yh
|
7e41b9d355
|
1.新增CMRC2018的metric,loss,pipe,loader等; 2. 新增中文Bert的使用教程
|
6 years ago |
yunfan
|
c73f327a3e
|
[bugfix] fix test cases
|
6 years ago |
Yige Xu
|
b3718b10dc
|
1. rename init_embed to embed in models/*; 2. update documents in models/bert.py; 3. update tutorial six.
|
6 years ago |
Yige Xu
|
091f24e393
|
fix some bugs in test code.
|
6 years ago |
yh
|
dc2885c6bc
|
修复sequence labeling 测试报错
|
6 years ago |
Yige Xu
|
4440801dbf
|
1. update bert.py and fix a bug in bert_embedding to adapt torch 1.2.0; 2. update models/bert.py and add BertForSentenceMatching model, now a BertEmbedding param should be passed to these five models; 3. create a small bert version for testing and modify test/models/test_bert.py; 4. move small glove and word2vec files to data_for_tests/embedding/small_static_embedding dir and fix relevant test codes; 5. delete some __init__.py files in test dir.
|
6 years ago |
xuyige
|
9560a4d367
|
update test codes in models/bert.py
|
6 years ago |
xuyige
|
c2d687528e
|
fix bugs and add test codes for: 1. models.snli; 2. core.metrics.extractive_qa; 3. io.data_loader.mnli
|
6 years ago |
xuyige
|
327833a77f
|
fix bug in test code
|
6 years ago |
xuyige
|
089009f9f2
|
大幅度更新:
1、更新requirements以及README.md
2、更新DataLoader
3、更新loss
4、更新model/bert.py内容及适配的测试代码
5、更新reproduction/README.md
6、修复其他测试代码的报错的地方
|
6 years ago |
yh_cc
|
2f5d8967a3
|
1. 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
|
6 years ago |
yh_cc
|
17b5fd0066
|
1. 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
|
6 years ago |
xuyige
|
1a4c3c2d20
|
fix some bugs in test
|
6 years ago |
xuyige
|
016f02be3b
|
fix bugs in model/bert.py and add testing codes
|
6 years ago |
ChenXin
|
902a3a6bcd
|
修改了一些注释
|
6 years ago |
yh_cc
|
af6a9da78d
|
Merge branch 'dev' of github.com:choosewhatulike/fastNLP-private into dev
|
6 years ago |
yh_cc
|
6d36dbe7fb
|
完善测试
|
6 years ago |
yunfan
|
702fa1d95c
|
- update attention
- fix tests
|
6 years ago |
yh
|
25565fe0c9
|
增加cnn的测试
|
6 years ago |
yunfan
|
4f65e17d1a
|
- add model runner for easier test models
- add model tests
|
6 years ago |
yunfan
|
799d4dbc68
|
- add test
- fix jsonloader
|
6 years ago |
yunfan
|
001586fa3e
|
- add document
|
6 years ago |
FengZiYjun
|
f5ab7a5d45
|
* 将enas相关代码放到automl目录下
* 修复fast_param_mapping的一个bug
* Trainer添加自动创建save目录
* Vocabulary的打印,显示内容
|
6 years ago |
chenkaiyu1997
|
251088ac4b
|
Add ENAS (Efficient Neural Architecture Search)
|
6 years ago |
FengZiYjun
|
bfaf09df8c
|
add BERT model
* load pre-trained BERT weights from local binary
* add tests
|
6 years ago |
FengZiYjun
|
e4f997d52a
|
refactor type system in FieldArray:
* 重构dtype的检测代码,在FieldArray的初始化和append两处,达到更好的代码复用
* 类型检测的责任完全落在FieldArray,DataSet与之配合
测试:
* 整理dtype相关的测试代码
* 给所有tutorial添加测试
其他:
* 完善一个完整的Conll dataset loader
* 升级POS tag model训练脚本
|
6 years ago |
yunfan
|
2e9e6c6c20
|
- fix trainer with validate_every > 0
- refine & fix Transformer Encoder
- refine & speed up biaffine parser
|
6 years ago |
yunfan
|
ba28702e68
|
update Biaffine Parser, Variational RNN
add parser API
|
6 years ago |
yunfan
|
c91696e1ee
|
update parser, optimize embed_loader
|
6 years ago |