x54-729
|
e734d11d3b
|
transformers 添加 AutoTokenizer
|
2022-05-25 15:30:14 +00:00 |
|
Letian Li
|
005b0e055e
|
实现 jittor driver 多卡训练 (#418)
|
2022-05-25 19:09:09 +08:00 |
|
x54-729
|
296c1acc31
|
修复mixdataloader在torch1.6下参数不匹配的问题
|
2022-05-25 08:21:33 +00:00 |
|
Letian Li
|
5425095cac
|
[bugfix] 修复 tests/core/controllers/_test_trainer_jittor.py,使其可以正常运行 (#415)
* 修复 tests/core/controllers/_test_trainer_jittor.py,使其可以正常运行
Trainer 并不接收 validate_dataloaders 参数,改为 evaluate_dataloaders 即可。
* jittor single driver 支持 cpu 和 gpu 的切换
|
2022-05-25 14:07:37 +08:00 |
|
x54-729
|
da0b747b30
|
修改test_mixdataloader中的工具函数名,防止被pytest执行
|
2022-05-23 14:38:49 +00:00 |
|
x54-729
|
756bd09d1a
|
1.删除不必要的测试标签 2.为test_mixdataloader添加torch标签
|
2022-05-23 13:17:45 +00:00 |
|
x54-729
|
f333090149
|
1.删掉不必要的temp标签 2.为test_mixdataloader添加torch标签
|
2022-05-23 13:06:35 +00:00 |
|
x54-729
|
b43e21960d
|
fastNLP/core/dataset/dataset.py
|
2022-05-23 11:46:21 +00:00 |
|
yh
|
388baf589c
|
Merge branch 'dev0.8.0' of github.com:fastnlp/fastNLP into dev0.8.0
|
2022-05-23 19:43:35 +08:00 |
|
yh
|
ff50437702
|
1.修复数据处理的时候,多进程print会报错的问题;2.Trainer/Evaluator增加check_dataloader_legality
|
2022-05-23 19:43:29 +08:00 |
|
x54-729
|
dd0e02b244
|
1.tokenization_utils_base.py中一处logger.warn改为warning_once 2. mix_modules/utils.py torch转换时添加.cpu() 3. dataset中将输出重定向到DEVNULL而不是Null
|
2022-05-22 16:07:58 +00:00 |
|
MorningForest
|
a9316b0209
|
dataset增加set_pad, set_ignore
|
2022-05-22 21:21:36 +08:00 |
|
MorningForest
|
b9be645ed0
|
dataset 增加 set_pad, set_ignore
|
2022-05-22 21:12:42 +08:00 |
|
x54-729
|
cfe389631a
|
Merge branch 'dev0.8.0' of github.com:fastnlp/fastNLP into dev0.8.0
|
2022-05-22 12:57:55 +00:00 |
|
x54-729
|
639de99563
|
为transformers 添加AutoModel
|
2022-05-22 12:57:38 +00:00 |
|
MorningForest
|
a788fd17be
|
增加metric文档
|
2022-05-22 19:50:10 +08:00 |
|
MorningForest
|
e0f49d887b
|
Merge branch 'dev0.8.0' of github.com:fastnlp/fastNLP into dev0.8.0
|
2022-05-22 19:12:55 +08:00 |
|
MorningForest
|
724b112a34
|
增加dataset, collator文档
|
2022-05-22 19:12:47 +08:00 |
|
x54-729
|
a03558a5f8
|
Merge branch 'dev0.8.0' of github.com:fastnlp/fastNLP into dev0.8.0
|
2022-05-22 08:42:48 +00:00 |
|
yh
|
099b5b598e
|
Merge branch 'dev0.8.0' of github.com:fastnlp/fastNLP into dev0.8.0
|
2022-05-22 16:29:52 +08:00 |
|
yh
|
653b129e60
|
Evaluator新增load_model函数
|
2022-05-22 16:29:45 +08:00 |
|
x54-729
|
70b0aaa78f
|
调整文档侧边栏的显示
|
2022-05-21 07:51:11 +00:00 |
|
x54-729
|
1a1b1e8c9e
|
迁移transformers的modeling_auto
|
2022-05-21 07:50:53 +00:00 |
|
MorningForest
|
7ca1abfba5
|
Merge branch 'dev0.8.0' of github.com:fastnlp/fastNLP into dev0.8.0
|
2022-05-20 18:04:09 +08:00 |
|
MorningForest
|
9bb1ed4ccf
|
增加 mixdataloader 文档, 修改mix_sampler, mixdataloader代码, 增加相应测试用例
|
2022-05-20 18:04:00 +08:00 |
|
x54-729
|
1ccb200c5c
|
1.为Jenkins添加30min的超时时间 2.跟进driver关于checkpoint的改动
|
2022-05-20 08:02:25 +00:00 |
|
x54-729
|
f8015d050b
|
Merge branch 'dev0.8.0' of github.com:fastnlp/fastNLP into dev0.8.0
|
2022-05-20 07:28:00 +00:00 |
|
x54-729
|
06551964a3
|
paddle 不再为非fp16的driver加载fp16状态
|
2022-05-20 07:27:48 +00:00 |
|
yh
|
466e924d2e
|
1.修复classify_f1_pre_rec_metric在batch中target不包含某类class时的计算错误;2.增加对FairScale的支持, 同时修改TorchDriver以实现兼容
|
2022-05-20 01:57:06 +08:00 |
|
x54-729
|
5455663093
|
1.修改.Jenkinsfile移动html页面的步骤至post中 2.修改torch_dataloader/test_fdl.py的部分测试例 3.在torch版本较低时跳过worker_inif_function的测试
|
2022-05-19 13:34:46 +00:00 |
|
x54-729
|
cb8eb8040a
|
集成测试时添加对飞书bot的支持和生成结果的html页面
|
2022-05-19 12:00:31 +00:00 |
|
x54-729
|
0838529873
|
Merge branch 'dev0.8.0' of github.com:fastnlp/fastNLP into dev0.8.0
|
2022-05-19 10:15:32 +00:00 |
|
yh
|
c04fc2dc51
|
Merge branch 'dev0.8.0' of github.com:fastnlp/fastNLP into dev0.8.0
|
2022-05-19 17:07:51 +08:00 |
|
yh
|
4d3a93964f
|
修复ClassifyFPreRecMetric代码中的bug
|
2022-05-19 17:07:39 +08:00 |
|
x54-729
|
be94e5b686
|
1.增加构建多分支文档的命令和文件 2.重命名Jenkinsfile为.Jenkinsfile
|
2022-05-19 09:04:58 +00:00 |
|
MorningForest
|
916c113322
|
修改增加 jittor fdl 的文档
|
2022-05-19 01:47:53 +08:00 |
|
MorningForest
|
744f6de6a0
|
修改 fdl 文档
|
2022-05-19 01:37:55 +08:00 |
|
MorningForest
|
6592794d1e
|
Merge branch 'dev0.8.0' of github.com:fastnlp/fastNLP into dev0.8.0
|
2022-05-19 01:08:37 +08:00 |
|
MorningForest
|
10aed5a3c0
|
torch fdl的文档修改
|
2022-05-19 01:08:29 +08:00 |
|
x54-729
|
1096fd8b7b
|
Merge branch 'dev0.8.0' of github.com:fastnlp/fastNLP into dev0.8.0
|
2022-05-18 14:08:16 +00:00 |
|
MorningForest
|
d2672c62b1
|
修改fdl train_sampler
|
2022-05-18 21:54:12 +08:00 |
|
yh
|
696f6b89f8
|
修改参数train_batch_size
|
2022-05-18 21:49:47 +08:00 |
|
MorningForest
|
b76c6816be
|
增加fdl文档
|
2022-05-18 20:52:03 +08:00 |
|
x54-729
|
f183c59cce
|
修改torch_dataloader/test_fdl.py的test_Version_111测试例的参数;删除一处不必要的temp标签
|
2022-05-18 12:33:57 +00:00 |
|
x54-729
|
e21589a0d9
|
暂时跳过jittor的测试;增加torch1.11的共享内存
|
2022-05-18 10:03:18 +00:00 |
|
x54-729
|
3143cc76cc
|
修改dataloaders/test_utils.py
|
2022-05-18 09:33:07 +00:00 |
|
yh
|
0a5a4fde51
|
让DataLoader兼容1.6版本
|
2022-05-18 17:13:45 +08:00 |
|
x54-729
|
4387172bf6
|
Merge branch 'dev0.8.0' of github.com:fastnlp/fastNLP into dev0.8.0
|
2022-05-18 08:33:59 +00:00 |
|
x54-729
|
af36813296
|
1. test_fleet.py 和 test_ddp.py 中的 load_checkpoint 相关的测试例添加同步 2.修改 Jenkinsfile 中 Test Other 的 pytest 标签形式
|
2022-05-18 08:33:43 +00:00 |
|
x54-729
|
daac74f25c
|
1.为torch1.11的测试增加共享内存为256M 2.prepare_dataloader判断backend时添加dist-packages路径
|
2022-05-18 07:58:42 +00:00 |
|