0d832a2539
Enable profiler, for both static and dynamic training. ( #729 )
...
* Enable profiler for static training.
* Polish the initialize of ProfilerOptions.
* Enable profiler for dynamic mode.
2021-05-21 10:31:43 +08:00
c3d401b7ea
add multi xpu support for PaddleClas ( #678 )
2021-04-14 22:31:36 +08:00
4ba3d47e31
Merge branch 'develop' into cp_fp16_training
2021-03-01 16:18:26 +08:00
aeccae2128
fix oom for batch_size=208 ( #618 )
2021-03-01 12:37:17 +08:00
4e43ec6995
new usage of amp training. ( #564 )
...
* new usage of amp training.
* change the usage of amp and pure fp16 training.
* modified code as reviews
2021-02-26 09:25:54 +00:00
918e68a934
fix bug for kunlun ( #518 )
2020-12-28 13:43:05 +08:00
8fd56a4503
fix static train ( #478 )
2020-12-15 16:43:15 +08:00
dc3020ab4a
support fp16 training ( #435 )
...
* support fp16 training
* Use compiled training program
* Change timing ips.
* Use dali
* add pure fp16 training
* fix a bug, which will not use fuse pass using pure fp16 training.
* modify code as review
* modify loss, so that it will use different loss when using pure fp16 training.
* remove some fluid API
* add static optimizer.
2020-12-11 11:04:51 +08:00
066d53f8ec
support cpu/xpu/gpu in static graph ( #460 )
2020-12-08 20:59:23 +08:00
e92cb0b93c
fix init model in static mode ( #444 )
2020-12-03 12:50:33 +08:00
2b77c71459
Support DALI ( #442 )
2020-12-02 22:06:23 +08:00
832364e191
support static graph train for kunlun ( #441 )
2020-12-02 18:36:51 +08:00
e83e3038e1
fix local rank get word size in dist ( #402 )
...
* fix local rank
* fix export model
2020-11-18 13:59:34 +08:00
6a5f4626d7
add static running in dygraph ( #399 )
...
* add static running in dygraph
2020-11-18 09:48:56 +08:00
Yiqun Liu
liuyuhui
littletomatodonkey
Zhang Ting
huangxu96
QingshuChen
QingshuChen
Tingquan Gao