yh_cc
|
e1ed6f16e4
|
move model to device in DistTrainer
|
2021-10-15 11:01:01 +08:00 |
|
yh_cc
|
972185dc6c
|
增加DistTrainer中的batch_per_epoch属性
|
2021-09-25 22:53:05 +08:00 |
|
hw
|
c0aa5bd932
|
增加Mr, R8, R52, Ohsumed, 20ng数据集的loader ,pipe (#382)
* 增加Mr, R8, R52, Ohsumed, 20ng数据集的loader ,pipe
* "更改_20ng为NG20"
|
2021-09-23 14:40:07 +08:00 |
|
yh_cc
|
a5d608c781
|
Merge branch 'dev' of github.com:fastnlp/fastNLP into dev
|
2021-09-22 20:06:14 +08:00 |
|
yh_cc
|
c72bc070a9
|
1.修复yxg同学发现的DataSet copy Field导致的bug; 2. 修改DistTrainer中update_every的bug; 3.Trainer和DistTrainer修改pin_memory默认值为True
|
2021-09-18 00:00:38 +08:00 |
|
yh_cc
|
7ce03a601b
|
1.修改Trainer中的pin_memory参数; 2.修改DistTrainer使得DistTrainer和Trainer的api使用可以尽量接近
|
2021-09-13 10:26:45 +08:00 |
|
Xipeng Qiu
|
b127963f21
|
Update README.md
|
2021-07-13 13:14:00 +08:00 |
|
yh_cc
|
e9c6bf751b
|
Merge branch 'dev' of gitee.com:fastnlp/fastNLP into dev
|
2021-04-29 19:39:57 +08:00 |
|
yh_cc
|
62dd8b433c
|
1.trainer中修改evaluate的逻辑让valid_every不等于-1时在结束epoch需要evaluate; 2.修复BCELoss中的bug
|
2021-04-29 19:39:50 +08:00 |
|
LouChao98
|
595fc1d455
|
mirror fix (#361)
|
2021-04-22 09:57:46 +08:00 |
|
WillQvQ
|
099a14402c
|
!9 一些修复汇总
Merge pull request !9 from WillQvQ/dev
|
2021-04-12 21:45:28 +08:00 |
|
WillQvQ
|
0f4dad107e
|
Merge branch 'dev' of github.com:fastnlp/fastNLP into dev
|
2021-04-12 10:31:38 +08:00 |
|
WillQvQ
|
d9e0450ae7
|
Merge branch 'master' of github.com:fastnlp/fastNLP into dev
|
2021-04-12 10:30:28 +08:00 |
|
yh_cc
|
3e9330d250
|
解决https://github.com/fastnlp/fastNLP/issues/355 的问题
|
2021-03-30 11:12:26 +08:00 |
|
sleepy-owl
|
2b91830c26
|
Flaky test fix (#352)
|
2021-02-23 19:04:05 +08:00 |
|
yh_cc
|
7b4e099c52
|
修复spacy版本升级导致的模型加载失败
|
2021-02-22 12:01:18 +08:00 |
|
yh_cc
|
04ad8e604e
|
DataSet apply的时候可以传入use_tqdm和tqdm_desc
|
2021-01-05 23:45:15 +08:00 |
|
yh_cc
|
4886dbff8b
|
1.trainer增加异常时输出当前batch的indice; 2.BERTEmbedding修改一个typo; 3.Loss函数进行update;4.seq2seq_generator修改bias防止没有输出eos
|
2021-01-05 19:31:00 +08:00 |
|
yh_cc
|
33a94a3d1b
|
update for inf
|
2020-12-27 20:27:42 +08:00 |
|
yh_cc
|
1ec10c48e6
|
Merge branch 'dev' of gitee.com:fastnlp/fastNLP into dev
|
2020-12-27 16:02:06 +08:00 |
|
yh_cc
|
0a2f546b70
|
1.更新权重下载url; 2.更新seq2seq,方式第一个位置预测eos
|
2020-12-27 16:01:51 +08:00 |
|
WillQvQ
|
22839c215a
|
update .Jenkinsfile.
|
2020-12-27 15:52:39 +08:00 |
|
Yige Xu
|
030e0aa3ee
|
update some function about bert and roberta
|
2020-12-26 09:29:17 +08:00 |
|
ROGERDJQ
|
f17343e19b
|
[bugfix]修复fitlogcallback在disttrainner中无法添加dev_data 的问题 (#348)
fix the distTrainer dev_data
|
2020-12-16 18:04:22 +08:00 |
|
willqvq
|
84776696cd
|
给 _get_dataset_url 增加了使用自定义数据集下载路径的能力
|
2020-12-16 10:40:00 +08:00 |
|
Yige Xu
|
bffde7857a
|
fix a bug in early exit of bert
|
2020-12-14 16:06:44 +08:00 |
|
Yige Xu
|
bf9d834821
|
Merge branch 'dev' of https://gitee.com/fastnlp/fastNLP into dev
|
2020-12-14 15:34:50 +08:00 |
|
yh_cc
|
d4fda68840
|
Merge branch 'dev' of github.com:fastnlp/fastNLP into dev
|
2020-12-11 14:20:32 +08:00 |
|
yh_cc
|
057fa63d7e
|
解决dytpe的问题
|
2020-12-11 14:15:41 +08:00 |
|
yh_cc
|
dc280fa1cb
|
删除对spacy的依赖
|
2020-12-11 14:03:10 +08:00 |
|
yh_cc
|
7ca8d045e7
|
update for basemodel bug
|
2020-12-11 13:42:18 +08:00 |
|
yh_cc
|
50d7cfdd7a
|
在dataset中添加concat函数,支持将两个dataset concat起来
|
2020-12-09 19:45:55 +08:00 |
|
yh_cc
|
d6072ba1d3
|
update tutorial
|
2020-12-08 15:30:49 +08:00 |
|
yh_cc
|
afe3af76de
|
update tutorial
|
2020-12-08 14:24:19 +08:00 |
|
johnson7788
|
9fdcafff6a
|
"bugs:fix lstm rnn.pack_padded_sequence RuntimeError, 详细信息: https://github.com/pytorch/pytorch/issues/43227" (#345)
zhicai.guo <zhicai.guo@lavector.com>
|
2020-12-07 13:16:17 +08:00 |
|
yh_cc
|
24694a8485
|
update tutorial
|
2020-12-06 18:22:35 +08:00 |
|
yh_cc
|
46c3020656
|
update tutorial
|
2020-12-06 14:06:31 +08:00 |
|
yh_cc
|
801d19d340
|
Merge branch 'dev' of gitee.com:fastnlp/fastNLP into dev
|
2020-12-06 13:27:06 +08:00 |
|
yh_cc
|
9939528760
|
修复Trainer fp16的bug; 添加使用中文词进行分类的例子
|
2020-12-06 13:27:00 +08:00 |
|
yunfan
|
f711d3070a
|
dist_trainer for fp16
|
2020-12-06 03:48:08 +00:00 |
|
yh_cc
|
69a6cbbf09
|
bug fix for Trainer fp16
|
2020-12-06 00:42:42 +08:00 |
|
yh_cc
|
2e8350cd16
|
Trainer支持fp16
|
2020-12-05 22:30:23 +08:00 |
|
yh_cc
|
35d0371955
|
update DistTrainer
|
2020-12-05 13:33:02 +08:00 |
|
yh_cc
|
3e2380aa0c
|
t Merge branch 'dev' of gitee.com:fastnlp/fastNLP into dev
|
2020-11-28 20:56:37 +08:00 |
|
yh_cc
|
eaff7cafc6
|
update to avoid warning
|
2020-11-28 20:53:24 +08:00 |
|
WillQvQ
|
22c6e6d59c
|
!5 修复了一些错误
Merge pull request !5 from WillQvQ/dev
|
2020-11-23 18:50:37 +08:00 |
|
willqvq
|
350772f518
|
修改读取测试数据的位置
|
2020-11-23 13:34:05 +08:00 |
|
willqvq
|
fc8f8f6913
|
把测试文件夹 test 改名为 tests
|
2020-11-23 13:00:31 +08:00 |
|
yh_cc
|
2ba336945c
|
1.在DistTrainer中增加对sampler的控制; 2.在trainer和disttrainer中支持其它optimizer
|
2020-11-22 20:14:44 +08:00 |
|
WillQvQ
|
c49c849aee
|
!4 update fastNLP/modules/encoder/seq2seq_encoder.py.
Merge pull request !4 from 刘向阳/N/A
|
2020-11-18 12:13:51 +08:00 |
|