Commit Graph

2579 Commits

Author SHA1 Message Date
SSK015 5fa223a306 docs: comment RunGraph() More 2023-09-25 22:49:33 +08:00
SSK015 48ccd45285 docs: comment more functions in CompileGraphs Func 2023-09-02 22:28:58 +08:00
i-robot 1a2235358a
!32626 Refactor KernelMod and its sub class.
Merge pull request !32626 from TronZhang/make_kernel_mod_better
2022-04-09 05:36:57 +00:00
i-robot 10274506ef
!32686 fix the dynamic shape with heterogeneity
Merge pull request !32686 from limingqi107/bug_fix4
2022-04-08 09:17:48 +00:00
tronzhang 27893cde7c change kernelmod interface and gpu/cpu kernel mod 2022-04-08 16:28:02 +08:00
i-robot 6b082e6ebe
!32416 RL add TensorsQueue
Merge pull request !32416 from VectorSL/grads_queue_for_merge
2022-04-08 01:32:28 +00:00
i-robot efd13aa953
!32551 Change the types of input and output of tcp message handler
Merge pull request !32551 from chengang/add_retrieve_for_cgn
2022-04-08 01:25:14 +00:00
limingqi107 90158f8aee fix the dynamic shape with heterogeneity 2022-04-08 09:00:16 +08:00
VectorSL 0f9148ea5b add tensorsqueue gpu kernel 2022-04-07 10:02:54 +08:00
i-robot d3e5b08302
!32550 add precision flag for op which raise or reduce precision
Merge pull request !32550 from liubuyu/code_dex
2022-04-06 07:05:05 +00:00
i-robot 050a0e7052
!32543 [Bugfix][PyNative] Coredump
Merge pull request !32543 from caifubi/master-pynative-async-coredump
2022-04-06 03:22:16 +00:00
Parallels 18eaf2d9dd Add return value for user defined message handler 2022-04-06 10:02:08 +08:00
lby 519e78bd4c change log level 2022-04-06 09:35:21 +08:00
caifubi 65f29c5af2 Add lock for PyNative OpExecutor 2022-04-05 18:29:29 +08:00
i-robot d6edfcabed
!32470 Optimize rpc actor in multiple-worker case.
Merge pull request !32470 from ZPaC/sync-bug-fix-for-new-ps
2022-04-02 06:24:50 +00:00
ZPaC b96da9c3de Optimize rpc actor in multiple-worker case. 2022-04-02 10:28:38 +08:00
i-robot 2cf884634f
!32385 update owners
Merge pull request !32385 from kisnwang/modify-owners
2022-04-02 02:11:53 +00:00
i-robot b0043637d9
!30709 Delete DynamicKernel
Merge pull request !30709 from TuDouNi/gpu_cpu_delete_dynamic_kernel
2022-04-02 01:33:13 +00:00
i-robot 18d3e7c467
!32413 Add return value for tcp message handler
Merge pull request !32413 from chengang/add_retrieve_api
2022-04-01 09:33:53 +00:00
kswang 296fb69b54 update owners 2022-04-01 15:25:54 +08:00
i-robot e239e26598
!32298 move the graph_kernel_flags.cc/h to graph_kernel's directory
Merge pull request !32298 from DeshiChen/0329_flags
2022-04-01 06:06:56 +00:00
Parallels 7fa00586af Add return message for message handler 2022-04-01 10:48:12 +08:00
ttudu 4fe7b0d350 delete dynamic_kernel 2022-03-31 20:50:52 +08:00
i-robot aa6981ffa6
!32291 [master]Bugfix for graph with tuple inputs
Merge pull request !32291 from caifubi/master-graph-tuple-input
2022-03-31 08:17:08 +00:00
i-robot fcb0d72e38
!31737 adapt csrtensor runtime
Merge pull request !31737 from wangrao124/pr_csr_param
2022-03-31 03:30:20 +00:00
dayschan 6dd20326b7 move the graph_kernel_flags.cc/h to graph_kernel's directory 2022-03-31 11:23:16 +08:00
i-robot f3bca456bf
!31684 unified runtime support the execution order running
Merge pull request !31684 from limingqi107/new_actor_runtime
2022-03-31 02:02:06 +00:00
caifubi bffe839332 Bugfix for graph with tuple inputs 2022-03-31 09:18:37 +08:00
i-robot 92e70a7e3c
!32212 【bugfix】Use group rank id replace global rank id to check unique id valid
Merge pull request !32212 from zyli2020/worker_failover_bp
2022-03-30 09:20:25 +00:00
limingqi107 d3e6080a9b unified runtime support the execution order running 2022-03-30 17:10:41 +08:00
i-robot 95f02f851e
!32253 Eliminate duplicate control arrows for stack actors.
Merge pull request !32253 from gaoyong10/dynamic_shape_03
2022-03-30 07:14:55 +00:00
gaoyong10 2c79588336 Eliminate duplicate control arrows for stack actors. 2022-03-30 11:01:57 +08:00
ZPaC 35e4e9aed6 Encapsulate methods and optimize logs. 2022-03-30 09:34:31 +08:00
i-robot d22da105c8
!31752 unified runtime enable the dynamic shape in the heterogeneous
Merge pull request !31752 from limingqi107/bug_fix3
2022-03-29 12:49:00 +00:00
lizhenyu de60949038 [bugfix] use group rank id replace global rank id for broadcast unique id 2022-03-29 19:39:32 +08:00
limingqi107 f50834e8e0 unified runtime enable the dynamic shape in the heterogeneous 2022-03-29 12:50:33 +08:00
i-robot a5696e2e3b
!31994 code clean
Merge pull request !31994 from liubuyu/code_dex
2022-03-29 01:46:55 +00:00
wangrao124 ef214dbf4f fix csrtensor parameter 2022-03-28 16:42:24 +08:00
i-robot 42d6188254
!31912 update the output size after updateop
Merge pull request !31912 from TuDouNi/dynamic_shape_stage2
2022-03-26 10:53:51 +00:00
i-robot e271e3a1f4
!31947 bugfix for recovery: add actor set name for finish transform status
Merge pull request !31947 from zyli2020/worker_failover_bp
2022-03-26 10:22:10 +00:00
lby b8351d65c0 code clean 2022-03-26 16:57:25 +08:00
ttudu 138560daac update the output size after updateop 2022-03-26 15:53:55 +08:00
lizhenyu 405465242f bugfix:add actor set name for finish transform status 2022-03-26 14:59:42 +08:00
i-robot d40dc4f997
!31482 Fix load tensor into mem twice for ascend kernel by kernel dump
Merge pull request !31482 from TinaMengtingZhang/kernel_dump
2022-03-26 01:53:33 +00:00
i-robot 0e591fa834
!31937 Fix eval after train issue.
Merge pull request !31937 from ZPaC/add-dist-execution-mode
2022-03-26 01:40:36 +00:00
TinaMengtingZhang 9e717bd7fa fix load and convert tensor twice in ascend kernel by kernel dump 2022-03-25 11:35:11 -04:00
i-robot 54b12d016e
!31929 insert transdata for pynative output
Merge pull request !31929 from yuchaojie/ir_fusion
2022-03-25 11:21:54 +00:00
ZPaC 1d6a72b435 Fix eval after train issue. 2022-03-25 18:13:38 +08:00
i-robot ea4f720184
!31830 Fix bert boost hccl memory error.
Merge pull request !31830 from linqingke/bert_boost_memory
2022-03-25 09:53:40 +00:00
yuchaojie aae5cb4e31 insert transdata for pynative output 2022-03-25 17:46:47 +08:00