xuyige
|
5dc43c6e5a
|
create load dataset tutorial
|
2019-07-07 22:42:56 +08:00 |
|
Yige XU
|
3e37356f2a
|
Merge pull request #173 from zide05/tutorials
修改docs/source/user/tutorials.rst和docs/source/tutorials/tutorial_1_bat…
|
2019-07-05 21:38:00 +08:00 |
|
陈怡然
|
1fef29e372
|
修改docs/source/user/tutorials.rst和docs/source/tutorials/tutorial_1_batcher_loss_optimizer.rst
|
2019-07-05 20:11:24 +08:00 |
|
ChenXin
|
63e00bd84d
|
add more detail in README.md
|
2019-07-05 14:29:25 +08:00 |
|
ChenXin
|
6e6a31183f
|
make tutorials folder
|
2019-07-04 11:49:50 +08:00 |
|
ChenXin
|
1ccc730c42
|
basic framework of docs folder
|
2019-07-04 11:33:33 +08:00 |
|
ChenXin
|
aa5f67ef55
|
first commit on tutorials branch
|
2019-07-04 10:05:22 +08:00 |
|
yh_cc
|
4f91fb1ac1
|
Merge branch 'dev0.5.0' of github.com:fastnlp/fastNLP into dev0.5.0
|
2019-07-03 10:20:40 +08:00 |
|
yh_cc
|
84b18890a1
|
1.增加AdamW的optimizer;2.修复Trainer中metric_key的bug;3.静态embedding初始化修改;4.CrossEntropyLoss增加对reduction的支持'
|
2019-07-03 10:20:15 +08:00 |
|
xuyige
|
1bc780a2bd
|
update framework in matching tasks
|
2019-07-02 13:38:05 +08:00 |
|
xuyige
|
8e7a604b29
|
update documents in predictor
|
2019-07-02 13:34:26 +08:00 |
|
yh_cc
|
9b31e6e213
|
ner的readme更新
|
2019-07-02 11:32:45 +08:00 |
|
xuyige
|
ba34a56fa7
|
Create README.md
create reproduction/matching/README.md
|
2019-07-01 19:36:31 +08:00 |
|
yh_cc
|
2858ff84cf
|
更新data folder读取代码
|
2019-07-01 14:11:52 +08:00 |
|
yh_cc
|
999f8ac33f
|
增加joint_cws_parse的代码
|
2019-07-01 14:04:49 +08:00 |
|
yh
|
b0fe264e42
|
Tester中predict function的DataParallel并行
|
2019-07-01 12:04:17 +08:00 |
|
yh
|
3c984872d3
|
Tester数据并行
|
2019-07-01 01:03:30 +08:00 |
|
yh
|
f68b2c5382
|
Tester支持predict数据并行
|
2019-07-01 00:33:31 +08:00 |
|
yh
|
5f19601d20
|
支持predict数据并行
|
2019-07-01 00:01:45 +08:00 |
|
xuyige
|
15d9581e6d
|
fix a bug in predictor
|
2019-06-30 15:44:26 +08:00 |
|
yh_cc
|
9c1b4914d8
|
1.修复trainer中潜在多步更新bug; 2. LSTM的数据并行修改;3. embed_loader中bug修复, 且允许手动初始化;
|
2019-06-30 09:52:01 +08:00 |
|
yh
|
8b8d184026
|
Merge branch 'dev0.5.0' of https://github.com/fastnlp/fastNLP into dev0.5.0
|
2019-06-26 13:43:48 +08:00 |
|
yh
|
40c4d216d1
|
修改staticEmbedding的初始化方式,显示通过这种初始化在esmi上的snli更容易达到88的test acc
|
2019-06-26 13:39:14 +08:00 |
|
xuyige
|
50faa936b4
|
add RTE and QNLI loader
|
2019-06-25 17:22:10 +08:00 |
|
xuyige
|
bc5e071253
|
Delete matching.py
|
2019-06-24 21:56:14 +08:00 |
|
xuyige
|
238d4fbcd0
|
Merge remote-tracking branch 'origin/dev0.5.0' into dev0.5.0
|
2019-06-24 21:47:53 +08:00 |
|
xuyige
|
e0b23b16db
|
update data loader of matching
|
2019-06-24 21:44:43 +08:00 |
|
yh_cc
|
43d3380b73
|
1.修复Trainer初始化的多device bug; 2.在CrossEntropyLoss中增加seq_len
|
2019-06-24 18:31:38 +08:00 |
|
yh_cc
|
39dd086262
|
1.修改CrossEntropyLoss中存在的反直觉bug; 2.更新sequence labeling
|
2019-06-24 09:56:28 +08:00 |
|
yh_cc
|
add8edf701
|
Merge branch 'dev0.5.0' of github.com:fastnlp/fastNLP into dev0.5.0
|
2019-06-23 19:14:05 +08:00 |
|
yh_cc
|
e9137349c3
|
解决在DataParallel模型场景下无法进行参数匹配的问题
|
2019-06-23 19:13:49 +08:00 |
|
xuyige
|
4d9eb7c64a
|
update framework of matching
|
2019-06-23 18:44:36 +08:00 |
|
xuyige
|
3593f0a545
|
fix bugs in matching dataloader
|
2019-06-23 18:42:57 +08:00 |
|
xuyige
|
d1f531c049
|
update matching dataloader in reproduction/matching
|
2019-06-23 18:25:04 +08:00 |
|
yh_cc
|
5d0877583e
|
conflict merge
|
2019-06-23 15:04:33 +08:00 |
|
yh_cc
|
4b0c26d338
|
1.修复BertEmbedding中的bug; 2. 修复Batch, Field在进行类型转换时的bug
|
2019-06-23 15:02:34 +08:00 |
|
xuyige
|
0a2122a7f7
|
Merge branch 'dev0.5.0' of https://github.com/fastnlp/fastNLP into dev0.5.0
|
2019-06-22 22:54:17 +08:00 |
|
xuyige
|
d34c739a39
|
Update embedding.py
|
2019-06-22 22:53:52 +08:00 |
|
yh_cc
|
e57b8e4fd3
|
seq_len_to_mask修复测试失败的问题
|
2019-06-21 11:24:42 +08:00 |
|
yh_cc
|
8f7ed07441
|
1. 在vocabulary的from_dataset中增加no_create_entry_dataset选项,用于传递dev和test
2. 调整各种Embedding的实现,使得确保来自dev和test的未发现词使用unk的表示
3. 在Embedding中增加dropout_word的选项,使得可以随机drop掉词语
4. 以及其它若干小的bug
|
2019-06-21 11:06:35 +08:00 |
|
yh
|
0f4cf30301
|
LSTM修改错误
|
2019-06-19 23:59:40 +08:00 |
|
yh
|
6b9bc007ee
|
LSTM中修复潜在的DataParallel可能存在的问题, 并且删除init_method参数
|
2019-06-19 23:49:01 +08:00 |
|
yh
|
8a766f070b
|
seq_len_to_mask修改为直接使用max_len而不再和句中最大长度对比
|
2019-06-19 23:20:12 +08:00 |
|
yh
|
76e2330a2e
|
增加seq_len_to_mask对多卡场景的支持
|
2019-06-19 23:13:53 +08:00 |
|
yh
|
1167d3b587
|
再次修改elmo的dataparallel问题
|
2019-06-19 22:35:44 +08:00 |
|
yh
|
c4e131a0c5
|
重新修改ELMO与LSTM DataParallel的问题
|
2019-06-19 22:19:41 +08:00 |
|
yh
|
a137038eb2
|
修复ELMO与LSTM无法使用nn.DataParallel的问题
|
2019-06-19 19:43:53 +08:00 |
|
yunfan
|
15c7c073be
|
fix embed_loader
|
2019-06-19 17:04:52 +08:00 |
|
yh_cc
|
4533427ea3
|
sequence labeling更新
|
2019-06-19 11:14:41 +08:00 |
|
yh_cc
|
4d138ed7f8
|
Merge branch 'dev0.5.0' of github.com:fastnlp/fastNLP into dev0.5.0
|
2019-06-18 10:02:29 +08:00 |
|