Commit Graph

306 Commits (af952673f758c71126b27de8b32bdf5df8f74b69)

Author SHA1 Message Date
HELSON dddacd2d2c
[hotfix] add norm clearing for the overflow step (#2416) 2023-01-10 15:43:06 +08:00
HELSON 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
Frank Lee 551cafec14
[doc] updated kernel-related optimisers' docstring (#2385)
* [doc] updated kernel-related optimisers' docstring

* polish doc
2023-01-09 17:13:53 +08:00
eric8607242 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
Frank Lee 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
HELSON 48d33b1b17
[gemini] add get static torch model (#2356) 2023-01-06 13:41:19 +08:00
Jiarui Fang 16cc8e6aa7
[builder] MOE builder (#2277) 2023-01-03 20:29:39 +08:00
zbian e94c79f15b improved allgather & reducescatter for 3d 2023-01-03 17:46:08 +08:00
Jiarui Fang af32022f74
[Gemini] fix the convert_to_torch_module bug (#2269) 2023-01-03 15:55:35 +08:00
HELSON 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
Jiarui Fang 355ffb386e
[builder] unified cpu_optim fused_optim inferface (#2190) 2022-12-23 20:57:41 +08:00
Jiarui Fang 9587b080ba
[builder] use runtime builder for fused_optim (#2189) 2022-12-23 17:07:03 +08:00
Jiarui Fang d42afd30f8
[builder] runtime adam and fused_optim builder (#2184) 2022-12-23 14:14:21 +08:00
Tongping Liu ab54fed292
[hotfix] add kwargs for colo_addmm (#2171) 2022-12-22 13:25:30 +08:00
アマデウス 622f863291
[hotfix] Jit type hint #2161 (#2164) 2022-12-22 10:17:03 +08:00
Jiarui Fang 2827f41898
[Gemini] GeminiDPP convert to PyTorch Module. (#2151) 2022-12-20 10:19:36 +08:00
Jiarui Fang bdef9dfdbe
[NFC] remove useless graph node code (#2150) 2022-12-20 00:33:58 +08:00
Jiarui Fang 9214d1fe28
[Gemini] chunk init using runtime visited param order (#2115) 2022-12-12 18:06:16 +08:00
HELSON 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
Jiarui Fang e5aa8333e4
[NFC] update chunk manager API (#2119) 2022-12-12 16:57:22 +08:00
Jiarui Fang e99edfcb51
[NFC] polish comments for Chunk class (#2116) 2022-12-12 15:39:31 +08:00
HELSON 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
Jiarui Fang 1f99205827
[Gemini] remove static tracer (#2083) 2022-12-06 12:53:58 +08:00
Jiarui Fang b3b89865e2
[Gemini] ParamOpHook -> ColoParamOpHook (#2080) 2022-12-05 17:11:06 +08:00
HELSON e37f3db40c
[gemini] add arguments (#2046)
* [zero] fix testing parameters

* [gemini] add arguments

* add docstrings
2022-11-30 16:40:13 +08:00
Jiarui Fang 96134e7be3
[hotfix] add bert test for gemini fwd bwd (#2035) 2022-11-29 11:19:52 +08:00
Jiarui Fang 8daf1b4db1
[Gemini] patch for supporting orch.add_ function for ColoTensor (#2003) 2022-11-25 20:06:35 +08:00
Jiarui Fang a2d3266648
[hotfix] make Gemini work for conv DNN (#1998) 2022-11-22 14:52:36 +08:00
Jiarui Fang cc0ed7cf33
[Gemini] ZeROHookV2 -> GeminiZeROHook (#1972) 2022-11-17 14:43:49 +08:00
ver217 f8a7148dec
[kernel] move all symlinks of kernel to `colossalai._C` (#1971) 2022-11-17 13:42:33 +08:00
Jiarui Fang f7e276fa71
[Gemini] add GeminiAdamOptimizer (#1960) 2022-11-16 14:44:28 +08:00
アマデウス e52f9d9109
[tensorparallel] fixed tp layers (#1938) 2022-11-14 17:34:03 +08:00
Jiarui Fang 986f8cbaa7
[inference] overlap comm and compute in Linear1D_Row when stream_chunk_num > 1 (#1876) 2022-11-10 17:36:42 +08:00
Jiarui Fang c2947dadf1
[inference] streaming Linear 1D Row inference (#1874) 2022-11-10 17:03:21 +08:00
zbian 653b0a620e added skip_bias_add for non-tp linear 2022-11-09 15:41:08 +08:00
アマデウス 4268ae017b
[kernel] added jit warmup (#1792) 2022-11-08 16:22:23 +08:00
Jiarui Fang cd5a0d56fa
[Gemini] make gemini usage simple (#1821) 2022-11-08 15:53:13 +08:00
Zihao 20e255d4e8
MemStatsCollectorStatic (#1765) 2022-11-07 16:49:03 +08:00
HELSON 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
kurisusnowdeng 0b8161fab8 updated tp layers 2022-11-02 12:19:38 +08:00
Sze-qq 23703c9dd6 [NFC] polish colossalai/nn/metric/_utils.py code style (#1727) 2022-10-19 12:20:51 +08:00
Ofey Chan 7e62af28a0 [NFC] polish accuracy_2d.py code style (#1719) 2022-10-19 12:20:51 +08:00
yuxuan-lou 2b49ca80a3 [NFC] polish colossalai/nn/lr_scheduler/linear.py code style (#1716) 2022-10-19 12:20:51 +08:00
shenggan e1d780030d [NFC] polish colossalai/nn/metric/accuracy_2p5d.py code style (#1714) 2022-10-19 12:20:51 +08:00
HELSON 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
binmakeswell 5f41463a76
add optimizer README for tutorials (#1707) 2022-10-14 09:10:18 +00:00
Jiarui Fang 21962e1593
[embedding] rename FreqAwareEmbedding -> CachedEmbedding (#1699) 2022-10-13 22:22:27 +08:00
Jiarui Fang 363fc2861a
[embeddings] more detailed timer (#1692) 2022-10-12 12:01:21 +08:00
jim e5ab6be72e
[hotfix[ fix colotensor.type() raise NotImplementedError (#1682) 2022-10-10 10:13:31 +08:00
HELSON b28991dd0a
[feature] A new ZeRO implementation (#1644) 2022-10-09 09:18:51 +08:00
Jiarui Fang c638bec028
[embedding] polish async copy (#1657) 2022-09-27 14:37:03 +08:00
Jiarui Fang 988570e4a6
[embedding] add more detail profiling (#1656) 2022-09-27 13:43:59 +08:00
Jiarui Fang e1f97fd2b8
[embedding] print profiling results (#1654) 2022-09-27 12:50:33 +08:00
Jiarui Fang 04443605a5
[embedding] non-blocking cpu-gpu copy (#1647) 2022-09-26 14:57:57 +08:00
CsRic 0767f67a0f
[embedding] isolate cache_op from forward (#1645)
Co-authored-by: ric <mkkt_bkkt@mail.ustc.edu.cn>
2022-09-26 11:18:59 +08:00
Jiarui Fang c5d39215f6
Revert "[feature] new zero implementation (#1623)" (#1643)
This reverts commit 5be118f405.
2022-09-26 10:06:03 +08:00
HELSON 5be118f405
[feature] new zero implementation (#1623) 2022-09-24 19:58:18 +08:00
Jiarui Fang e57df80325
[embeddings] cache option (#1635) 2022-09-23 16:40:18 +08:00
HELSON a088022efc
[moe] fix moe bugs (#1633) 2022-09-23 15:33:57 +08:00
HELSON f7f2248771
[moe] fix MoE bugs (#1628)
* remove forced FP32 modules

* correct no_shard-contexts' positions
2022-09-22 13:56:30 +08:00
Jiarui Fang 38c68b5b9a
[embedding] rollback for better FAW performance (#1625) 2022-09-22 11:16:25 +08:00
Jiarui Fang 504ff1d101
[embeddings] use cache_ratio instead of cuda_row_num (#1611) 2022-09-20 14:33:04 +08:00
Jiarui Fang a19eb80998
[embedding] updates some default parameters 2022-09-15 15:45:17 +08:00
CsRic f3403ff98e
[embeddings] add already_split_along_rank flag for tablewise mode (#1584) 2022-09-13 10:50:34 +08:00
Sze-qq 2144cbae8c [NFC] polish colossalai/nn/lr_scheduler/multistep.py code style (#1572) 2022-09-08 22:11:04 +08:00
superhao1995 e4bf7ae667 [NFC] polish colossalai/nn/lr_scheduler/torch.py code style (#1571)
Co-authored-by: Research <research@soccf-snr3-017.comp.nus.edu.sg>
2022-09-08 22:11:04 +08:00
Jiatong Han 3263cdf57f [NFC] polish colossalai/nn/parallel/data_parallel.py code style (#1570)
Co-authored-by: JThh <jiatong.han@u.nus.edu>
2022-09-08 22:11:04 +08:00
DouJS f586887a90 [NFC] polish colossalai/nn/layer/colossalai_layer/dropout.py code style (#1568) 2022-09-08 22:11:04 +08:00
BigOneLiXiaoMing 0c4c9aa6e0 [NFC] polish colossalai/nn/_ops/embedding.py code style (#1561) 2022-09-08 22:11:04 +08:00
Ofey Chan 7cc052f6c0 [NFC] polish colossalai/nn/layer/colossalai_layer/linear.py (#1556) 2022-09-08 22:11:04 +08:00
yuxuan-lou 413f9c19f4 [NFC] polish colossalai/nn/_ops/layernorm.py code style (#1555) 2022-09-08 22:11:04 +08:00
shenggan 8edb777cc2 [NFC] polish colossalai/nn/loss/loss_2p5d.py code style (#1553) 2022-09-08 22:11:04 +08:00
Maruyama_Aya bd2d789832 [NFC] polish colossalai/nn/_ops/embedding_bag.py code style (#1552) 2022-09-08 22:11:04 +08:00
binmakeswell 73e9eb13b7 [NFC] polish colossalai/nn/lr_scheduler/cosine.py code style 2022-09-08 22:11:04 +08:00
CsRic a389ac4ec9
[embedding] cache_embedding small improvement (#1564) 2022-09-08 16:41:19 +08:00
ver217 10dd8226b1
add gather_output for VocabParallelClassifier1D (#1569) 2022-09-08 16:40:56 +08:00
ver217 ae71036cd2
[utils] refactor parallel layers checkpoint and bcast model on loading checkpoint (#1548)
* refactor parallel layer

* broadcast rank0 model after load ckpt
2022-09-06 20:18:35 +08:00
Jiarui Fang 64169f3e8f
[embedding] polish parallel embedding tablewise (#1545) 2022-09-06 10:41:20 +08:00
CsRic 964123ae0f
[embedding] freq_aware_embedding: add small functions for caller application (#1537) 2022-09-05 15:12:53 +08:00
Jiarui Fang 521078ffc9
[embedding] fix a bug in table wise sharding (#1538) 2022-09-02 15:48:35 +08:00
Jiarui Fang 87134524fd
[embedding] tablewise sharding polish (#1535) 2022-09-02 11:09:37 +08:00
CsRic 5156d5b4f8
[embedding] add tablewise sharding for FAW (#1526) 2022-09-01 17:55:41 +08:00
Jiarui Fang 4537d39df9
[doc] docstring for FreqAwareEmbeddingBag (#1525) 2022-08-31 13:52:30 +08:00
Jiarui Fang 9a9ef65313
[FAW] cpu caching operations (#1520) 2022-08-30 14:50:02 +08:00
Jiarui Fang af5438caa2
[FAW] refactor reorder() for CachedParamMgr (#1514) 2022-08-29 14:22:07 +08:00
Jiarui Fang 9feee6d06b
[FAW] LFU initialize with dataset freq (#1513) 2022-08-29 12:52:53 +08:00
CsRic 1b8fee8e9c
[FAW] shrink freq_cnter size (#1509) 2022-08-29 11:44:55 +08:00
Jiarui Fang ba61109b6c
[FAW] remove code related to chunk (#1501) 2022-08-26 14:23:30 +08:00
Jiarui Fang d5085bb317
[FAW] add more docs and fix a warning (#1500) 2022-08-26 14:10:21 +08:00
CsRic 0ed2f46131
[FAW] FAW embedding use LRU as eviction strategy intialized with dataset stats (#1494) 2022-08-26 11:24:12 +08:00
CsRic b8d0e39eaf
[FAW] LFU cache for the FAW 2022-08-25 13:08:46 +08:00
Jiarui Fang cde7b8a5b8
[FAW] init an LFU implementation for FAW (#1488) 2022-08-24 17:37:22 +08:00
Geng Zhang 0aad53c62b
[FCE] update interface for frequency statistics in FreqCacheEmbedding (#1462) 2022-08-23 17:38:24 +08:00
Jiarui Fang a1476ea882
[NFC] polish doc style for ColoTensor (#1457) 2022-08-16 09:21:05 +08:00
ver217 367c615818
fix nvme docstring (#1450) 2022-08-12 18:01:02 +08:00
Geng Zhang 9f3eed66eb
[FAW] reorganize the inheritance struct of FreqCacheEmbedding (#1448) 2022-08-12 15:55:46 +08:00
Frank Lee ae1b58cd16
[tensor] added linear implementation for the new sharding spec (#1416)
* [tensor] added linear implementation for the new sharding spec

* polish code
2022-08-12 11:33:09 +08:00
Jiarui Fang 30b4dd17c0
[FAW] export FAW in _ops (#1438) 2022-08-11 13:43:24 +08:00
Jiarui Fang c9427a323f
hotfix #1434 (#1437) 2022-08-11 13:14:25 +08:00
Jiarui Fang 10b3df65c8
[FAW] move coloparam setting in test code. (#1429) 2022-08-10 14:31:53 +08:00