Commit Graph
-
1bc780a2update framework in matching tasks by 2019-07-02 13:38:05 +0800 -
8e7a604bupdate documents in predictor by 2019-07-02 13:34:26 +0800 -
9b31e6e2ner的readme更新 by 2019-07-02 11:32:45 +0800 -
ba34a56fCreate README.md by 2019-07-01 19:36:31 +0800 -
2858ff84更新data folder读取代码 by 2019-07-01 14:11:52 +0800 -
999f8ac3增加joint_cws_parse的代码 by 2019-07-01 14:04:49 +0800 -
b0fe264eTester中predict function的DataParallel并行 by 2019-07-01 12:04:17 +0800 -
cdcc235dMerge pull request #3 from fastnlp/dev0.5.0 by 2019-07-01 11:42:14 +0800 -
610791adupdate Readme.md by 2019-07-01 11:39:02 +0800 -
204bb067fix random process by 2019-07-01 02:24:18 +0000 -
205f8fe6set random seed by 2019-06-30 17:45:09 +0000 -
406e1f7ffix and improve star_trans on SST by 2019-06-30 17:42:09 +0000 -
3c984872Tester数据并行 by 2019-07-01 01:03:30 +0800 -
f68b2c53Tester支持predict数据并行 by 2019-07-01 00:33:31 +0800 -
5f19601d支持predict数据并行 by 2019-07-01 00:01:45 +0800 -
fcc5a9f8Merge pull request #7 from fastnlp/dev0.5.0 by 2019-06-30 16:18:14 +0800 -
15d9581efix a bug in predictor by 2019-06-30 15:44:26 +0800 -
2e4d84d9Merge pull request #5 from fastnlp/dev0.5.0 by 2019-06-30 15:43:29 +0800 -
9c1b49141.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； by 2019-06-30 09:52:01 +0800 -
8b8d1840Merge branch 'dev0.5.0' of https://github.com/fastnlp/fastNLP into dev0.5.0 by 2019-06-26 13:43:48 +0800 -
40c4d216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 by 2019-06-26 13:39:14 +0800 -
50faa936add RTE and QNLI loader by 2019-06-25 17:22:10 +0800 -
bc5e0712Delete matching.py by 2019-06-24 21:56:14 +0800 -
238d4fbcMerge remote-tracking branch 'origin/dev0.5.0' into dev0.5.0 by 2019-06-24 21:47:53 +0800 -
e0b23b16update data loader of matching by 2019-06-24 21:44:43 +0800 -
79762c4cAdd Summarization framework by 2019-06-24 19:24:14 +0800 -
43d3380b1.修复Trainer初始化的多device bug; 2.在CrossEntropyLoss中增加seq_len by 2019-06-24 18:31:38 +0800 -
3710a306Merge pull request #2 from fastnlp/dev0.5.0 by 2019-06-24 16:54:59 +0800 -
39dd08621.修改CrossEntropyLoss中存在的反直觉bug; 2.更新sequence labeling by 2019-06-24 09:56:28 +0800 -
add8edf7Merge branch 'dev0.5.0' of github.com:fastnlp/fastNLP into dev0.5.0 by 2019-06-23 19:14:05 +0800 -
e9137349解决在DataParallel模型场景下无法进行参数匹配的问题 by 2019-06-23 19:13:49 +0800 -
4d9eb7c6update framework of matching by 2019-06-23 18:44:36 +0800 -
3593f0a5fix bugs in matching dataloader by 2019-06-23 18:42:57 +0800 -
d1f531c0update matching dataloader in reproduction/matching by 2019-06-23 18:25:04 +0800 -
5d087758conflict merge by 2019-06-23 15:04:33 +0800 -
4b0c26d31.修复BertEmbedding中的bug; 2. 修复Batch, Field在进行类型转换时的bug by 2019-06-23 15:02:34 +0800 -
0a2122a7Merge branch 'dev0.5.0' of https://github.com/fastnlp/fastNLP into dev0.5.0 by 2019-06-22 22:54:17 +0800 -
d34c739aUpdate embedding.py by 2019-06-22 22:53:52 +0800 -
e57b8e4fseq_len_to_mask修复测试失败的问题 by 2019-06-21 11:24:42 +0800 -
8f7ed0741. 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 by 2019-06-21 11:06:35 +0800 -
0f4cf303LSTM修改错误 by 2019-06-19 23:59:40 +0800 -
6b9bc007LSTM中修复潜在的DataParallel可能存在的问题, 并且删除init_method参数 by 2019-06-19 23:49:01 +0800 -
8a766f07seq_len_to_mask修改为直接使用max_len而不再和句中最大长度对比 by 2019-06-19 23:20:12 +0800 -
76e2330a增加seq_len_to_mask对多卡场景的支持 by 2019-06-19 23:13:53 +0800 -
1167d3b5再次修改elmo的dataparallel问题 by 2019-06-19 22:35:44 +0800 -
c4e131a0重新修改ELMO与LSTM DataParallel的问题 by 2019-06-19 22:19:41 +0800 -
a137038e修复ELMO与LSTM无法使用nn.DataParallel的问题 by 2019-06-19 19:43:53 +0800 -
15c7c073fix embed_loader by 2019-06-19 17:04:25 +0800 -
4533427esequence labeling更新 by 2019-06-19 11:14:41 +0800 -
4d138ed7Merge branch 'dev0.5.0' of github.com:fastnlp/fastNLP into dev0.5.0 by 2019-06-18 10:02:29 +0800 -
9a8fe42c新增NER的数据加载与模型代码; 修改metric中的typo; 修改LSTM中的默认初始化将forget gate设置为1. by 2019-06-18 10:02:24 +0800 -
93620e76update framework of matching by 2019-06-18 02:04:53 +0800 -
342b7026Merge remote-tracking branch 'origin/dev0.5.0' into dev0.5.0 by 2019-06-17 21:48:42 +0800 -
39388567update matching.py by 2019-06-17 21:48:18 +0800 -
2f5d89671. 适配将Batch修改为pytorch的DataLoader的修改 2. 修改embedding.py中的bug 3. ConllReader默认跳过所有的DOCSTART标签 4. 交换bert的heavy lifting到_bert, 将BertEncoder在bert.py中暴露 5. crf中allow_transition的include_end_start修改为false，以与CRF的默认值适配 6. allow_transition与SpanMetric支持BIOES类型的tag 7. datainfo中增加打印格式化输出 by 2019-06-17 20:18:07 +0800 -
839d7124增强field中的value_count支持对nested的field的支持 by 2019-06-17 16:46:39 +0800 -
66f51395Merge pull request #166 from fastnlp/batch by 2019-06-17 16:42:28 +0800 -
95e8e2afMerge branch 'dev0.5.0' of github.com:fastnlp/fastNLP into dev0.5.0 by 2019-06-17 13:27:59 +0800 -
30b012ac修复metric和loss在映射时出现重复同名输入时会覆盖的bug by 2019-06-17 13:27:52 +0800 -
0e6a2a74Merge pull request #168 from lyhuang18/lyhuang-reproduction by 2019-06-16 23:51:51 +0800 -
c78811f8add TC/MTL16Loader by 2019-06-16 23:43:37 +0800 -
97b5909fMerge pull request #3 from fastnlp/dev0.5.0 by 2019-06-16 21:22:11 +0800 -
4b5113cbprefecth变更为deprecated warning; by 2019-06-15 14:17:48 +0800 -
17b5fd001. 删除Trainer中对train_data必须为DataSet的assert 2. 删除Trainer的prefetch参数; 在注释中增加num_workers参数 3. Trainer中默认sampler为RandomSampler by 2019-06-15 13:10:28 +0800 -
efe35740Merge remote-tracking branch 'origin/dev0.5.0' into batch by 2019-06-15 12:23:59 +0800 -
7564818f[unstable] change Batch to torch's DataLoader by 2019-06-15 12:12:57 +0800 -
fabc692bMerge branch 'dev0.5.0' of github.com:fastnlp/fastNLP into dev0.5.0 by 2019-06-13 21:26:24 +0800 -
acf18e2e修改DataSet split的一个注释错误 by 2019-06-13 21:25:47 +0800 -
1a4c3c2dfix some bugs in test by 2019-06-13 20:33:15 +0800 -
e5e4381eadd fasttext embedding by 2019-06-13 20:05:59 +0800 -
ba9c591dMerge remote-tracking branch 'origin/dev0.5.0' into batch by 2019-06-13 13:38:10 +0800 -
4124b385Update quickstart.rst by 2019-06-13 11:17:02 +0800 -
e4b07824增加requests依赖 by 2019-06-13 11:09:26 +0800 -
a366f156update documents in embedding.py by 2019-06-13 00:36:13 +0800 -
0f7c732fupdate embedding.py by 2019-06-12 19:08:36 +0800 -
7b8b5884修复_ignore_type的bug. by 2019-06-12 14:09:04 +0800 -
ed3098e11. 修改bert，elmo的cache方式; 这样不需要使用sentence_index这种方式进行索引 by 2019-06-12 12:21:13 +0800 -
6309eafd1. 在fieldarray中支持split，int等handy的function 2. 重大更新，支持ElmoEmbedding, BertEmbedding by 2019-06-12 11:10:33 +0800 -
37c50d66Merge branch 'dev0.5.0' of github.com:fastnlp/fastNLP into dev0.5.0 by 2019-06-11 16:42:11 +0800 -
a504c416Merge pull request #2 from fastnlp/dev0.5.0 by 2019-06-10 20:41:26 +0800 -
bc31b28bMerge pull request #163 from violetyao/dev0.5.0 by 2019-06-09 22:37:35 +0800 -
83729dfcmoved test to reproduction folder by 2019-06-09 21:46:34 +0800 -
6a9ebd5aMerge branch 'dev0.5.0' of https://github.com/violetyao/fastNLP into dev0.5.0 by 2019-06-09 21:31:37 +0800 -
74925f82wrong folder by 2019-06-09 21:30:40 +0800 -
e1869ec0Delete yelp_sample.json by 2019-06-09 21:25:56 +0800 -
189a0a06removed ast by 2019-06-09 16:45:54 +0800 -
d17acd0cremove ast by 2019-06-09 16:45:08 +0800 -
ad6a55bafixed comment format by 2019-06-08 14:32:25 +0800 -
2edb2a1aadded yelpLoader by 2019-06-08 14:27:52 +0800 -
9e5c4f66Merge branch 'dev0.5.0' of github.com:fastnlp/fastNLP into dev0.5.0 by 2019-06-08 09:48:11 +0800 -
bddce51bmerge update by 2019-06-08 09:47:39 +0800 -
e90bbbb3！！！重要更新，DataSet理论上支持任意类型的数据了，但是因为改动非常大，所以可能会有bug by 2019-06-07 23:28:05 +0800 -
9f5c00b1change Batch to torch.DataLoader by 2019-06-06 01:40:46 +0800 -
420112e0Merge branch 'dev0.5.0' of https://github.com/fastnlp/fastNLP into dev0.5.0 by 2019-06-06 00:28:26 +0800 -
0257dc6d修复Trainer无法捕获Exception的bug by 2019-06-06 00:28:12 +0800 -
8e82c917update bert for nli in reproduction/matching by 2019-06-06 00:09:25 +0800 -
999a1438fix a bug in metrics.py by 2019-06-05 20:13:23 +0800 -
9bfedbdf修复了 metrics 文档中的公式 by 2019-06-05 19:36:40 +0800 -
eca04cd0Merge branch 'dev0.5.0' of https://github.com/fastnlp/fastNLP into dev0.5.0 by 2019-06-05 18:22:19 +0800 -
e40447cc禁止travis测试非 test之外的文件 by 2019-06-05 18:21:35 +0800