Commit Graph

269 Commits (e8cdb49aa9c29390d036d0a9984b4b458a506908)

Author SHA1 Message Date
Leo Chen 0f1fde5102
fix the modification of set_expected_place (#31177)
4 years ago
Qi Li 1d996637e6
[ROCM] update fluid imperative for rocm (part1), test=develop (#31017)
4 years ago
ShenLiang 9401173e3a
Remove scale loss before reduce in dygraph (#30807)
4 years ago
ShenLiang dae3e1f337
Solve inconsistent order in each card in dynamic graph (#30931)
4 years ago
liuyuhui 87197f8c2e
[kunlun]fix sync in multi kunlun xpu dygraph training. (#30943)
4 years ago
WangXi 6e3856d3fb
fix xpu dygraph place (#30868)
4 years ago
wanghuancoder 35c5b23f68
use iwyu clean include second time, test=develop (#30829)
4 years ago
WangXi b1026f64af
【kunlun】dygraph supports multi xpu card training (#30671)
4 years ago
ShenLiang 3858f458ea
rm Singleton of reducer (#30775)
4 years ago
wanghuancoder d1b25ed9d7
add some RecordEvent, for dygraph timeline (#30299)
4 years ago
WangXi 572c466d19
[Prepare for MultiProcess xpu] unified gen nccl id, refine imperative reducer (#30455)
4 years ago
Zhou Wei fb20ec9a4e
fix bug of multicard grad ncclAllReduce (#30553)
4 years ago
pangyoki 00554b3f6b
fix error message of Inplace strategy (#30520)
4 years ago
Leo Chen 7043b8cfc6
support layer_norm fp16 in dygraph amp (#30430)
4 years ago
pangyoki 13d757362c
Add Inplace strategy (Output reuse Input Varbase) in dygraph (#30103)
4 years ago
ShenLiang a60f17b89d
Support unused parameters in dynamic graph distributed (#30224)
4 years ago
石晓伟 8ce2482b80
fix header file paths of gflags, commit 1, test=develop (#30271)
4 years ago
Leo Chen 8696335f86
Fix dtype of ungenerated grad var (#28511)
4 years ago
Leo Chen 1f97d61c68
Add callback after TensorCopy (#30123)
4 years ago
Chen Weihang d0fb06b27f
[Complex] Simplify prepared op impl to improve performance (#30153)
4 years ago
hong 297fff1a79
support dygraph in xpu place (#30051)
4 years ago
Chen Weihang a1d9a14e89
support grad accumulated across batch (#29942)
5 years ago
Chen Weihang a6072055be
[Complex] Handle complex to real after type promotion (#29855)
5 years ago
Chen Weihang 1a304e6c06
[Complex] Add support for complex grad accumulated (#29889)
5 years ago
ShenLiang f65f1caad3
opt sparse allreduce using ncclgather (#29819)
5 years ago
ShenLiang 01e2874a0e
Support multi-stream communication for dynamic graph distributed (#29525)
5 years ago
Zhou Wei e74e1a226c
support deepcopy for Layer/Tensor/Paramerbase (#29387)
5 years ago
ShenLiang 2ef9e0e23c
Rebuild group automatically in dynamic graph distributed (#29255)
5 years ago
Zhou Wei 24ba9ed436
fix that parameters'grad has grad var (#29408)
5 years ago
Leo Chen b58cfff89d
use has_grad instead of train_mode (#29309)
5 years ago
ShenLiang 696dc4bb13
fix the warning of reducer (#29323)
5 years ago
Zhou Wei c0a991c874
accumulate gradient for leaf tensor with previous graph and expose leaf tensor concept (#28429)
5 years ago
liym27 865a45984f
Check whether there is any inplace operation affecting gradient calculation. (#27901)
5 years ago
ShenLiang e2d01eb650
Support dynamic graph distributed (#28997)
5 years ago
Leo Chen 770395cb93
Split train_mode and has_grad for tracer (#29064)
5 years ago
Chen Weihang 7eeb99fe02
Add basic hook classes for dygraph & implement reduce hook (#28584)
5 years ago
danleifeng a24d186814
fix nccl init failed in parallel dygraph mode (#28497)
5 years ago
Chen Weihang 155b4f9b6c
Remove selected rows all reduce over height check (#28460)
5 years ago
Chen Weihang c42e656179
Add retry for dygraph parallel socket bind (#28404)
5 years ago
Leo Chen 44a476c2ab
support cuda pinned place (#28416)
5 years ago
lidanqing 4ea2330759
use FLAGS_use_mkldnn to prevent unnecessary attrs copy (#28146)
5 years ago
danleifeng f29fb396df
dygraph nccl init support host domain name (#28107)
5 years ago
Leo Chen 049696bf67
Refine the format of printing tensor (#27673)
5 years ago
arlesniak 0ecf441af1
Add support for mkldnn ops types selection with FLAGS in dygraph (#27482)
5 years ago
Leo Chen a5b3263782
Refine error msg in paddle/fluid/imperative (#27521)
5 years ago
wanghuancoder df43905f12
use iwyu clean include (#27267)
5 years ago
arlesniak 885c61f086
Add use of global flag 'use_mkldnn' to layer_helper (#26497)
5 years ago
Zhen Wang f32ae272ec
Remove `sorted_sum_gradient_` form BasicEngine and PartialGradTask. (#26766)
5 years ago
Zhen Wang f9066e6a6f
Update the demo code and the doc of varbase.backward. (#26506)
5 years ago
QingshuChen 138ecf24aa
support Baidu Kunlun AI Accelerator (#25959)
5 years ago