +02:00
368670ac31
Merge pull request #866 from xanlsh/master
...
Rework how PreTrainedModel.from_pretrained handles its arguments
2019-07-23 18:05:30 +02:00
1383c7b87a
Fix #869
2019-07-23 17:52:20 +02:00
6070b55443
fix #868
2019-07-23 17:46:01 +02:00
2c9a3115b7
fix #858
2019-07-23 16:45:55 +02:00
4fb56c7729
Remove unused *args parameter from PreTrainedConfig.from_pretrained
2019-07-23 10:43:01 -04:00
e179c55490
Add docs for from_pretrained functions, rename return_unused_args
2019-07-23 10:43:01 -04:00
fec76a481d
Update readme
2019-07-23 16:05:29 +02:00
859c441776
Merge pull request #872 from huggingface/saving_schedules
...
Updating schedules for state_dict saving/loading
2019-07-23 16:03:06 +02:00
0740e63e49
updating schedules for state_dict saving
2019-07-23 15:57:18 +02:00
268c6cc160
Merge pull request #845 from rabeehk/master
...
fixed version issues in run_openai_gpt
2019-07-23 15:29:31 +02:00
1d7d01c080
Merge pull request #847 from lpq29743/master
...
typos
2019-07-23 15:28:31 +02:00
c4bc66886d
Merge pull request #860 from Yiqing-Zhou/patch-1
...
read().splitlines() -> readlines()
2019-07-23 15:24:25 +02:00
ba52fe69d5
update breaking change section regarding from_pretrained keyword arguments
2019-07-23 15:10:02 +02:00
b1019d2a8e
token[-1] -> token.rstrip('\n')
2019-07-23 20:41:26 +08:00
0227b4a940
fix #827
2019-07-23 14:06:43 +02:00
490ebbdcf7
Fix PretrainedModel.from_pretrained not passing cache_dir forward
2019-07-22 18:03:08 -04:00
b8009cb0da
Make PreTrainedModel.from_pretrained pass unused arguments to model
2019-07-22 18:03:08 -04:00
bef0c629ca
fix
...
Remove '\n' before adding token into vocab
2019-07-22 22:30:49 +08:00
897d0841be
read().splitlines() -> readlines()
...
splitlines() does not work as what we expect here for bert-base-chinese because there is a '\u2028' (unicode line seperator) token in vocab file. Value of '\u2028'.splitlines() is ['', ''].
Perhaps we should use readlines() instead.
2019-07-22 20:49:09 +08:00
2f869dc665
Fixed typo
2019-07-21 11:05:36 -04:00
76be189b08
typos
2019-07-21 20:39:42 +08:00
f63ff536ad
fixed version issues in run_openai_gpt
2019-07-20 12:43:07 +02:00
a615499076
Merge pull request #797 from yzy5630/fix-examples
...
fix some errors for distributed lm_finetuning
2019-07-18 23:32:33 +02:00
dbecfcf321
Merge pull request #815 from praateekmahajan/update-readme-link
...
Update Readme link for Fine Tune/Usage section
2019-07-18 18:30:32 +02:00
acc48a0cc9
typos
2019-07-18 09:54:04 -04:00
a1fe4ba9c9
use new API for save and load
2019-07-18 15:45:23 +08:00
0d46b17553
Update Readme
...
Incorrect link for `Quick tour: Fine-tuning/usage scripts`
2019-07-17 22:50:10 -07:00
a7ba27b1b4
add parser for adam
2019-07-18 08:52:51 +08:00
d6522e2873
change loss and optimizer to new API
2019-07-17 21:22:34 +08:00
Ethan Perez