8593ae1a3f
[autoparallel] rotor solver refactor ( #2813 )
...
* [autoparallel] rotor solver refactor
* [autoparallel] rotor solver refactor
2023-02-18 11:30:15 +08:00
56ddc9ca7a
[hotfix] add correct device for fake_param ( #2796 )
2023-02-17 15:29:07 +08:00
a2b43e393d
[autoparallel] Patch meta information of `torch.nn.Embedding` ( #2760 )
...
* [autoparallel] embedding metainfo
* [autoparallel] fix function name in test_activation_metainfo
* [autoparallel] undo changes in activation metainfo and related tests
2023-02-17 10:39:48 +08:00
8e3f66a0d1
[zero] fix wrong import ( #2777 )
2023-02-17 10:26:07 +08:00
01066152f1
Don't use `torch._six` ( #2775 )
...
* Don't use `torch._six`
This is a private API which is gone after https://github.com/pytorch/pytorch/pull/94709
* Update common.py
2023-02-17 09:22:45 +08:00
93b788b95a
Merge branch 'main' into fix/format
2023-02-15 20:23:51 +08:00
2fd528b9f4
[NFC] polish colossalai/auto_parallel/tensor_shard/deprecated/graph_analysis.py code style ( #2737 )
2023-02-15 22:57:45 +08:00
1dc003c169
[autoparallel] distinguish different parallel strategies ( #2699 )
2023-02-15 22:28:28 +08:00
ae86a29e23
Refact method of grad store ( #2687 )
2023-02-15 22:27:58 +08:00
c9e3ee389e
[NFC] polish colossalai/context/process_group_initializer/initializer_2d.py code style ( #2726 )
2023-02-15 22:27:13 +08:00
1819373e5c
[NFC] polish colossalai/auto_parallel/tensor_shard/deprecated/op_handler/batch_norm_handler.py code style ( #2728 )
2023-02-15 22:26:13 +08:00
8331420520
[NFC] polish colossalai/cli/cli.py code style ( #2734 )
2023-02-15 22:25:28 +08:00
d344313533
[NFC] polish colossalai/auto_parallel/tensor_shard/deprecated/op_handler/embedding_handler.py code style ( #2725 )
2023-02-15 16:31:40 +08:00
e81caeb4bc
[NFC] polish colossalai/auto_parallel/tensor_shard/deprecated/cost_graph.py code style ( #2720 )
...
Co-authored-by: Fuzhao Xue <fuzhao@login2.ls6.tacc.utexas.edu>
2023-02-15 16:12:45 +08:00
51c45c2460
[NFC] polish colossalai/auto_parallel/tensor_shard/deprecated/op_handler/where_handler.py code style ( #2723 )
2023-02-15 16:12:24 +08:00
21d6a48f4d
[autoparallel] add shard option ( #2696 )
...
* [autoparallel] add shard option
* polish
2023-02-15 13:48:28 +08:00
5b24987fa7
[autoparallel] fix parameters sharding bug ( #2716 )
2023-02-15 12:25:50 +08:00
4603538ddd
[NFC] posh colossalai/context/process_group_initializer/initializer_sequence.py code style ( #2712 )
...
Co-authored-by: Ziyue Jiang <ziyue.jiang@gmail.com>
2023-02-15 10:53:38 +08:00
cb2c6a2415
[autoparallel] refactor runtime pass ( #2644 )
...
* [autoparallel] refactor runtime pass
* add unit test
* polish
2023-02-15 10:36:19 +08:00
b3d10db5f1
[NFC] polish colossalai/cli/launcher/__init__.py code style ( #2709 )
2023-02-15 09:57:22 +08:00
0b2a738393
[autoparallel] remove deprecated codes ( #2664 )
2023-02-15 09:54:32 +08:00
7fa6be49d2
[autoparallel] test compatibility for gemini and auto parallel ( #2700 )
2023-02-15 09:43:29 +08:00
4ac8bfb072
[NFC] polish colossalai/engine/gradient_handler/utils.py code style ( #2708 )
2023-02-15 09:40:08 +08:00
6427c406cf
[NFC] polish colossalai/auto_parallel/tensor_shard/deprecated/op_handler/strategy_generator.py code style ( #2695 )
...
Co-authored-by: shenggan <csg19971016@gmail.com>
2023-02-14 21:30:25 +08:00
534f68c83c
[NFC] polish pipeline process group code style ( #2694 )
2023-02-14 18:12:01 +08:00
56ff1921e9
[NFC] polish colossalai/context/moe_context.py code style ( #2693 )
2023-02-14 18:02:45 +08:00
1712da2800
[NFC] polish colossalai/gemini/gemini_context.py code style ( #2690 )
2023-02-14 11:55:23 +08:00
df4f020ee3
[zero1&2] only append parameters with gradients ( #2681 )
2023-02-13 18:00:16 +08:00
f0aa191f51
[gemini] fix colo_init_context ( #2683 )
2023-02-13 17:53:15 +08:00
40c916b192
[autoparallel] Patch meta information of `torch.nn.functional.softmax` and `torch.nn.Softmax` ( #2674 )
...
* [autoparallel] softmax metainfo
* [autoparallel] softmax metainfo
2023-02-13 16:09:22 +08:00
8213f89fd2
[gemini] add fake_release_chunk for keep-gathered chunk in the inference mode ( #2671 )
2023-02-13 14:35:32 +08:00
9ab14b20b5
[doc] add CVPR tutorial ( #2666 )
2023-02-10 20:43:34 +08:00
0385b26ebf
[autoparallel] Patch meta information of `torch.nn.LayerNorm` ( #2647 )
...
* [autoparallel] layernorm metainfo patch
* [autoparallel] polish test
2023-02-10 14:29:24 +08:00
37df666f38
[autoparallel] refactor handlers which reshape input tensors ( #2615 )
...
* [autoparallel] refactor handlers which reshape input tensors
* polish
2023-02-08 15:02:49 +08:00
28398f1c70
add overlap option ( #2613 )
2023-02-08 15:02:31 +08:00
cb3d1bef62
[autoparallel] adapt autoparallel tests with latest api ( #2626 )
2023-02-08 15:02:12 +08:00
90a9fdd91d
[autoparallel] Patch meta information of `torch.matmul` ( #2584 )
...
* [autoparallel] matmul metainfo
* [auto_parallel] remove unused print
* [tests] skip test_matmul_handler when torch version is lower than 1.12.0
2023-02-08 11:05:31 +08:00
6ba8364881
[autochunk] support diffusion for autochunk ( #2621 )
...
* add alphafold benchmark
* renae alphafold test
* rename tests
* rename diffuser
* renme
* rename
* update transformer
* update benchmark
* update benchmark
* update bench memory
* update transformer benchmark
* rename
* support diffuser
* support unet metainfo prop
* fix bug and simplify code
* update linear and support some op
* optimize max region search, support conv
* update unet test
* support some op
* support groupnorm and interpolate
* update flow search
* add fix dim in node flow
* fix utils
* rename
* support diffusion
* update diffuser
* update chunk search
* optimize imports
* import
* finish autochunk
2023-02-07 16:32:45 +08:00
8518263b80
[test] fixed the triton version for testing ( #2608 )
2023-02-07 13:49:38 +08:00
552183bb74
[polish] polish ColoTensor and its submodules ( #2537 )
2023-02-03 11:44:10 +08:00
dd14783f75
[kernel] fixed repeated loading of kernels ( #2549 )
...
* [kernel] fixed repeated loading of kernels
* polish code
* polish code
2023-02-03 09:47:13 +08:00
5b1854309a
[hotfix] fix zero ddp warmup check ( #2545 )
2023-02-02 16:42:38 +08:00
fa3d66feb9
support unet metainfo prop ( #2544 )
2023-02-02 16:19:26 +08:00
05671fcb42
[autochunk] support multi outputs chunk search ( #2538 )
...
Support multi outputs chunk search. Previously we only support single output chunk search. It is more flexible and improve performance by a large margin. For transformer, we reduce memory by 40% than previous search strategy.
1. rewrite search strategy to support multi outputs chunk search
2. fix many, many bugs
3. update tests
2023-02-01 13:18:51 +08:00
63199c6687
[autochunk] support transformer ( #2526 )
2023-01-31 16:00:06 +08:00
a4ed9125ac
[hotfix] fix lightning error ( #2529 )
2023-01-31 10:40:39 +08:00
66dfcf5281
[gemini] update the gpt example ( #2527 )
2023-01-30 17:58:05 +08:00
b528eea0f0
[zero] add zero wrappers ( #2523 )
...
* [zero] add zero wrappers
* change names
* add wrapper functions to init
2023-01-29 17:52:58 +08:00
c198c7c0b0
[hotfix] meta tensor default device. ( #2510 )
2023-01-29 16:28:10 +08:00
077a5cdde4
[zero] fix gradient clipping in hybrid parallelism ( #2521 )
...
* [zero] fix gradient clipping in hybrid parallelism
* [testing] change model name to avoid pytest warning
* [hotfix] fix unit testing
2023-01-29 15:09:57 +08:00
aa0f6686f9
[autoparallel] accelerate gpt2 training ( #2495 )
2023-01-29 11:13:15 +08:00
707b11d4a0
[gemini] update ddp strict mode ( #2518 )
...
* [zero] add strict ddp mode for chunk init
* [gemini] update gpt example
2023-01-28 14:35:25 +08:00
2d1a7dfe5f
[zero] add strict ddp mode ( #2508 )
...
* [zero] add strict ddp mode
* [polish] add comments for strict ddp mode
* [zero] fix test error
2023-01-20 14:04:38 +08:00
c04f183237
[autochunk] support parsing blocks ( #2506 )
2023-01-20 11:18:17 +08:00
35c0c0006e
[utils] lazy init. ( #2148 )
...
* [utils] lazy init.
* [utils] remove description.
* [utils] complete.
* [utils] finalize.
* [utils] fix names.
2023-01-20 10:49:00 +08:00
72341e65f4
[auto-chunk] support extramsa ( #3 ) ( #2504 )
2023-01-20 10:13:03 +08:00
0f02b8c6e6
add avg partition ( #2483 )
...
Co-authored-by: Ziyue Jiang <ziyue.jiang@gmail.com>
2023-01-19 13:54:50 +08:00
99d9713b02
Revert "Update parallel_context.py ( #2408 )"
...
This reverts commit 7d5640b9db
2023-01-19 12:27:48 +08:00
ecccc91f21
[autochunk] support autochunk on evoformer ( #2497 )
2023-01-19 11:41:00 +08:00
5db3a5bf42
[fx] allow control of ckpt_codegen init ( #2498 )
...
* [fx] allow control of ckpt_codegen init
Currently in ColoGraphModule, ActivationCheckpointCodeGen will be set automatically in __init__. But other codegen can't be set if so.
So I add an arg to control whether to set ActivationCheckpointCodeGen in __init__.
* code style
2023-01-18 17:02:46 +08:00
d565a24849
[zero] add unit testings for hybrid parallelism ( #2486 )
2023-01-18 10:36:10 +08:00
4953b4ace1
[autochunk] support evoformer tracer ( #2485 )
...
support full evoformer tracer, which is a main module of alphafold. previously we just support a simplifed version of it.
1. support some evoformer's op in fx
2. support evoformer test
3. add repos for test code
2023-01-16 19:25:05 +08:00
67e1912b59
[autoparallel] support origin activation ckpt on autoprallel system ( #2468 )
2023-01-16 16:25:13 +08:00
fef5c949c3
polish pp middleware ( #2476 )
...
Co-authored-by: Ziyue Jiang <ziyue.jiang@gmail.com>
2023-01-13 16:56:01 +08:00
a5dc4253c6
[zero] polish low level optimizer ( #2473 )
2023-01-13 14:56:17 +08:00
8b7495dd54
[example] integrate seq-parallel tutorial with CI ( #2463 )
2023-01-13 14:40:05 +08:00
867c8c2d3a
[zero] low level optim supports ProcessGroup ( #2464 )
2023-01-13 10:05:58 +08:00
14d9299360
[cli] fixed hostname mismatch error ( #2465 )
2023-01-12 14:52:09 +08:00
9358262992
Fix False warning in initialize.py ( #2456 )
...
* Update initialize.py
* pre-commit run check
2023-01-12 13:49:01 +08:00
8221fd7485
[autoparallel] update binary elementwise handler ( #2451 )
...
* [autoparallel] update binary elementwise handler
* polish
2023-01-12 09:35:10 +08:00
2bfeb24308
[zero] add warning for ignored parameters ( #2446 )
2023-01-11 15:30:09 +08:00
39163417a1
[example] updated the hybrid parallel tutorial ( #2444 )
...
* [example] updated the hybrid parallel tutorial
* polish code
2023-01-11 15:17:17 +08:00
5521af7877
[zero] fix state_dict and load_state_dict for ddp ignored parameters ( #2443 )
...
* [ddp] add is_ddp_ignored
[ddp] rename to is_ddp_ignored
* [zero] fix state_dict and load_state_dict
* fix bugs
* [zero] update unit test for ZeroDDP
2023-01-11 14:55:41 +08:00
2731531bc2
[autoparallel] integrate device mesh initialization into autoparallelize ( #2393 )
...
* [autoparallel] integrate device mesh initialization into autoparallelize
* add megatron solution
* update gpt autoparallel examples with latest api
* adapt beta value to fit the current computation cost
2023-01-11 14:03:49 +08:00
c72c827e95
[cli] provided more details if colossalai run fail ( #2442 )
2023-01-11 13:56:42 +08:00
c41e59e5ad
[fx] allow native ckpt trace and codegen. ( #2438 )
2023-01-11 13:49:59 +08:00
41429b9b28
[autoparallel] add shard option ( #2423 )
2023-01-11 13:40:33 +08:00
7829aa094e
[ddp] add is_ddp_ignored ( #2434 )
...
[ddp] rename to is_ddp_ignored
2023-01-11 12:22:45 +08:00
bb4e9a311a
[zero] add inference mode and its unit test ( #2418 )
2023-01-11 10:07:37 +08:00
93f62dd152
[autochunk] add autochunk feature
2023-01-10 16:04:42 +08:00
dddacd2d2c
[hotfix] add norm clearing for the overflow step ( #2416 )
2023-01-10 15:43:06 +08:00
7ab2db206f
adapt new fx
2023-01-10 11:56:00 +08:00
e532679c95
Merge branch 'main' of https://github.com/oahzxl/ColossalAI into chunk
2023-01-10 11:29:01 +08:00
7d5640b9db
Update parallel_context.py ( #2408 )
2023-01-10 11:27:23 +08:00
fd818cf144
change imports
2023-01-10 11:10:45 +08:00
a591d45b29
add available
2023-01-10 10:56:39 +08:00
615e7e68d9
update doc
2023-01-10 10:44:07 +08:00
7d4abaa525
add doc
2023-01-10 09:59:47 +08:00
1be0ac3cbf
add doc for trace indice
2023-01-09 17:59:52 +08:00
0b6af554df
remove useless function
2023-01-09 17:46:43 +08:00
d914a21d64
rename
2023-01-09 17:45:36 +08:00
865f2e0196
rename
2023-01-09 17:42:25 +08:00
ea13a201bb
[polish] polish code for get_static_torch_model ( #2405 )
...
* [gemini] polish code
* [testing] remove code
* [gemini] make more robust
2023-01-09 17:41:38 +08:00
a4ed5b0d0d
rename in doc
2023-01-09 17:41:26 +08:00
1bb1f2ad89
rename
2023-01-09 17:38:16 +08:00
cb9817f75d
rename function from index to indice
2023-01-09 17:34:30 +08:00
0ea903b94e
rename trace_index to trace_indice
2023-01-09 17:25:13 +08:00
551cafec14
[doc] updated kernel-related optimisers' docstring ( #2385 )
...
* [doc] updated kernel-related optimisers' docstring
* polish doc
2023-01-09 17:13:53 +08:00
065f0b4c27
add doc for search
2023-01-09 17:11:51 +08:00
a68d240ed5
add doc for search chunk
2023-01-09 16:54:08 +08:00
1951f7fa87
code style
2023-01-09 16:30:16 +08:00
212b5b1b5f
add comments
2023-01-09 16:29:33 +08:00
19cc64b1d3
remove autochunk_available
2023-01-09 16:06:58 +08:00
9880fd2cd8
Fix state_dict key missing issue of the ZeroDDP ( #2363 )
...
* Fix state_dict output for ZeroDDP duplicated parameters
* Rewrite state_dict based on get_static_torch_model
* Modify get_static_torch_model to be compatible with the lower version (ZeroDDP)
2023-01-09 14:35:14 +08:00
4d223e18a2
fix typo
2023-01-09 13:46:17 +08:00
ce08661eb1
[cli] updated installation check cli for aot/jit build ( #2395 )
2023-01-09 11:05:27 +08:00
69d9180c4b
[hotfix] issue #2388
2023-01-07 18:23:02 +08:00
4e96039649
[device] find best logical mesh
2023-01-07 14:04:30 +08:00
8f72b6f8fb
[hotfix] fix implement error in diffusers
2023-01-07 07:56:39 +08:00
40d376c566
[setup] support pre-build and jit-build of cuda kernels ( #2374 )
...
* [setup] support pre-build and jit-build of cuda kernels
* polish code
* polish code
* polish code
* polish code
* polish code
* polish code
2023-01-06 20:50:26 +08:00
33f3023e19
[hotfix] fix implement error in diffusers
2023-01-06 18:37:18 +08:00
12c8bf38d7
[Pipeline] Refine GPT PP Example
2023-01-06 18:03:45 +08:00
8a989a0d89
code style
2023-01-06 17:55:22 +08:00
c3a2bf48b4
code style
2023-01-06 17:31:59 +08:00
a6cdbf9161
seperate trace flow
2023-01-06 17:24:23 +08:00
4748967fb1
ad reorder graph
2023-01-06 17:13:18 +08:00
da4076846d
rename
2023-01-06 17:09:37 +08:00
c3d72f7db9
seperate reorder
2023-01-06 16:53:01 +08:00
a881d6d000
Revert "[NFC] polish code format" ( #2372 )
2023-01-06 16:01:09 +08:00
9ae9e74017
fix diff device in some partition
2023-01-06 15:59:06 +08:00
0dcc410f57
[NFC] polish code format
2023-01-06 15:54:06 +08:00
6685a9d022
seperate non chunk input
2023-01-06 15:53:24 +08:00
d634eae05b
Revert "[NFC] polish code format ( #2367 )" ( #2371 )
...
This reverts commit 1f8ab6f1f5
2023-01-06 15:52:16 +08:00
f856611d21
seperate prepose_nodes
2023-01-06 15:47:17 +08:00
d42aecdda1
[NFC] polish colossalai/auto_parallel/tensor_shard/deprecated/op_handler/embedding_handler.py code style ( #2368 )
2023-01-06 15:47:10 +08:00
1aaeb596c6
[example] gpt, shard init on all processes ( #2366 )
2023-01-06 15:44:50 +08:00
f4a1607e56
seperate input node dim search
2023-01-06 15:36:17 +08:00
1f8ab6f1f5
[NFC] polish code format ( #2367 )
2023-01-06 15:34:48 +08:00
ae27a8b26d
seperate flow tracer
2023-01-06 14:57:33 +08:00
fd87d78a28
rename ambiguous variable
2023-01-06 14:28:04 +08:00
2bde9d2b7f
code format
2023-01-06 14:21:49 +08:00
8a634af2f5
close mem and code print
2023-01-06 14:19:45 +08:00
1a6d2a740b
take apart chunk code gen
2023-01-06 14:14:45 +08:00
ac0d30fe2e
[NFC] polish batch_norm_handler.py code style ( #2359 )
2023-01-06 13:41:38 +08:00
48d33b1b17
[gemini] add get static torch model ( #2356 )
2023-01-06 13:41:19 +08:00
efb1c64c30
restruct dir
2023-01-06 11:39:26 +08:00
7080a8edb0
[workflow]New version: Create workflow files for examples' auto check ( #2298 )
...
* [workflows]bug_repair
* [workflow]new_pr_fixing_bugs
Co-authored-by: binmakeswell <binmakeswell@gmail.com>
2023-01-06 09:26:49 +08:00
e11a005c02
[NFC] polish colossalai/auto_parallel/tensor_shard/utils/factory.py code style ( #2349 )
2023-01-05 21:17:42 +08:00
b5a3a4a65f
[device] find best logical mesh
2023-01-05 17:21:29 +08:00
28e2d16794
[NFC] polish colossalai/auto_parallel/tensor_shard/deprecated/graph_analysis.py code style ( #2340 )
2023-01-05 16:53:24 +08:00
9c9246c0d9
[device] alpha beta profiler ( #2311 )
...
* [device] alpha beta profiler
* add usage
* fix variable name
2023-01-05 16:39:55 +08:00
bd12a49e2a
[NFC] polish <colossalai/auto_parallel/tensor_shard/deprecated/constants.py> code style ( #2339 )
2023-01-05 16:20:54 +08:00
35427bcab4
[NFC] polish colossalai/auto_parallel/tensor_shard/deprecated/op_handler/unary_elementwise_handler.py code style ( #2326 )
2023-01-05 12:18:08 +08:00
db6eea3583
[builder] reconfig op_builder for pypi install ( #2314 )
2023-01-04 16:32:32 +08:00
4a79c10750
[NFC] polish colossalai/cli/benchmark/__init__.py code style ( #2308 )
2023-01-04 15:09:57 +08:00
87d2defda6
[NFC] polish colossalai/auto_parallel/tensor_shard/deprecated/op_handler/layer_norm_handler.py code style ( #2305 )
2023-01-04 15:09:57 +08:00
116e3d0b8f
[NFC] polish communication/p2p_v2.py code style ( #2303 )
2023-01-04 15:09:57 +08:00
b965585d05
[NFC] polish colossalai/amp/torch_amp/torch_amp.py code style ( #2290 )
2023-01-04 15:09:57 +08:00
d1e5bafcd4
[NFC] polish colossalai/auto_parallel/tensor_shard/deprecated/__init__.py code style ( #2291 )
2023-01-04 15:09:57 +08:00
950685873f
[NFC] polish colossalai/auto_parallel/tensor_shard/deprecated/op_handler/reshape_handler.py code style ( #2292 )
2023-01-04 15:09:57 +08:00
3041014089
[NFC] polish colossalai/amp/naive_amp/grad_scaler/dynamic_grad_scaler.py code style ( #2299 )
...
Co-authored-by: henryqin1997 <henryqin1997@gamil.com>
2023-01-04 15:09:57 +08:00
49715a78f0
[NFC] polish colossalai/cli/benchmark/benchmark.py code style ( #2287 )
2023-01-04 15:09:57 +08:00
1c29b173c9
[NFC] polish colossalai/auto_parallel/tensor_shard/node_handler/getitem_handler.py code style ( #2289 )
2023-01-04 15:09:57 +08:00
3a02b46447
[auto-parallel] refactoring ColoTracer ( #2118 )
...
* add meta_data_computing
* add checkpoint_annotation
* rename proxy.data to proxy.meta_data and add bias addition pass
* polish code
* delete meta_prop_pass invoke and rename ori_node to orig_node
* add TracerType
* unify meta data computing
* delete TracerType
* handle setitem operation
* operator.setitem
2023-01-04 14:44:22 +08:00
5d3a2be3af
[amp] add gradient clipping for unit tests ( #2283 )
...
* [amp] add gradient clipping in unit tests
* fix bugs
2023-01-04 11:59:56 +08:00
d45695d94e
Merge pull request #2258 from hpcaitech/debug/ckpt-autoparallel
...
[autockpt] provide option for activation checkpoint search in SPMD solver
2023-01-04 11:37:28 +08:00
16cc8e6aa7
[builder] MOE builder ( #2277 )
2023-01-03 20:29:39 +08:00
b904748210
[autoparallel] bypass MetaInfo when unavailable and modify BCAST_FUNC_OP metainfo ( #2293 )
...
* [autoparallel] align the data_ptr with the old version of auto activation checkpoint pipeline
* [autoparallel] using fwd_time and bwd_time instead of fwd_flop and bwd_flop
* [autoparallel] specifycomm nodes' memory cost in construct chain
* [autoparallel] fix wrong runtime apply calculation
* [autoparallel] fix wrong runtime apply calculation
* [autoparallel] fix wrong runtime apply calculation
* [autoparallel] bypass metainfo when available and modify BCAST_FUNC_OP
2023-01-03 20:28:01 +08:00
8ea50d999e
[hotfix] pass a parameter. ( #2288 )
...
* [autockpt] make it work.
* [autockpt] linearize / merge shape-consistency nodes.
* [autockpt] considering parameter and optimizer weights.
* [hotfix] pass a parameter.
2023-01-03 18:05:06 +08:00
e94c79f15b
improved allgather & reducescatter for 3d
2023-01-03 17:46:08 +08:00
62c38e3330
[zero] polish low level zero optimizer ( #2275 )
2023-01-03 17:22:34 +08:00
ac863a01d6
[example] add benchmark ( #2276 )
...
* add benchmark
* merge common func
* add total and avg tflops
Co-authored-by: Ziyue Jiang <ziyue.jiang@gmail.com>
2023-01-03 17:20:59 +08:00
22e947f982
[autoparallel] fix runtime apply memory estimation ( #2281 )
...
* [autoparallel] align the data_ptr with the old version of auto activation checkpoint pipeline
* [autoparallel] using fwd_time and bwd_time instead of fwd_flop and bwd_flop
* [autoparallel] specifycomm nodes' memory cost in construct chain
* [autoparallel] fix wrong runtime apply calculation
* [autoparallel] fix wrong runtime apply calculation
* [autoparallel] fix wrong runtime apply calculation
2023-01-03 17:18:07 +08:00
8e8900ff3f
[autockpt] considering parameter and optimizer weights. ( #2279 )
...
* [autockpt] make it work.
* [autockpt] linearize / merge shape-consistency nodes.
* [autockpt] considering parameter and optimizer weights.
2023-01-03 16:55:49 +08:00
f027ef7913
[hotfix] fix fp16 optimzier bug ( #2273 )
2023-01-03 16:53:43 +08:00
fb87322773
[autoparallel] fix spelling error ( #2270 )
2023-01-03 16:13:00 +08:00
af32022f74
[Gemini] fix the convert_to_torch_module bug ( #2269 )
2023-01-03 15:55:35 +08:00
b0d21d0c4f
[autockpt] linearize / merge shape-consistency nodes. ( #2271 )
...
* [autockpt] make it work.
* [autockpt] linearize / merge shape-consistency nodes.
2023-01-03 14:54:22 +08:00
4b29112ab2
[autoparallel] gpt2 autoparallel examples ( #2267 )
...
* [autoparallel] gpt2 autoparallel examples
* polish code
* polish code
2023-01-03 14:23:33 +08:00
8b045b3c1f
[Pipeline Middleware] Reduce comm redundancy by getting accurate output ( #2232 )
...
* move to cpu to avoid dead lock
* get output by offsets
Co-authored-by: Ziyue Jiang <ziyue.jiang@gmail.com>
2023-01-03 13:43:57 +08:00
5c2ef9fc76
[autoparallel] modify comm nodes' memory cost in construct chain ( #2263 )
...
* [autoparallel] align the data_ptr with the old version of auto activation checkpoint pipeline
* [autoparallel] using fwd_time and bwd_time instead of fwd_flop and bwd_flop
* [autoparallel] specifycomm nodes' memory cost in construct chain
2023-01-03 11:38:48 +08:00
1ea99b869e
[autoparallel] align the data_ptr with the old version of auto activation checkpoint pipeline ( #2261 )
2023-01-03 10:30:15 +08:00
3ccf58aa76
[autockpt] make it work. ( #2257 )
2023-01-02 23:37:45 +08:00
ac3739930d
[autoparallel] modify construct chain in rotor solver ( #2254 )
2023-01-02 16:26:12 +08:00
ab38aebace
[autoparallel] Hook all meta information on ResNet nodes for auto activation checkpoint ( #2248 )
...
* [autoparallel] hook node meta on graph nodes for checkpoint solver
* [autoparallel] polish code
* [autoparallel] restore some node handlers
* colossalai/auto_parallel/passes/meta_info_prop.py
* [autoparallel] remove some unused import
* [autoparallel] hook bwd_mem_out
2023-01-02 16:25:18 +08:00
c8c79102f0
[autoparallel] patch torch.flatten metainfo for autoparallel ( #2247 )
...
* [autoparallel] patch torch.flatten
2023-01-02 15:51:03 +08:00
8897b8f753
[autoparallel] autoparallel initialize ( #2238 )
2022-12-31 01:02:14 +08:00
85178a397a
[hotfix] fix error for torch 2.0 ( #2243 )
2022-12-30 23:11:55 +08:00
b7d0990c61
[autoparallel] fix construct meta info. ( #2245 )
2022-12-30 19:56:44 +08:00
57929a6210
fix type of num_worker_threads ( #2237 )
...
Co-authored-by: Ziyue Jiang <ziyue.jiang@gmail.com>
2022-12-30 11:04:01 +08:00
db4cbdc7fb
[builder] builder for scaled_upper_triang_masked_softmax ( #2234 )
2022-12-30 09:58:00 +08:00
78483a9fdd
[logger] hotfix, missing _FORMAT ( #2231 )
2022-12-29 22:59:39 +08:00
54de05da5d
[builder] polish builder with better base class ( #2216 )
...
* [builder] polish builder
* remove print
2022-12-28 19:45:49 +08:00
3b1b91eaf4
[autoparallel] record parameter attribute in colotracer ( #2217 )
...
* [autoparallel] record parameter attribute in collotracer
* [autoparallel] fix construct_meta_info bug
2022-12-28 19:29:08 +08:00
7675792100
[builder] raise Error when CUDA_HOME is not set ( #2213 )
2022-12-28 16:07:08 +08:00
d5e3e3ec01
[example] update gpt example for larger model scale ( #2211 )
2022-12-28 13:54:08 +08:00
24246f7aa5
[autoparallel] Attach input, buffer and output tensor to MetaInfo class ( #2162 )
...
* [fx] metainfo class for auto parallel
* [fx] add unit test for linear metainfo
* [fx] fix bwd param for linear
* [fx] modify unit test
* [fx] modify unit test
* [fx] modify import
* [fx] modify import
* [fx] modify import
* [fx] move meta profiler to auto parallel
* [fx] add conv metainfo class
* [fx] restore profiler
* [fx] restore meta profiler
* [autoparallel] modify unit test
* [fx] modify unit test
* [autoparallel] add batchnorm metainfo class
* [autoparallel] fix batchnorm unit test function declaration
* [fx] restore profiler
* [fx] add relu metainfo class
* [fx] restore profiler
* [autoparallel] modify metainfo input
* [autoparallel] add pooling metainfo
* [autoparallel] add F.linear metainfo generator
* [autoparallel] add binary elementwise metainfo
* [fx] recover profiler
* [autoparallel] fix forward memory calculation
* [autoparallel] modify constants.py
* [autoparallel] remove redundant print
* [autoparallel] add F.conv metainfo
* [autoparallel] linear fix
* [autoparallel] memory estimation for communication actions
* [autoparallel] fix docstring
* [autoparallel] fix variables name
* [autoparallel] attach tensor to metainfo class
* [autoparallel] fix dangerous try except
* [autoparallel] attach memory cost to shape consistency node
* [autoparallel] attach shape consistency node's metainfo to the node
* [autoparallel] remove todo in shape consistency memory estimation
* [autoparallel] fix the annotation
2022-12-28 13:37:40 +08:00
d0bc5a1b34
[autoparallel] new metainfoprop based on metainfo class ( #2179 )
...
* [autoparallel] new metainfoprop to combine SPMD solver and checkpoint solver
* [autoparallel] new metainfoprop to combine SPMD solver and checkpoint solver
* [autoparallel] modify placeholder handler
* [autoparallel] modify metainfoprop
* [autoparallel] fix function typo
* [autoparallel] fix placeholder handler
2022-12-28 13:35:08 +08:00
78509124d3
[autoparallel] update getitem handler ( #2207 )
2022-12-27 19:58:32 +08:00
1cb532ffec
[builder] multihead attn runtime building ( #2203 )
...
* [hotfix] correcnt cpu_optim runtime compilation
* [builder] multihead attn
* fix bug
* fix a bug
2022-12-27 16:06:09 +08:00
8e22c38b89
[hotfix] Fixing the bug related to ipv6 support
...
Co-authored-by: ByteDance <tongping.liu@bytedance.com>
2022-12-27 12:42:46 +08:00
4851f2d607
[autoparallel] update_getattr_handler ( #2193 )
2022-12-26 21:57:39 +08:00
5682e6d346
[hotfix] correcnt cpu_optim runtime compilation ( #2197 )
2022-12-26 16:45:14 +08:00
2458659919
[zero] fix error for BEiT models ( #2169 )
...
* [zero] fix error for BEiT models
* [ColoParameter] add unpack operation for tuple arguments
* fix bugs
* fix chunkv2 unit testing
* add assertion for gradient state
2022-12-26 15:03:54 +08:00
355ffb386e
[builder] unified cpu_optim fused_optim inferface ( #2190 )
2022-12-23 20:57:41 +08:00
9587b080ba
[builder] use runtime builder for fused_optim ( #2189 )
2022-12-23 17:07:03 +08:00
bc0e271e71
[buider] use builder() for cpu adam and fused optim in setup.py ( #2187 )
2022-12-23 16:05:13 +08:00
d42afd30f8
[builder] runtime adam and fused_optim builder ( #2184 )
2022-12-23 14:14:21 +08:00
550f8f8905
[autoparallel] integrate_gpt_related_tests ( #2134 )
...
* [autoparallel] integrate_gpt_related_tests
* polish code
* polish code
* add GPT2Model into runtime test
2022-12-23 12:36:59 +08:00
59e343328d
[Pipeline Middleware ] Fix deadlock when num_microbatch=num_stage ( #2156 )
...
* add splitter
* polish code
* remove comment
* fix async nan by moving to cpu first
Co-authored-by: Ziyue Jiang <ziyue.jiang@gmail.com>
2022-12-23 11:38:43 +08:00
ab54fed292
[hotfix] add kwargs for colo_addmm ( #2171 )
2022-12-22 13:25:30 +08:00
622f863291
[hotfix] Jit type hint #2161 ( #2164 )
2022-12-22 10:17:03 +08:00
12e7bcd720
register meta func for rnn ( #2159 )
2022-12-21 23:06:18 +08:00
cfe2a9bd90
[autoparallel] memory estimation for shape consistency ( #2144 )
...
* [fx] metainfo class for auto parallel
* [fx] add unit test for linear metainfo
* [fx] fix bwd param for linear
* [fx] modify unit test
* [fx] modify unit test
* [fx] modify import
* [fx] modify import
* [fx] modify import
* [fx] move meta profiler to auto parallel
* [fx] add conv metainfo class
* [fx] restore profiler
* [fx] restore meta profiler
* [autoparallel] modify unit test
* [fx] modify unit test
* [autoparallel] add batchnorm metainfo class
* [autoparallel] fix batchnorm unit test function declaration
* [fx] restore profiler
* [fx] add relu metainfo class
* [fx] restore profiler
* [autoparallel] modify metainfo input
* [autoparallel] add pooling metainfo
* [autoparallel] add F.linear metainfo generator
* [autoparallel] add binary elementwise metainfo
* [fx] recover profiler
* [autoparallel] fix forward memory calculation
* [autoparallel] modify constants.py
* [autoparallel] remove redundant print
* [autoparallel] add F.conv metainfo
* [autoparallel] linear fix
* [autoparallel] memory estimation for communication actions
* [autoparallel] fix docstring
* [autoparallel] fix variables name
2022-12-21 10:39:37 +08:00
b87496a66b
[hotfix] fix auto policy of test_sharded_optim_v2 ( #2157 )
2022-12-20 23:03:18 +08:00
16335cb537
[hotfix] fix aten default bug ( #2158 )
2022-12-20 22:40:46 +08:00
a7d95b7024
[example] add zero1, zero2 example in GPT examples ( #2146 )
...
* [example] add zero1 and zero2 for GPT
* update readme in gpt example
* polish code
* change init value
* update readme
2022-12-20 14:30:27 +08:00
1cce6e36ca
[autoparallel] use metainfo in handler ( #2149 )
2022-12-20 10:31:22 +08:00
2827f41898
[Gemini] GeminiDPP convert to PyTorch Module. ( #2151 )
2022-12-20 10:19:36 +08:00
bdef9dfdbe
[NFC] remove useless graph node code ( #2150 )
2022-12-20 00:33:58 +08:00
b3f73ce1c8
[Gemini] Update coloinit_ctx to support meta_tensor ( #2147 )
2022-12-19 22:37:07 +08:00
a128eec9d5
register aten._convolution.default ( #2137 )
2022-12-18 19:27:01 +08:00
ee287620f0
[Gemini] revert ZeROInitCtx related tracer ( #2138 )
2022-12-16 12:37:06 +08:00
077a66dd81
updated attention kernel ( #2133 )
2022-12-16 10:54:03 +08:00
a3c6924deb
[autoparallel] process size nodes in runtime pass ( #2130 )
...
* [autoparallel] process size nodes in runtime pass
* polish code
2022-12-14 16:10:50 +08:00
536560ccc0
[autoparallel] implement softmax handler ( #2132 )
2022-12-14 16:09:53 +08:00
c89c66a858
[Gemini] update API of the chunkmemstatscollector. ( #2129 )
2022-12-14 00:47:06 +08:00
2938edf446
[Gemini] update the non model data record method in runtime memory tracer ( #2128 )
2022-12-13 17:11:31 +08:00
8fac837679
[Gemini] update non model data calculation method ( #2126 )
2022-12-13 15:44:07 +08:00
5efda69735
[Gemini] hotfix the unittest bugs ( #2125 )
2022-12-13 14:14:55 +08:00
05bb28aacf
[Gemini] mapping of preop timestep and param ( #2124 )
2022-12-13 12:50:24 +08:00
cd0af9f7f6
[autoparallel] gpt2lp runtimee test ( #2113 )
2022-12-12 18:06:40 +08:00
9214d1fe28
[Gemini] chunk init using runtime visited param order ( #2115 )
2022-12-12 18:06:16 +08:00
e7d3afc9cc
[optimizer] add div_scale for optimizers ( #2117 )
...
* [optimizer] add div_scale for optimizers
* [zero] use div_scale in zero optimizer
* fix testing error
2022-12-12 17:58:57 +08:00
e5aa8333e4
[NFC] update chunk manager API ( #2119 )
2022-12-12 16:57:22 +08:00
e99edfcb51
[NFC] polish comments for Chunk class ( #2116 )
2022-12-12 15:39:31 +08:00
09d69e1c25
[PP Middleware] Add bwd and step for PP middleware ( #2111 )
...
* add bwd and step for PP middleware
* pre-commit
Co-authored-by: Ziyue Jiang <ziyue.jiang@gmail.com>
2022-12-12 12:40:03 +08:00
8afc001f4f
[Gemini] chunk init use OrderedParamGenerator ( #2110 )
2022-12-11 21:41:13 +08:00
63fbba3c19
[zero] add L2 gradient clipping for ZeRO ( #2112 )
...
* [zero] add L2 gradient clipping
* [testing] add MlpModel
* [zero] add unit test for grad clipping
* fix atol
2022-12-09 18:09:17 +08:00
70a8556946
[gemini] get the param visited order during runtime ( #2108 )
2022-12-09 16:13:03 +08:00
61f31c3cf0
[Gemini] NFC, polish search_chunk_configuration ( #2107 )
2022-12-09 15:00:39 +08:00
8e14344ec9
[hotfix] fix a type in ColoInitContext ( #2106 )
2022-12-09 11:44:39 +08:00
05545bfee9
[ColoTensor] throw error when ColoInitContext meets meta parameter. ( #2105 )
2022-12-09 11:39:46 +08:00
d87baa85d9
[autoparallel] support linear function bias addition ( #2104 )
2022-12-09 10:31:36 +08:00
0fecbb9e20
[autoparallel] support addbmm computation ( #2102 )
2022-12-08 21:15:11 +08:00
d3d4630495
[autoparallel] add sum handler ( #2101 )
2022-12-08 17:02:54 +08:00
e4705ba4e2
[Pipeline Middleware] fix data race in Pipeline Scheduler for DAG ( #2087 )
...
* add DAG test case
* fix datarace by adjusting theposition of lock
* polish code
* fix pytest for middleware
* remove test
Co-authored-by: Ziyue Jiang <ziyue.jiang@gmail.com>
2022-12-08 13:32:27 +08:00
b175e6d58e
[autoparallel] add bias addtion function class ( #2098 )
...
* [autoparallel] add bias addtion function class
* polish code
* polish
2022-12-08 11:31:51 +08:00
3af7e65dea
[autoparallel] complete gpt related module search ( #2097 )
2022-12-08 10:04:09 +08:00
85efb7ac2e
[Gemini] gemini use the runtime memory tracer (RMT) ( #2099 )
2022-12-07 23:04:02 +08:00
2bf2d1cd3b
[fx] An experimental version of ColoTracer.' ( #2002 )
...
* [fx] add a symbolic_trace api.
* [fx] fix import errors.
* [fx] ColoTracer experimental.
2022-12-07 18:36:17 +08:00
4b055351b0
[Gemini] make RuntimeMemTracer work correctly ( #2096 )
2022-12-07 16:59:59 +08:00
7f72eb0510
[autoparallel]add embedding handler ( #2089 )
...
* [autoparallel] add embedding handler
* fix bugs
2022-12-07 09:41:46 +08:00
1fca5d79ea
[Gemini] remove GLOBAL_MODEL_DATA_TRACER ( #2091 )
2022-12-06 22:30:16 +08:00
28e55c2530
[Gemini] remove GLOBAL_CUDA_MEM_INFO ( #2090 )
2022-12-06 22:10:47 +08:00
25abae6d7f
[Gemini] use MemStats in Runtime Memory tracer ( #2088 )
2022-12-06 19:48:20 +08:00
33f4412102
[Gemini] use MemStats to store the tracing data. Seperate it from Collector. ( #2084 )
2022-12-06 16:43:06 +08:00
1f99205827
[Gemini] remove static tracer ( #2083 )
2022-12-06 12:53:58 +08:00
0e9db368ef
[autoparallel] add tensor constructor handler ( #2082 )
2022-12-06 10:20:10 +08:00
cdf537a648
[autoparallel] add non_split linear strategy ( #2078 )
...
* [autoparallel] add non_split linear stategy
* polish
2022-12-06 10:19:33 +08:00
cf0268da93
[autoparallel] Add F.conv metainfo ( #2069 )
...
* [fx] metainfo class for auto parallel
* [fx] add unit test for linear metainfo
* [fx] fix bwd param for linear
* [fx] modify unit test
* [fx] modify unit test
* [fx] modify import
* [fx] modify import
* [fx] modify import
* [fx] move meta profiler to auto parallel
* [fx] add conv metainfo class
* [fx] restore profiler
* [fx] restore meta profiler
* [autoparallel] modify unit test
* [fx] modify unit test
* [autoparallel] add batchnorm metainfo class
* [autoparallel] fix batchnorm unit test function declaration
* [fx] restore profiler
* [fx] add relu metainfo class
* [fx] restore profiler
* [autoparallel] modify metainfo input
* [autoparallel] add pooling metainfo
* [autoparallel] add F.linear metainfo generator
* [autoparallel] add binary elementwise metainfo
* [fx] recover profiler
* [autoparallel] fix forward memory calculation
* [autoparallel] modify constants.py
* [autoparallel] remove redundant print
* [autoparallel] add F.conv metainfo
* [autoparallel] linear fix
2022-12-06 10:17:57 +08:00
f123476666
[autoparallel] complete gpt block searching ( #2065 )
...
* [autoparallel] complete gpt block searching
* fix test
2022-12-06 10:17:10 +08:00
597cdd3006
[Pipeline Middleware] Adapt scheduler for Topo ( #2066 )
...
* adapt scheduler for Topo
* remoove comment
* fix set input
Co-authored-by: Ziyue Jiang <ziyue.jiang@gmail.com>
2022-12-05 20:23:41 +08:00
b3b89865e2
[Gemini] ParamOpHook -> ColoParamOpHook ( #2080 )
2022-12-05 17:11:06 +08:00
677e1e20d4
[device] update flatten device mesh usage ( #2079 )
2022-12-05 16:16:07 +08:00
a7adad9ccb
[Gemini] rename hooks related to runtime mem tracer ( #2076 )
2022-12-05 15:00:03 +08:00
223332ff7e
[Gemini] rename ParamTracerWrapper -> RuntimeMemTracer ( #2073 )
2022-12-05 12:45:11 +08:00
9f828ef36f
[Gemini] remove not used MemtracerWrapper ( #2072 )
2022-12-05 11:57:59 +08:00
616da17fab
[autoparallel] add binary elementwise metainfo for auto parallel ( #2058 )
...
* [fx] metainfo class for auto parallel
* [fx] add unit test for linear metainfo
* [fx] fix bwd param for linear
* [fx] modify unit test
* [fx] modify unit test
* [fx] modify import
* [fx] modify import
* [fx] modify import
* [fx] move meta profiler to auto parallel
* [fx] add conv metainfo class
* [fx] restore profiler
* [fx] restore meta profiler
* [autoparallel] modify unit test
* [fx] modify unit test
* [autoparallel] add batchnorm metainfo class
* [autoparallel] fix batchnorm unit test function declaration
* [fx] restore profiler
* [fx] add relu metainfo class
* [fx] restore profiler
* [autoparallel] modify metainfo input
* [autoparallel] add pooling metainfo
* [autoparallel] add F.linear metainfo generator
* [autoparallel] add binary elementwise metainfo
* [fx] recover profiler
* [autoparallel] fix forward memory calculation
* [autoparallel] modify constants.py
* [autoparallel] remove redundant print
2022-12-04 15:18:51 +08:00
4b40fbd743
[autoparallel] fix forward memory calculation ( #2062 )
2022-12-04 15:00:16 +08:00
44ea461890
[Pipeline] Add Topo Class ( #2059 )
...
* use Topo class to rewrite DAG
* polish code
* polish code
* polish code
* add comment
* add else to unended if
Co-authored-by: Ziyue Jiang <ziyue.jiang@gmail.com>
2022-12-02 18:13:20 +08:00
e4293e5077
[hotfix] update test for latest version ( #2060 )
2022-12-02 18:12:30 +08:00
38ea4ba1bd
[Gemini] fix grad unreleased issue and param recovery issue ( #2052 )
2022-12-02 16:04:19 +08:00
1c1fe44305
[autoparallel] adapt solver with self attention ( #2037 )
...
* [autoparallel] adapt solver with self attention
* polish code
2022-12-01 17:53:15 +08:00
ea74a3b9cc
[cli] updated installation cheheck with more inforamtion ( #2050 )
...
* [cli] updated installation cheheck with more inforamtion
* polish code
* polish code
2022-11-30 17:53:55 +08:00
f6178728a0
[gemini] fix init bugs for modules ( #2047 )
...
* [gemini] fix init bugs for modules
* fix bugs
2022-11-30 17:06:10 +08:00
81e0da7fa8
[setup] supported conda-installed torch ( #2048 )
...
* [setup] supported conda-installed torch
* polish code
2022-11-30 16:45:15 +08:00
e37f3db40c
[gemini] add arguments ( #2046 )
...
* [zero] fix testing parameters
* [gemini] add arguments
* add docstrings
2022-11-30 16:40:13 +08:00
6a9158f1fa
[Gemini] free and allocate cuda memory by tensor.storage, add grad hook ( #2040 )
2022-11-30 15:57:45 +08:00
31c644027b
[hotfix] hotfix Gemini for no leaf modules bug ( #2043 )
2022-11-30 14:53:41 +08:00
a1ce02d740
[zero] test gradient accumulation ( #1964 )
...
* [zero] fix memory leak for zero2
* [zero] test gradient accumulation
* [zero] remove grad clip test
2022-11-29 13:00:30 +08:00
b0936e4a44
[rpc] split with dag ( #2028 )
...
* add DAG to split_module
* add comment
* add test case for DAG
* remove print
* add DAG middleware in scheduler
* add test case for scheduler
* remove break
* recover old lifecycle
Co-authored-by: Ziyue Jiang <ziyue.jiang@gmail.com>
2022-11-29 11:36:28 +08:00
96134e7be3
[hotfix] add bert test for gemini fwd bwd ( #2035 )
2022-11-29 11:19:52 +08:00
0dbcd4a6f5
[autoparallel] add split handler ( #2032 )
...
* [autoparallel] add split handler
* add numerical test and runtime passes
2022-11-29 11:03:51 +08:00
28aa9a4294
[Gemini] more rigorous unit tests for run_fwd_bwd ( #2034 )
2022-11-29 09:26:06 +08:00
81330b0352
[autoparallel] add experimental permute handler ( #2029 )
2022-11-27 20:26:52 +08:00
95c4532fff
[Gemini] paramWrapper paramTracerHook unitest ( #2030 )
2022-11-26 13:30:24 +08:00
8daf1b4db1
[Gemini] patch for supporting orch.add_ function for ColoTensor ( #2003 )
2022-11-25 20:06:35 +08:00
632753abbc
[fx]Split partition with DAG information ( #2025 )
...
* add DAG to split_module
* add comment
* add test case for DAG
* remove print
Co-authored-by: Ziyue Jiang <ziyue.jiang@gmail.com>
2022-11-25 17:42:48 +08:00
ea0f6b8df9
[autoparallel] add runtime pass and numerical test for view handler ( #2018 )
2022-11-25 15:50:16 +08:00
a719b89a41
[gemini] param_trace_hook ( #2020 )
2022-11-24 18:08:36 +08:00
0b0d8f9e17
[hotfix] revert bug PRs ( #2016 )
2022-11-24 15:28:58 +08:00
aba3db464d
[Gemini] ParamMemHook ( #2008 )
2022-11-24 15:22:51 +08:00
0160a62a3c
[Gemini] param_tracer_wrapper and test case ( #2009 )
2022-11-24 14:40:33 +08:00
1438993113
[autoparallel] add experimental view handler ( #2011 )
...
* [autoparallel] add experimental view handler
* polish
* polish
* polish code
* rename variables
2022-11-24 11:34:41 +08:00
d655eea515
[autoparallel] mix gather ( #1977 )
...
* Add mix-gather
* Add comments
* Add comments
* Polish comments
* Change the global rank assumption
* Add tests
* Add two-step tests
* Fix 10 and 01
* Skip test becasue the number of GPUs
2022-11-23 21:49:17 +08:00
2bab6f512c
[release] release v0.1.11rc4 ( #2007 )
2022-11-23 17:14:32 +08:00
6cd784ffee
[autoparallel] Add metainfo support for F.linear ( #1987 )
...
* [fx] metainfo class for auto parallel
* [fx] add unit test for linear metainfo
* [fx] fix bwd param for linear
* [fx] modify unit test
* [fx] modify unit test
* [fx] modify import
* [fx] modify import
* [fx] modify import
* [fx] move meta profiler to auto parallel
* [fx] add conv metainfo class
* [fx] restore profiler
* [fx] restore meta profiler
* [autoparallel] modify unit test
* [fx] modify unit test
* [autoparallel] add batchnorm metainfo class
* [autoparallel] fix batchnorm unit test function declaration
* [fx] restore profiler
* [fx] add relu metainfo class
* [fx] restore profiler
* [autoparallel] modify metainfo input
* [autoparallel] add pooling metainfo
* [autoparallel] add F.linear metainfo generator
2022-11-23 14:12:34 +08:00
2edbef13cc
[fx] add more meta_registry for MetaTensor execution. ( #2000 )
...
* [sc] add examples for auto checkpoint.
* merge upstream
* [fx] add more meta_registry for MetaTensor execution.
2022-11-23 10:55:46 +08:00
a2d3266648
[hotfix] make Gemini work for conv DNN ( #1998 )
2022-11-22 14:52:36 +08:00
155891113e
[autoparallel] use pytree map style to process data ( #1989 )
2022-11-21 10:44:22 +08:00
35e6b9ec82
[autoparallel] adapt handlers with attention block ( #1990 )
...
* [autoparallel] adapt handlers with attention block
* polish
2022-11-21 10:44:11 +08:00
05020e50d0
[autoparallel] support more flexible data type ( #1967 )
2022-11-18 17:01:06 +08:00
c26f21d365
[autoparallel] add pooling metainfo ( #1968 )
...
* [fx] metainfo class for auto parallel
* [fx] add unit test for linear metainfo
* [fx] fix bwd param for linear
* [fx] modify unit test
* [fx] modify unit test
* [fx] modify import
* [fx] modify import
* [fx] modify import
* [fx] move meta profiler to auto parallel
* [fx] add conv metainfo class
* [fx] restore profiler
* [fx] restore meta profiler
* [autoparallel] modify unit test
* [fx] modify unit test
* [autoparallel] add batchnorm metainfo class
* [autoparallel] fix batchnorm unit test function declaration
* [fx] restore profiler
* [fx] add relu metainfo class
* [fx] restore profiler
* [autoparallel] modify metainfo input
* [autoparallel] add pooling metainfo
2022-11-18 15:13:03 +08:00
3712ac7f90
[Gemini] add bert for MemtracerWrapper unintests ( #1982 )
2022-11-18 14:58:28 +08:00
e481489aa6
[Gemini] MemtracerWrapper unittests ( #1981 )
2022-11-18 14:19:40 +08:00
31922110ad
[Gemini] memory trace hook ( #1978 )
2022-11-18 11:52:55 +08:00
0529fcde06
[Gemini] independent runtime tracer ( #1974 )
2022-11-18 10:53:42 +08:00
0da1d00399
[autoparallel] support distributed dataloader option ( #1906 )
...
* [autoparallel] support distributed dataloader option
* update output handler to support ddp dataloader
* poish code
2022-11-17 20:11:53 +08:00
6630d45546
[autoparallel] Add alpha beta ( #1973 )
...
* Add alpha beta
* Fix test
* Fix test
2022-11-17 16:01:14 +08:00
cc0ed7cf33
[Gemini] ZeROHookV2 -> GeminiZeROHook ( #1972 )
2022-11-17 14:43:49 +08:00
f8a7148dec
[kernel] move all symlinks of kernel to `colossalai._C` ( #1971 )
2022-11-17 13:42:33 +08:00
7e24b9b9ee
[Gemini] clean no used MemTraceOp ( #1970 )
2022-11-17 13:41:54 +08:00
7c7921f71b
[autoparallel] add torch.nn.ReLU metainfo ( #1868 )
...
* [fx] metainfo class for auto parallel
* [fx] add unit test for linear metainfo
* [fx] fix bwd param for linear
* [fx] modify unit test
* [fx] modify unit test
* [fx] modify import
* [fx] modify import
* [fx] modify import
* [fx] move meta profiler to auto parallel
* [fx] add conv metainfo class
* [fx] restore profiler
* [fx] restore meta profiler
* [autoparallel] modify unit test
* [fx] modify unit test
* [autoparallel] add batchnorm metainfo class
* [autoparallel] fix batchnorm unit test function declaration
* [fx] restore profiler
* [fx] add relu metainfo class
* [fx] restore profiler
* [autoparallel] modify metainfo input
2022-11-16 23:12:31 +08:00
8c66a1d0aa
[polish] remove useless file _mem_tracer_hook.py ( #1963 )
2022-11-16 15:55:10 +08:00
c4739a725a
[Gemini] polish memstats collector ( #1962 )
2022-11-16 15:45:57 +08:00
fea3cb661c
[autoparallel] support addmm in tracer and solver ( #1961 )
...
* [fx] patch addmm
* [autoparallel] support addmm in tracer and solver
2022-11-16 14:59:18 +08:00
f7e276fa71
[Gemini] add GeminiAdamOptimizer ( #1960 )
2022-11-16 14:44:28 +08:00
7066dfbf82
[zero] fix memory leak for zero2 ( #1955 )
2022-11-16 11:43:24 +08:00
52c6ad26e0
[ColoTensor] reconfig ColoInitContext, decouple default_pg and default_dist_spec. ( #1953 )
2022-11-15 16:24:16 +08:00
598d456d0e
fixed logger
2022-11-15 16:00:07 +08:00
6877121377
updated flash attention api
2022-11-15 15:25:39 +08:00
36c0f3ea5b
[autoparallel] remove redundancy comm node ( #1893 )
2022-11-15 10:53:41 +08:00
e52f9d9109
[tensorparallel] fixed tp layers ( #1938 )
2022-11-14 17:34:03 +08:00
9f4fb3f28a
[ColoTensor] ColoInitContext initialize parameters in shard mode. ( #1937 )
2022-11-14 16:05:09 +08:00
d5c5bc219e
[SC] add GPT example for auto checkpoint ( #1889 )
...
* [sc] SC tutorial for auto checkpoint
* [sc] polish examples
* [sc] polish readme
* [sc] polish readme and help information
* [sc] polish readme and help information
2022-11-11 23:17:25 +08:00
14a0b18305
[NFC] polish colossalai/amp/naive_amp/__init__.py code style ( #1905 )
2022-11-11 17:49:18 +08:00
6e51d296f0
[zero] migrate zero1&2 ( #1878 )
...
* add zero1&2 optimizer
* rename test ditectory
* rename test files
* change tolerance in test
2022-11-11 09:26:40 +08:00
cc55ff0aa4
[autoparallel] user-friendly API for CheckpointSolver. ( #1879 )
...
Merge for SC tutorial
2022-11-10 20:59:28 +08:00
448248b27c
[fx] metainfo_trace as an API. ( #1873 )
...
* [fx] metainfo_trace as an API.
* [fx] add return.
2022-11-10 20:58:37 +08:00
986f8cbaa7
[inference] overlap comm and compute in Linear1D_Row when stream_chunk_num > 1 ( #1876 )
2022-11-10 17:36:42 +08:00
1b494ad73c
[autoparallel] fix linear logical convert issue ( #1857 )
2022-11-10 17:19:22 +08:00
c2947dadf1
[inference] streaming Linear 1D Row inference ( #1874 )
2022-11-10 17:03:21 +08:00
e6ec99d389
[utils] fixed lazy init context ( #1867 )
2022-11-10 15:17:20 +08:00
653b0a620e
added skip_bias_add for non-tp linear
2022-11-09 15:41:08 +08:00
94329fc139
[NFC] polish colossalai/amp/apex_amp/__init__.py code style ( #1853 )
2022-11-09 14:49:42 +08:00
1559a09fb7
[NFC] polish amp.naive_amp.grad_scaler code style
2022-11-09 13:38:15 +08:00
72c9448920
[NFC] polish colossalai/auto_parallel/tensor_shard/deprecated/op_handler/operator_handler.py code style ( #1845 )
2022-11-09 12:08:47 +08:00
b25030cc07
[NFC] polish ./colossalai/amp/torch_amp/__init__.py code style ( #1836 )
2022-11-09 12:08:47 +08:00
95ac4f88ea
[NFC] polish colossalai/auto_parallel/tensor_shard/deprecated/op_handler/conv_handler.py code style ( #1829 )
...
Co-authored-by: siqi <siqi@siqis-MacBook-Pro.local>
2022-11-09 12:08:47 +08:00
5da03c936d
[NFC] polish colossalai/amp/torch_amp/_grad_scaler.py code style ( #1823 )
...
Co-authored-by: Ziyue Jiang <ziyue.jiang@gmail.com>
2022-11-09 12:08:47 +08:00
399f84d8f6
[NFC] polish colossalai/amp/naive_amp/_fp16_optimizer.py code style ( #1819 )
2022-11-09 12:08:47 +08:00
9623ec1b02
[NFC] polish colossalai/amp/naive_amp/_utils.py code style ( #1816 )
...
* [NFC] polish colossalai/nn/metric/accuracy_2p5d.py code style (#1714 )
* [NFC] polish colossalai/zero/sharded_param/__init__.py code style
* [NFC] polish colossalai/amp/naive_amp/_utils.py code style
Co-authored-by: shenggan <csg19971016@gmail.com>
Co-authored-by: ric <mkkt_bkkt@mail.ustc.edu.cn>
2022-11-09 12:08:47 +08:00
3c3714fc2a
[NFC] polish strategies_constructor.py code style ( #1806 )
2022-11-09 12:08:47 +08:00
3ce4463fe6
[utils] remove lazy_memory_allocate from ColoInitContext ( #1844 )
2022-11-09 11:50:33 +08:00
fba34efb5a
version to 0.1.11rc2 ( #1832 )
2022-11-08 17:25:15 +08:00
49216d7ab1
[autoparallel] fix bugs caused by negative dim key ( #1808 )
...
* [autoparallel] fix bugs caused by negative dim key
* fix import error
* fix matmul test issue
* fix unit test issue
2022-11-08 17:03:50 +08:00
4268ae017b
[kernel] added jit warmup ( #1792 )
2022-11-08 16:22:23 +08:00
f6032ddb17
[autoparallel] fix bias addition module ( #1800 )
2022-11-08 16:21:25 +08:00
cd5a0d56fa
[Gemini] make gemini usage simple ( #1821 )
2022-11-08 15:53:13 +08:00
99870726b1
[CheckpointIO] a uniform checkpoint I/O module ( #1689 )
2022-11-08 15:15:13 +08:00
629172b319
[autoparallel] add batch norm metainfo ( #1815 )
...
* [fx] metainfo class for auto parallel
* [fx] add unit test for linear metainfo
* [fx] fix bwd param for linear
* [fx] modify unit test
* [fx] modify unit test
* [fx] modify import
* [fx] modify import
* [fx] modify import
* [fx] move meta profiler to auto parallel
* [fx] add conv metainfo class
* [fx] restore profiler
* [fx] restore meta profiler
* [autoparallel] modify unit test
* [fx] modify unit test
* [autoparallel] add batchnorm metainfo class
* [autoparallel] fix batchnorm unit test function declaration
* [fx] restore profiler
2022-11-08 15:05:26 +08:00
441d584e4a
[fx] add a symbolic_trace api. ( #1812 )
...
* [fx] add a symbolic_trace api.
* [fx] fix import errors.
2022-11-08 13:59:20 +08:00
e0da01ea71
[hotfix] fix build error when torch version >= 1.13 ( #1803 )
2022-11-08 09:40:24 +08:00
9639ea88fc
[kernel] more flexible flashatt interface ( #1804 )
2022-11-07 17:02:09 +08:00
20e255d4e8
MemStatsCollectorStatic ( #1765 )
2022-11-07 16:49:03 +08:00
327d07c44a
[autoparallel] add conv metainfo class for auto parallel ( #1796 )
...
* [fx] metainfo class for auto parallel
* [fx] add unit test for linear metainfo
* [fx] fix bwd param for linear
* [fx] modify unit test
* [fx] modify unit test
* [fx] modify import
* [fx] modify import
* [fx] modify import
* [fx] move meta profiler to auto parallel
* [fx] add conv metainfo class
* [fx] restore profiler
* [fx] restore meta profiler
* [autoparallel] modify unit test
* [fx] modify unit test
2022-11-07 16:15:35 +08:00
501a9e9cd2
[hotfix] polish flash attention ( #1802 )
2022-11-07 14:30:22 +08:00
218c75fd9d
[NFC] polish type hint for shape consistency ( #1801 )
...
* [NFC] polish type hint for shape consistency
* polish code
* polish code
2022-11-07 14:13:03 +08:00
c248800359
[kernel] skip tests of flash_attn and triton when they are not available ( #1798 )
2022-11-07 13:41:13 +08:00
e34e850a4c
[autoparallel]add essential CommActions for broadcast oprands ( #1793 )
2022-11-04 18:36:42 +08:00
05ce3d369f
[fx] Add linear metainfo class for auto parallel ( #1783 )
...
* [fx] metainfo class for auto parallel
* [fx] add unit test for linear metainfo
* [fx] fix bwd param for linear
* [fx] modify unit test
* [fx] modify unit test
* [fx] modify import
* [fx] modify import
* [fx] modify import
* [fx] move meta profiler to auto parallel
2022-11-04 10:55:09 +08:00
e8a9bebc87
[autoparallel] refactor and add rotorc. ( #1789 )
...
* [autoparallel] refactor and add rotorc.
* [autoparallel] refactor and add rotorc.
2022-11-03 12:32:51 +08:00
2c4c7b3618
[autoparallel] add getattr handler ( #1767 )
...
* [autoparallel] add getattr haandler
* polish code
* add extra processes for Parameters
* add unit test for param resharding cost
* add docstring and polish test
2022-11-03 12:31:33 +08:00
c6a1a62636
[hotfix] fix zero's incompatibility with checkpoint in torch-1.12 ( #1786 )
...
* [hotfix] fix zero's incompatibility with checkpoint in torch-1.12
* [zero] add cpu shard init
* [zero] add tiny example test
* [colo_tensor] fix bugs for torch-1.11
2022-11-02 16:11:34 +08:00
0b8161fab8
updated tp layers
2022-11-02 12:19:38 +08:00
cb5a587e9a
[hotfix] polish chunk import ( #1787 )
2022-11-02 12:10:52 +08:00
e859380bf7
[fx] support module with bias addition ( #1780 )
...
* [autoparallel] refactor tracer to fix bias addition issue
* [fx] support module with bias addition
* create bias_addition_module
* refactor file structure
* polish code
* fix unit test
2022-11-01 22:53:51 +08:00
f3f19a5c47
[autoparallel] added matmul handler ( #1763 )
...
* [autoparallel] added matmul handler
* polish code
2022-11-01 15:14:53 +08:00
4df0194976
[Pipeline]Adapt to Pipelinable OPT ( #1782 )
2022-11-01 14:18:50 +08:00
27de252334
[autoparallel] fix conv handler numerical test ( #1771 )
2022-11-01 10:43:44 +08:00
1e88811c7a
[autoparallel] move ckpt solvers to autoparallel folder / refactor code ( #1764 )
...
* [autoparallel] first move.
* [autoparallel] add solver rotor.
* [autoparallel] add ckpt solvers.
* [autoparallel] modify codegen.
* [fx] fix annotation in test.
* [fx] remove check.
* [autoparallel] polish docstring.
* [fx] refactor MetaTensor.
2022-11-01 10:43:15 +08:00
f34dab4270
[compatibility] ChunkMgr import error ( #1772 )
2022-10-28 14:48:54 +08:00
b0f7c8bde8
[autoparallel] update CommSpec to CommActions ( #1768 )
...
* [autoparallel] update CommSpec to CommActions
* polish code
2022-10-28 09:57:43 +08:00
b4cc59b61e
[autoparallel] add numerical test for node strategies ( #1760 )
...
* [autoparallel] add numerical test for node strategies
* polish code
* polish code
2022-10-27 10:42:54 +08:00
25952b67d7
[feat] add flash attention ( #1762 )
2022-10-26 16:15:52 +08:00
0584654c79
[fx] refactor memory utils and extend shard utils. ( #1754 )
...
* [fx] change memory.py to memory_utils.py.
* [fx] add shard utils.
* [fx] fix import.
* [fx] check code style.
* [fx] add comment.
* [autoparallel] first move.
* [fx] add time computations.
2022-10-26 14:24:41 +08:00
63f250bbd4
fix file name ( #1759 )
...
Co-authored-by: Ziyue Jiang <ziyue.jiang@gmail.com>
2022-10-25 16:48:48 +08:00
314d8c497f
[autoparallel] refactor the runtime apply pass and add docstring to passes ( #1757 )
...
* [autoparallel] refactor the runtime apply pass and add doc string to passes
* fix unit test
* polish
2022-10-25 14:32:22 +08:00
f9a613d660
[autoparallel] added binary elementwise node handler ( #1758 )
...
* [autoparallel] added binary elementwise node handler
* polish code
2022-10-25 14:32:01 +08:00
d2fc067231
[autoparallel] fix param hook issue in transform pass ( #1755 )
2022-10-24 13:13:38 +08:00
262652c8bc
[autoparallel] added addbmm handler ( #1751 )
2022-10-21 18:55:48 +08:00
980ed21723
[autoparallel] shard param and buffer as expected ( #1753 )
...
* [autoparallel] shard param and buffer as expected
* fix unit test issue
2022-10-21 15:45:13 +08:00
cdb7d5e7d2
[hotfix] autoparallel unit test ( #1752 )
2022-10-20 19:51:38 +08:00
a4ce180e85
[autoparallel] add sequential order to communication actions ( #1735 )
2022-10-20 18:48:18 +08:00
474111ecb5
[autoparallel] fixed wrong sharding strategy in conv handler ( #1747 )
...
* [autoparallel] fixed wrong sharding strategy in conv handler
* polish code
2022-10-20 16:12:39 +08:00
8b8937d901
[autoparallel] fixed wrong generated strategy for dot op ( #1746 )
...
* [autoparallel] fixed wrong generated strategy for dot op
* polish code
2022-10-20 15:18:16 +08:00
993b8875b6
[autoparallel] handled illegal sharding strategy in shape consistency ( #1744 )
...
* [autoparallel] handled illegal sharding strategy in shape consistency
* polish code
2022-10-20 12:06:25 +08:00
88a79814fb
[autoparallel] handled illegal strategy in node handler ( #1743 )
...
* [autoparallel] handled illegal strategy in node handler
* polish code
2022-10-19 17:08:52 +08:00
30874f1692
[fx/profiler] debug the fx.profiler / add an example test script for fx.profiler ( #1730 )
...
* [fx/profiler] add test.
* [fx] fix file names.
* [fx] add docstring and comment.
* [fx] polish profiler.py.
* [fx] fix import errors.
* [fx] fix profiler.
* [fx] fix names.
2022-10-19 14:24:51 +08:00
eee84908d4
[autoparallel] handled illegal sharding strategy ( #1728 )
...
* [autoparallel] handled illegal sharding strategy
* polish code
2022-10-19 12:53:06 +08:00
23703c9dd6
[NFC] polish colossalai/nn/metric/_utils.py code style ( #1727 )
2022-10-19 12:20:51 +08:00
7e62af28a0
[NFC] polish accuracy_2d.py code style ( #1719 )
2022-10-19 12:20:51 +08:00
730f88f8e1
[NFC] polish _checkpoint_hook.py code style ( #1722 )
2022-10-19 12:20:51 +08:00
ea961d8fd1
[NFC] polish colossalai/zero/sharded_param/__init__.py code style ( #1717 )
...
Co-authored-by: ric <mkkt_bkkt@mail.ustc.edu.cn>
2022-10-19 12:20:51 +08:00
2b49ca80a3
[NFC] polish colossalai/nn/lr_scheduler/linear.py code style ( #1716 )
2022-10-19 12:20:51 +08:00
e1d780030d
[NFC] polish colossalai/nn/metric/accuracy_2p5d.py code style ( #1714 )
2022-10-19 12:20:51 +08:00
d373e67b99
[hotfix] resharding cost issue ( #1742 )
2022-10-19 11:33:43 +08:00
24e84eba60
upgrade version to 0.1.11rc1 ( #1739 )
2022-10-19 11:26:00 +08:00
d2e0e39c9d
[release] update to v0.1.11 ( #1736 )
2022-10-19 00:28:00 +08:00
f69f9bf223
[zero] add chunk init function for users ( #1729 )
...
* add chunk manager init function
* fix unit tests
* add comment
* add flush=True
2022-10-18 16:31:22 +08:00
51b89d2202
[autoparallel] runtime_backward_apply ( #1720 )
2022-10-18 10:44:58 +08:00
393f594051
[fx/meta/rpc] move _meta_registration.py to fx folder / register fx functions with compatibility checks / remove color debug ( #1710 )
...
* [fx] move meta registration
* [fx] fix tests.
* [fx] fix test.
* [fx] fix.
* [meta] refactor meta registration.py.
* [fx] add compatibility descriptions.
* [fx] polish import.
* [fx] add a decorator.
* [fx] fix tests.
* [fx] remove print.
* [fx] edit raise error.
* [fx] edit raise error.
* [fx] add type hint.
* [fx] fix import in experimental.
* [rpc] remove color debug.
* [meta] fix naming.
2022-10-18 10:44:23 +08:00
845ff4a47a
[autoparallel] resnet block runtime apply ( #1709 )
...
* [autoparallel] resnet block runtime apply
* seperate buffer and parameter in MemoryCost
* polish code
* add comments and todos
* fix test issue
2022-10-17 13:37:38 +08:00
22a115406b
[autoparallel] fixed broken node handler tests ( #1708 )
2022-10-14 18:25:59 +08:00
1468e4bcfc
[zero] add constant placement policy ( #1705 )
...
* fixes memory leak when paramter is in fp16 in ZeroDDP init.
* bans chunk releasement in CUDA. Only when a chunk is about to offload, it is allowed to release.
* adds a constant placement policy. With it, users can allocate a reserved caching memory space for parameters.
2022-10-14 17:53:16 +08:00
5f41463a76
add optimizer README for tutorials ( #1707 )
2022-10-14 09:10:18 +00:00
6c331a5a09
[autoparallel] refactored the autoparallel module for organization ( #1706 )
...
* [autoparallel] refactored the autoparallel module for organization
* polish code
2022-10-14 13:27:00 +08:00
91cd34e6e0
[unittest] added doc for the pytest wrapper ( #1704 )
2022-10-14 10:56:17 +08:00
451cd72dea
[autoparallel] adapt runtime passes ( #1703 )
...
* [autoparallel] adapt runtime passes v2
* polish code
2022-10-14 10:14:07 +08:00
Boyuan Yao