c5ea728016
[pipeline] add bert_for_pretraining bert_lmhead forward and policy ( #4172 )
...
* add pipeline policy and bert forward to be done
* add bertmodel pipeline forward and make tests
* add Bert_Policy and test for policy
* update formatting
* update formatting
* update the code
* fix bugs
* fix name confilt
* add bloom model and policy ,revise the base class of policy
* revise
* revision
* add bert_for_pretraining
* add bert_for_pretraining forward and policy
* fix typos
* cancel warning
* change the imediate output to default dict
* change the default output of get_shared_params
2023-08-15 23:25:14 +08:00
d35bd7d0e6
[shardformer] fix type hint
2023-08-15 23:25:14 +08:00
1ed3f8a24f
[shardformer] rename policy file name
2023-08-15 23:25:14 +08:00
b0b8ad2823
[pipeline] update shardformer docstring
2023-08-15 23:25:14 +08:00
59f6f573f1
[pipeline] update shardformer policy
2023-08-15 23:25:14 +08:00
90a65ea682
[pipeline] build bloom model and policy , revise the base class of policy ( #4161 )
...
* add pipeline policy and bert forward to be done
* add bertmodel pipeline forward and make tests
* add Bert_Policy and test for policy
* update formatting
* update formatting
* update the code
* fix bugs
* fix name confilt
* add bloom model and policy ,revise the base class of policy
* revise
* revision
* add bert_for_pretraining
2023-08-15 23:25:14 +08:00
e8e7e49243
[pipeline]add pipeline policy and bert forward ( #4130 )
...
* add pipeline policy and bert forward to be done
* add bertmodel pipeline forward and make tests
* add Bert_Policy and test for policy
* update formatting
* update formatting
* update the code
* fix bugs
* fix name confilt
2023-08-15 23:25:14 +08:00
f51ce1bc8e
[pipeline] refactor 1f1b schedule ( #4115 )
...
* [api] update optimizer wrapper to fit pipeline
* [pipeline] add base schedule
* [pipeline] add 1f1b schedule
* [test] add pipeline schedule utils test
* [pipeline] fix import
2023-08-15 23:25:14 +08:00
45fdc9b42c
[pipeline] implement p2p communication ( #4100 )
...
* [pipeline] add p2p communication
* [test] add p2p communication test
* [test] add rerun decorator
* [test] rename to avoid conflict
2023-08-15 23:25:14 +08:00
422544222f
[pipeline] add stage manager ( #4093 )
...
* [pipeline] add stage manager
* [test] add pipeline stage manager test
* [pipeline] add docstring for stage manager
2023-08-15 23:25:14 +08:00
5e1a9d48dd
[cluster] add process group mesh ( #4039 )
...
* [cluster] add process group mesh
* [test] add process group mesh test
* force sync
2023-08-15 23:25:14 +08:00
d86ddd9b29
[hotfix] fix unsafe async comm in zero ( #4404 )
...
* improve stablility of zero
* fix wrong index
* add record stream
2023-08-11 15:09:24 +08:00
6ccecc0c69
[gemini] fix tensor storage cleaning in state dict collection ( #4396 )
2023-08-10 15:36:46 +08:00
089c365fa0
[doc] add Series A Funding and NeurIPS news ( #4377 )
...
* [doc] add Series A Funding and NeurIPS news
* [kernal] fix mha kernal
* [CI] skip moe
* [CI] fix requirements
2023-08-04 17:42:07 +08:00
38b792aab2
[coloattention] fix import error ( #4380 )
...
fixed an import error
2023-08-04 16:28:41 +08:00
25c57b9fb4
[fix] coloattention support flash attention 2 ( #4347 )
...
Improved ColoAttention interface to support flash attention 2. Solved #4322
2023-08-04 13:46:22 +08:00
16bf4c0221
[test] remove useless tests ( #4359 )
...
* [test] remove legacy zero test
* [test] remove lazy distribute test
* [test] remove outdated checkpoint io
2023-08-01 18:52:14 +08:00
03654c0ce2
fix localhost measurement ( #4320 )
2023-08-01 10:14:00 +08:00
45b08f08cb
[zero] optimize the optimizer step time ( #4221 )
...
* optimize the optimizer step time
* fix corner case
* polish
* replace all-reduce with all-gather
* set comm device to cuda
2023-07-31 22:13:29 +08:00
1a49a5ea00
[zero] support shard optimizer state dict of zero ( #4194 )
...
* support shard optimizer of zero
* polish code
* support sync grad manually
2023-07-31 22:13:29 +08:00
dd7cc58299
[zero] add state dict for low level zero ( #4179 )
...
* add state dict for zero
* fix unit test
* polish
2023-07-31 22:13:29 +08:00
c668801d36
[zero] allow passing process group to zero12 ( #4153 )
...
* allow passing process group to zero12
* union tp-zero and normal-zero
* polish code
2023-07-31 22:13:29 +08:00
79cf1b5f33
[zero]support no_sync method for zero1 plugin ( #4138 )
...
* support no sync for zero1 plugin
* polish
* polish
2023-07-31 22:13:29 +08:00
c6ab96983a
[zero] refactor low level zero for shard evenly ( #4030 )
...
* refactor low level zero
* fix zero2 and support cpu offload
* avg gradient and modify unit test
* refactor grad store, support layer drop
* refactor bucket store, support grad accumulation
* fix and update unit test of zero and ddp
* compatible with tp, ga and unit test
* fix memory leak and polish
* add zero layer drop unittest
* polish code
* fix import err in unit test
* support diffenert comm dtype, modify docstring style
* polish code
* test padding and fix
* fix unit test of low level zero
* fix pad recording in bucket store
* support some models
* polish
2023-07-31 22:13:29 +08:00
a50d39a143
[NFC] fix: format ( #4270 )
...
* [NFC] polish colossalai/fx/profiler/experimental/profiler_module/embedding.py code style
* [NFC] polish colossalai/communication/utils.py code style
---------
Co-authored-by: Minghao Huang <huangminghao@luchentech.com>
2023-07-26 14:12:57 +08:00
fee553288b
[NFC] polish runtime_preparation_pass style ( #4266 )
2023-07-26 14:12:57 +08:00
3883db452c
[NFC] polish unary_elementwise_generator.py code style ( #4267 )
...
Co-authored-by: aye42 <aye42@gatech.edu>
2023-07-26 14:12:57 +08:00
abe4f971e0
[NFC] polish colossalai/booster/plugin/low_level_zero_plugin.py code style ( #4256 )
...
Co-authored-by: supercooledith <893754954@qq.com>
2023-07-26 14:12:57 +08:00
c614a99d28
[NFC] polish colossalai/auto_parallel/offload/amp_optimizer.py code style ( #4255 )
2023-07-26 14:12:57 +08:00
85774f0c1f
[NFC] polish colossalai/cli/benchmark/utils.py code style ( #4254 )
2023-07-26 14:12:57 +08:00
86cf6aed5b
Fix/format ( #4261 )
...
* revise shardformer readme (#4246 )
* [example] add llama pretraining (#4257 )
* [NFC] polish colossalai/communication/p2p.py code style
---------
Co-authored-by: Jianghai <72591262+CjhHa1@users.noreply.github.com>
Co-authored-by: binmakeswell <binmakeswell@gmail.com>
Co-authored-by: Qianran Ma <qianranm@luchentech.com>
2023-07-26 14:12:57 +08:00
b366f1d99f
[NFC] Fix format for mixed precision ( #4253 )
...
* [NFC] polish colossalai/booster/mixed_precision/mixed_precision_base.py code style
2023-07-26 14:12:57 +08:00
c6f6005990
[checkpointio] Sharded Optimizer Checkpoint for Gemini Plugin ( #4302 )
...
* sharded optimizer checkpoint for gemini plugin
* modify test to reduce testing time
* update doc
* fix bug when keep_gatherd is true under GeminiPlugin
2023-07-21 14:39:01 +08:00
fc5cef2c79
[lazy] support init on cuda ( #4269 )
...
* [lazy] support init on cuda
* [test] update lazy init test
* [test] fix transformer version
2023-07-19 16:43:01 +08:00
4b977541a8
[Kernels] added triton-implemented of self attention for colossal-ai ( #4241 )
...
* added softmax kernel
* added qkv_kernel
* added ops
* adding tests
* upload tets
* fix tests
* debugging
* debugging tests
* debugging
* added
* fixed errors
* added softmax kernel
* clean codes
* added tests
* update tests
* update tests
* added attention
* add
* fixed pytest checking
* add cuda check
* fix cuda version
* fix typo
2023-07-18 23:53:38 +08:00
9a4842c571
revise shardformer readme ( #4246 )
2023-07-17 17:30:57 +08:00
58913441a1
Next commit [checkpointio] Unsharded Optimizer Checkpoint for Gemini Plugin ( #4141 )
...
* [checkpointio] unsharded optimizer checkpoint for Gemini plugin
* [checkpointio] unsharded optimizer checkpoint for Gemini using all_gather
2023-07-07 16:33:06 +08:00
190a6ea9c2
[dtensor] fixed readme file name and removed deprecated file ( #4162 )
2023-07-04 18:21:11 +08:00
1908caad38
[cli] hotfix launch command for multi-nodes ( #4165 )
2023-07-04 17:54:40 +08:00
2ac24040eb
fix some typo colossalai/shardformer ( #4160 )
2023-07-04 17:53:39 +08:00
c77b3b19be
[format] applied code formatting on changed files in pull request 4152 ( #4157 )
...
Co-authored-by: github-actions <github-actions@github.com>
2023-07-04 16:07:47 +08:00
89f45eda5a
[shardformer] added development protocol for standardization ( #4149 )
2023-07-04 16:05:01 +08:00
1fb0d95df0
[shardformer] made tensor parallelism configurable ( #4144 )
...
* [shardformer] made tensor parallelism configurable
* polish code
2023-07-04 16:05:01 +08:00
74257cb446
[shardformer] refactored some doc and api ( #4137 )
...
* [shardformer] refactored some doc and api
* polish code
2023-07-04 16:05:01 +08:00
7f9b30335b
[shardformer] write an shardformer example with bert finetuning ( #4126 )
...
* [shardformer] add benchmark of shardformer
* [shardformer] add benchmark of shardformer
2023-07-04 16:05:01 +08:00
ae035d305d
[shardformer] added embedding gradient check ( #4124 )
2023-07-04 16:05:01 +08:00
44a190e6ac
[shardformer] import huggingface implicitly ( #4101 )
2023-07-04 16:05:01 +08:00
6a88bae4ec
[shardformer] integrate with data parallelism ( #4103 )
2023-07-04 16:05:01 +08:00
f3b6aaa6b7
[shardformer] supported fused normalization ( #4112 )
2023-07-04 16:05:01 +08:00
b1c2901530
[shardformer] supported bloom model ( #4098 )
2023-07-04 16:05:01 +08:00
8af29ee47a
[shardformer] support vision transformer ( #4096 )
...
* first v of vit shardformer
* keep vit
* update
* vit shard add vitattention vitlayer
* update num head shard para
* finish test for vit
* add new_model_class & postprocess
* add vit readme
* delete old files & fix the conflict
* fix sth
2023-07-04 16:05:01 +08:00
ac80937138
[shardformer] shardformer support opt models ( #4091 )
...
* [shardformer] shardformer support opt models
* [shardformer] shardformer support opt models, fix
* [shardformer] shardformer support opt models, fix
* [shardformer] shardformer support opt models, fix
2023-07-04 16:05:01 +08:00
d33a44e8c3
[shardformer] refactored layernorm ( #4086 )
2023-07-04 16:05:01 +08:00
c4b1b65931
[test] fixed tests failed due to dtensor change ( #4082 )
...
* [test] fixed tests failed due to dtensor change
* polish code
2023-07-04 16:05:01 +08:00
92f6791095
[shardformer] Add layernorm ( #4072 )
...
* add layernorm to bert
* add layernorm test
* add layernorm test with load state dict
* add use_mixedfusedLN in shard config
* refactor policy to support fused_layernorm
2023-07-04 16:05:01 +08:00
70c58cfd4f
[shardformer] supported fused qkv checkpoint ( #4073 )
2023-07-04 16:05:01 +08:00
0803a61412
[shardformer] add linearconv1d test ( #4067 )
...
* add linearconv1d test
* add linearconv1d test
2023-07-04 16:05:01 +08:00
8eb09a4c69
[shardformer] support module saving and loading ( #4062 )
...
* [shardformer] support module saving and loading
* polish code
2023-07-04 16:05:01 +08:00
7740c55c55
support kit use for bert/gpt test ( #4055 )
...
* support kit use for bert test
* support kit test for gpt2
2023-07-04 16:05:01 +08:00
f22ddacef0
[shardformer] refactored the shardformer layer structure ( #4053 )
2023-07-04 16:05:01 +08:00
58df720570
[shardformer] adapted T5 and LLaMa test to use kit ( #4049 )
...
* [shardformer] adapted T5 and LLaMa test to use kit
* polish code
2023-07-04 16:05:01 +08:00
4021b9a8a2
[shardformer] add gpt2 test and layer class refactor ( #4041 )
...
* add gpt2 test and layer class refactor
* add dropout in gpt2 policy
2023-07-04 16:05:01 +08:00
d857f3dbba
[shardformer] supported T5 and its variants ( #4045 )
2023-07-04 16:05:01 +08:00
c1d5453e9f
[shardformer] adapted llama to the new API ( #4036 )
2023-07-04 16:05:01 +08:00
74d176c8d8
[shardformer] fix bert and gpt downstream with new api ( #4024 )
...
* fix bert downstream with new api
* remove comment line
2023-07-04 16:05:01 +08:00
e253a07007
[shardformer] updated doc ( #4016 )
2023-07-04 16:05:01 +08:00
df018fc305
support bert with new api
2023-07-04 16:05:01 +08:00
507c0ad368
add vocabembedding layer
2023-07-04 16:05:01 +08:00
45d9384346
[shardformer] removed inplace tensor sharding ( #4018 )
2023-07-04 16:05:01 +08:00
3893fa1a8d
[shardformer] refactored embedding and dropout to parallel module ( #4013 )
...
* [shardformer] refactored embedding and dropout to parallel module
* polish code
2023-07-04 16:05:01 +08:00
dfca9678fa
integrate with dist layer ( #4011 )
2023-07-04 16:05:01 +08:00
015af592f8
[shardformer] integrated linear 1D with dtensor ( #3996 )
...
* [shardformer] integrated linear 1D with dtensor
* polish code
2023-07-04 16:05:01 +08:00
d3bc530849
[shardformer] Refactor shardformer api ( #4001 )
...
* fix an error in readme
* simplify code
* refactor shardformer
* add todo
* remove slicer
* resolve code review
2023-07-04 16:05:01 +08:00
611971248c
[device] support init device mesh from process group ( #3990 )
2023-07-04 16:05:01 +08:00
a2f9af810d
[shardformer] fix an error in readme ( #3988 )
...
* fix an error in readme
* simplify code
2023-07-04 16:05:01 +08:00
f7774ec0f3
[Shardformer] Downstream bert ( #3979 )
...
* add dist dropout in model
* update docstring and bert policy with dropout
* refactor basepolicy and sharded, update bert
* update format
* update gpt2 policy
* update bert policy
* remove unused code
* update readme for new policy usage
* add downstream model of bert
* remove unused code
2023-07-04 16:05:01 +08:00
c1c672d0f0
[shardformer] shardformer support t5 model ( #3994 )
...
test t5
2023-07-04 16:05:01 +08:00
6b30dfb7ce
[shardformer] support llama model using shardformer ( #3969 )
...
adjust layer attr
2023-07-04 16:05:01 +08:00
45927d5527
[shardformer] Add dropout layer in shard model and refactor policy api ( #3949 )
...
* add dist dropout in model
* update docstring and bert policy with dropout
* refactor basepolicy and sharded, update bert
* update format
* update gpt2 policy
* update bert policy
* remove unused code
* update readme for new policy usage
2023-07-04 16:05:01 +08:00
a73130482d
[shardformer] Unit test ( #3928 )
...
* fix bug in slicer, add slicer unit test
* add dropout test
* use pid as dropout seed
* updata dropout test with local pattern
* ad todo
2023-07-04 16:05:01 +08:00
f1cb5ac6bf
[shardformer] Align bert value ( #3907 )
...
* add bert align test, fix dist loss bug
* forward and backward align
* add ignore index
* add shardformer CI
* add gather_output optional for user in shardconfig
* update readme with optional gather_ouput
* add dist crossentropy loss test, remove unused files
* remove unused file
* remove unused file
* rename the file
* polish code
2023-07-04 16:05:01 +08:00
79f8d5d54b
[shardformer] add gpt2 policy and modify shard and slicer to support ( #3883 )
...
* add gpt2 policy and modify shard and slicer to support
* remove unused code
* polish code
2023-07-04 16:05:01 +08:00
70173e3123
update README ( #3909 )
2023-07-04 16:05:01 +08:00
ab8a47f830
[shardformer] add Dropout layer support different dropout pattern ( #3856 )
...
* add dropout layer, add dropout test
* modify seed manager as context manager
* add a copy of col_nn.layer
* add dist_crossentropy loss; separate module test
* polish the code
* fix dist crossentropy loss
2023-07-04 16:05:01 +08:00
c594dc2f1c
[shardformer] update readme with modules implement doc ( #3834 )
...
* update readme with modules content
* remove img
2023-07-04 16:05:01 +08:00
4972e1f40e
[shardformer] refactored the user api ( #3828 )
...
* [shardformer] refactored the user api
* polish code
2023-07-04 16:05:01 +08:00
235792f170
[shardformer] updated readme ( #3827 )
2023-07-04 16:05:01 +08:00
8cc11235c0
[shardformer]: Feature/shardformer, add some docstring and readme ( #3816 )
...
* init shardformer code structure
* add implement of sharder (inject and replace)
* add implement of replace layer to colossal layer
* separate different layer policy, add some notion
* implement 1d and 2d slicer, can tell col or row
* fix bug when slicing and inject model
* fix some bug; add inference test example
* add share weight and train example
* add train
* add docstring and readme
* add docstring for other files
* pre-commit
2023-07-04 16:05:01 +08:00
8d68de767d
[shardformer] init shardformer code structure ( #3731 )
...
* init shardformer code structure
* add implement of sharder (inject and replace)
* add implement of replace layer to colossal layer
* separate different layer policy, add some notion
* implement 1d and 2d slicer, can tell col or row
* fix bug when slicing and inject model
* fix some bug; add inference test example
2023-07-04 16:05:01 +08:00
1350ece492
[hotfix] fix import bug in checkpoint_io ( #4142 )
2023-07-03 22:14:37 +08:00
8abc87798f
fix Tensor is not defined ( #4129 )
2023-07-03 17:10:18 +08:00
7e46bc87b6
fix CheckpointIndexFile is not defined ( #4109 )
2023-07-03 17:09:06 +08:00
09fe9dc704
[nfc]fix ColossalaiOptimizer is not defined ( #4122 )
2023-06-30 17:23:22 +08:00
95e95b6d58
[testing] move pytest to be inside the function ( #4087 )
2023-06-27 11:02:25 +08:00
0bb0b481b4
[gemini] fix argument naming during chunk configuration searching
2023-06-25 13:34:15 +08:00
a52f62082d
[format] applied code formatting on changed files in pull request 4021 ( #4022 )
...
Co-authored-by: github-actions <github-actions@github.com>
2023-06-19 11:23:24 +08:00
822c3d4d66
[checkpointio] sharded optimizer checkpoint for DDP plugin ( #4002 )
2023-06-16 14:14:05 +08:00
725af3eeeb
[booster] make optimizer argument optional for boost ( #3993 )
...
* feat: make optimizer optional in Booster.boost
* test: skip unet test if diffusers version > 0.10.2
2023-06-15 17:38:42 +08:00
c9cff7e7fa
[checkpointio] General Checkpointing of Sharded Optimizers ( #3984 )
2023-06-15 15:21:26 +08:00
71fe52769c
[gemini] fixed the gemini checkpoint io ( #3934 )
2023-06-12 15:11:27 +08:00
Jianghai