Col_lin
/
mindspore2022
forked from huawei/mindspore2022
0
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
29,112 Commits 18 Branches 22 Tags 3.6 GiB
Commit Graph

38 Commits

Author SHA1 Message Date
Parastoo Ashtari ff83dee704 Add tensor base and stat info to offline debugger 2021-08-24 18:19:59 -04:00
sabrinasun 3150413259 change privare member name 2021-08-09 16:10:49 -04:00
He Wei b95e592324 Implement float16 to remove Eigen dependence 2021-07-28 16:46:46 +08:00
John Tzanakakis 3a6c88e3e6 unify overflow wp support for online and offline debugger 2021-07-27 00:53:53 -04:00
John Tzanakakis a0c6e29312 seg fault and existing dump fix 2021-07-08 16:53:48 -04:00
Parastoo Ashtari 81a289d084 Fix for load async tensor on acend issue 2021-07-05 19:02:04 -04:00
TinaMengtingZhang 9a68c26fc6 Fix issues: 1.missing parameter name in dump; 2.too much read tensor output in offline debugger 2021-06-30 10:29:40 -04:00
John Tzanakakis c85d65c0ef use -1 iteration for init state and fix missing tensor read outputs 2021-06-25 19:47:54 -04:00
TinaMengtingZhang 4e8db5e115 change arg 'is_parameter' to 'is_output' and rename device_id to rank_id in python api 2021-06-23 14:25:53 -04:00
TinaMengtingZhang 36b0517ac3 fix bug: stripped name 2021-06-14 20:59:06 -04:00
John Tzanakakis daedb7dd9f dump and offline fixes for sync op_debug_mode and input reading 2021-06-10 16:44:46 -04:00
John Tzanakakis ac1847ffac fix iter 0 and iter 1 being dumped in dir 0, make op_debug_mode optional for sync mode, read input files for offline debugger 2021-06-09 21:57:07 -04:00
Parastoo Ashtari 4120a122e3 Add multi thread for checkwatchpoints 
fixed the style issues

fixed the style header issue

Add tensor list byte size in debugger.cc and clean code for better style

Removed some comments

fixed the bug for multi thread code offline debugger

set default number of threads for checkwatchpoints to 32. move tensor byte size to debug_services.cc for offline and online debugger
 2021-06-07 11:50:34 -04:00
John Tzanakakis 7d460de8b4 adapt offline debuger to new dump, sync support for iteration ranges, GPU default dump mode correction 2021-06-04 08:38:50 -04:00
mindspore-ci-bot 045afa7686 !16033 [Debugger] Fixing ci alarm on typedef in master 
From: @tina_mengting_zhang
Reviewed-by: @john_tzanakakis,@tom__chen
Signed-off-by: @john_tzanakakis
 2021-05-08 04:49:27 +08:00
TinaMengtingZhang 248be56be7 change C style struct definition 2021-05-06 15:00:35 -04:00
John Tzanakakis b5e70f7231 use device type in dump file name when trans_flag=false 2021-05-06 14:07:20 -04:00
tinazhang 6b84149e2f Address ci alarm for master branch: pclint, CppCodeStyle and codex 2021-05-03 16:46:14 -04:00
Islam Amin 1b5ec2f3e1 NPY support 2021-04-16 11:56:14 -04:00
John Tzanakakis da3b13a0e1 Offline debugger 
Authors: John Tzanakakis, Adel Shafiei, Amir Lashkari, Islam Amin
 2021-04-05 18:01:38 -04:00
John Tzanakakis b291861e85 LoadTensors was using a 4-byte size field instead of an 8-byte size field 
don't load tensors if the type is none
 2020-12-22 11:10:30 -05:00
Harshvardhan Gupta dd0084c52b improve perf, keep consistent tensor state, fix recheck, check weights at step end 2020-12-08 15:22:40 -05:00
l00591931 1d1cab986d Codedex change for tensor_loader 2020-12-07 09:45:27 +08:00
Harshvardhan Gupta 7687a55c0a return error code if prev tensor not found and add validation checks 2020-11-30 14:51:40 -05:00
Harshvardhan Gupta 561f9082e9 init 2020-11-27 17:01:24 -05:00
mindspore-ci-bot f2cfa0f5bc !8238 change abs mean, update_weight_ratio, and weight and bias filter 
Merge pull request !8238 from john_tzanakakis/master_ms1_grpc
 2020-11-05 23:21:26 +08:00
Yi Huaijie d7faa77b5e support int64 shape 2020-11-05 21:15:49 +08:00
John Tzanakakis 8d436d5335 change abs mean, update_weight_ratio, and weight and bias filter 2020-11-04 20:57:18 -05:00
Harshvardhan Gupta 727d424553 support watchpoints on weights and bias, add support advanced watchpoints 2020-11-03 13:23:52 -05:00
Harshvardhan Gupta 744355a005 remove dbg args from runtime and remove needless argument from IsWatchpoint 2020-10-19 11:22:22 -04:00
Harshvardhan Gupta 18d1e8795a extend wp support to all types 2020-09-11 13:11:20 -04:00
Harshvardhan Gupta e751c2e069 add more wp conditions 2020-09-10 19:45:19 -04:00
Harshvardhan Gupta 3226e8407a fix gpu wp 2020-08-27 17:41:50 -04:00
Adel Shafiei 4834a3378b Op Overflow Watchpoint support for D-chip debugger 
Other Authors: Harshvardhan Gupta, Li Chen
 2020-08-17 18:06:47 -04:00
lichen_101010 7499c72d54 mindspore grpc implementation 
fix bugs for grpc implementation

addressed peer review comments

delete device_target code from Adel

add checksinglewatchpoint function for node level debugger

set the device target when sending metadata

add current node name

fix bugs for current node name

fix run_level_ bug

fix bugs for CheckSingleWatchpoint

fix multi-outputs node issue

fix num_step_ bug

fix continue_to previous node issue

fix run_level issue

fix merge conflict

smart kernel read, watch hit stop mid-sep, fix step number, read input tensors

cleanup the code and isolate UpdataStepNum function

do cpplint, Cppcheck and clang-format check

recover CMakeList.txt

mindspore grpc implementation

fix bugs for grpc implementation

addressed peer review comments

delete device_target code from Adel

add checksinglewatchpoint function for node level debugger

set the device target when sending metadata

add current node name

fix bugs for current node name

fix run_level_ bug

fix bugs for CheckSingleWatchpoint

fix multi-outputs node issue

fix num_step_ bug

fix continue_to previous node issue

fix run_level issue

fix merge conflict

smart kernel read, watch hit stop mid-sep, fix step number, read input tensors

cleanup the code and isolate UpdataStepNum function

do cpplint, Cppcheck and clang-format check

recover CMakeList.txt

only update step_num in one place

fix clang-format error

fix CI errors part2

update graphengine version

addressed comments
 2020-08-04 12:51:45 -04:00
John Tzanakakis 3d06b503d6 debugger compile fix 2020-07-27 21:36:09 -04:00
lichen_101010 6bb2182134 Add partial memory reuse support to debugger 
move pre-execution of debugger from rungraph to build/compile graph

support partial mem reuse for a scope of nodes

set default mem reuse to be true for debugger

remove some redundant lines

remove redundant code and fix a bug for supporting partial no mem reuse a scope of nodes

resolve CI errors

Solve CI errors

solve cpplint errors

solve CI build error

manually fix the CI compile UT error

Optimize code for mem reuse support

Debug optimization of debugger memory reuse

debug code for debugger memory reuse part2

address clang-format errors

Switch memory reuse on and off based on environment variable

Fix typo

Fix typo

Load watchpoint value only

fix bugs

Addressed comments from lupengcheng

fix typo

Fix typo

fix CI errors

refactor some code

fix typo

addressed comments from canadian teamates

remove locking from TensorLoader

fix CI errors

add lock to tensor_loader

fix rebase-to-master conflict

fix rebase conflicts

fix rebase conflicts part 2

fix rebase conflicts part 3
 2020-07-14 14:47:25 -04:00
Shida He 4c056855e0 Implementation for mindspore debugger 2020-06-22 17:15:30 -04:00