Commit Graph

  • 02949ac55b rm disable logger jiaruifang 2022-03-25 11:14:41 +0800
  • 495fd20d7e [zero] fix bug in zero context init test jiaruifang 2022-03-25 11:12:40 +0800
  • 6970d4ab60 polish jiaruifang 2022-03-25 10:43:33 +0800
  • fe243d10b0 polish jiaruifang 2022-03-25 10:41:48 +0800
  • 4c32e8e12b [zero] show model data cuda memory usage after zero context init. jiaruifang 2022-03-25 09:41:59 +0800
  • 77c17281be style number1roy 2022-03-25 00:10:04 +0800
  • a2e61d61d4
    [zero] zero init ctx enable rm_torch_payload_on_the_fly (#512) ver217 2022-03-24 23:44:00 +0800
  • eb99cf4873 sty;e number1roy 2022-03-24 22:39:50 +0800
  • dfd1981a71 style number1roy 2022-03-24 22:19:47 +0800
  • eefb8fe863
    Update README.md fastalgo 2022-03-24 21:44:07 +0800
  • ab30de3aa6 Merge branch 'main' into feature/docs number1roy 2022-03-24 21:43:53 +0800
  • 382ecf1c23 conflict solve number1roy 2022-03-24 21:41:24 +0800
  • 8d2cfed6be style number1roy 2022-03-24 18:43:36 +0800
  • 81145208d1
    [install] run with out rich (#513) Jiarui Fang 2022-03-24 17:39:50 +0800
  • 0f2d219162
    [MOE] add MOEGPT model (#510) HELSON 2022-03-24 17:39:21 +0800
  • d5ef4ba8ae run with out rich jiaruifang 2022-03-24 17:35:39 +0800
  • 57ba792424 add MOEGPT model 1SAA 2022-03-24 16:24:37 +0800
  • 057b6347ab polish docstr ver217 2022-03-24 17:23:06 +0800
  • bca0c49a9d
    [zero] use colo model data api in optimv2 (#511) Jiarui Fang 2022-03-24 17:19:34 +0800
  • c79e5893e5 [zero] use colo model data api in optimv2 jiaruifang 2022-03-24 17:17:17 +0800
  • bb78a1fea8 enable rm_torch_payload_on_the_fly ver217 2022-03-24 17:11:42 +0800
  • 9330be0f3c
    [memory] set cuda mem frac (#506) Jiarui Fang 2022-03-24 16:57:13 +0800
  • 97933b6710
    [devops] recover tsinghua pip source due to proxy issue (#509) Frank Lee 2022-03-24 16:11:49 +0800
  • bc1f206b92 google style number1roy 2022-03-24 15:41:29 +0800
  • 949e900e46 removed useless code lclgy 2022-03-24 15:30:53 +0800
  • f6cdb0efb2 [devops] recover tsinghua pip source due to proxy issue FrankLeeeee 2022-03-24 14:16:08 +0800
  • d15a59399c google style number1roy 2022-03-24 15:18:06 +0800
  • 2ab54a1e0a google style number1roy 2022-03-24 14:59:42 +0800
  • fe8c647cb5 google style number1roy 2022-03-24 14:52:01 +0800
  • 422e2ba375 Added loss scale test for cpu adam test lclgy 2022-03-24 14:51:36 +0800
  • bdcbe2ac9d polish code jiaruifang 2022-03-24 14:45:04 +0800
  • 63a8176109 google style number1roy 2022-03-24 14:40:28 +0800
  • 21fecb08e8 Merge branch 'main' of github.com:hpcaitech/ColossalAI into jiaruifang/set_cuda_mem_frac jiaruifang 2022-03-24 14:32:23 +0800
  • 0035b7be07
    [memory] add model data tensor moving api (#503) Jiarui Fang 2022-03-24 14:29:41 +0800
  • 168ec8d6c8 fix the hybrid dtype support of fused adam, added unit test lclgy 2022-03-24 14:17:32 +0800
  • 65ad47c35c
    [devops] remove tsinghua source for pip (#507) Frank Lee 2022-03-24 14:12:02 +0800
  • 7231b22df5 [devops] remove tsinghua source for pip FrankLeeeee 2022-03-24 14:02:04 +0800
  • d94109bbed google style number1roy 2022-03-24 14:05:42 +0800
  • d8b7d2caca [memory] set cuda mem fraction jiaruifang 2022-03-24 14:03:53 +0800
  • 44f7bcb277
    [devops] remove tsinghua source for pip (#505) Frank Lee 2022-03-24 14:03:05 +0800
  • 6fac6a0554 [devops] remove tsinghua source for pip FrankLeeeee 2022-03-24 14:02:04 +0800
  • 4733d5eb2e polish code jiaruifang 2022-03-24 12:43:45 +0800
  • af56c1d024
    fix discussion button in issue template (#504) binmakeswell 2022-03-24 12:25:00 +0800
  • 8c3f45fa9f polish jiaruifang 2022-03-24 12:20:42 +0800
  • 4b4160687e fix discussion button in issue template binmakeswell 2022-03-24 12:19:59 +0800
  • 10cece3c1c polish jiaruifang 2022-03-24 10:59:29 +0800
  • 8ae65b85cd Merge branch 'main' of https://github.com/hpcaitech/ColossalAI into jiaruifang/col_mem_move jiaruifang 2022-03-24 10:56:13 +0800
  • d2102fc232 polish code jiaruifang 2022-03-24 10:55:27 +0800
  • bb4c253f85 add model data tensor moving api jiaruifang 2022-03-24 10:52:22 +0800
  • a445e118cf
    [polish] polish singleton and global context (#500) Jiarui Fang 2022-03-23 18:03:39 +0800
  • 8eacb56b52 rm useless import jiaruifang 2022-03-23 17:03:54 +0800
  • bffa9a0112 polish code jiaruifang 2022-03-23 16:59:56 +0800
  • b9e5af1edd [polish] polish singleton usage jiaruifang 2022-03-23 16:58:24 +0800
  • 36ffee8ebc polish jiaruifang 2022-03-23 15:33:02 +0800
  • 5c2ccb9758 [WIP] demonstrate a new way to write parallel op with customized tensor structure. jiaruifang 2022-03-23 15:23:04 +0800
  • 9ec1ce6ab1
    [zero] sharded model support the reuse of fp16 shard (#495) ver217 2022-03-23 14:59:59 +0800
  • 31a126550b polish code ver217 2022-03-23 14:39:41 +0800
  • 5311ef7884 polish code ver217 2022-03-23 12:26:05 +0800
  • 2f5b85eb99 polish code ver217 2022-03-23 12:09:19 +0800
  • dee69f4454 rename variable ver217 2022-03-23 11:57:41 +0800
  • 1c6c4b1445 sharded model supports reuse fp16 shard ver217 2022-03-23 11:55:45 +0800
  • 0e121a256a Modified some description statements. chenjunejie 2022-03-22 17:46:53 +0800
  • deb7131787 Merge remote-tracking branch 'origin/main' chenjunejie 2022-03-22 17:45:12 +0800
  • 104dac49c6
    Merge branch 'hpcaitech:main' into main JunjieChen-2020 2022-03-22 17:43:54 +0800
  • 5990c97f50 import subprocess chenjunejie 2022-03-22 17:42:54 +0800
  • f24b5ed201
    [MOE] remove old MoE legacy (#493) HELSON 2022-03-22 17:37:16 +0800
  • 49fc73447b remove old MoE legacy 1SAA 2022-03-22 17:27:57 +0800
  • c4c02424f3
    [zero] sharded model manages ophooks individually (#492) ver217 2022-03-22 17:33:20 +0800
  • cbcadb6623 sharded model triggers ophooks' post_iter ver217 2022-03-22 17:12:03 +0800
  • c9023d4078
    [MOE] support PR-MOE (#488) HELSON 2022-03-22 16:48:22 +0800
  • 5c6902cf7c support PR-MOE 1SAA 2022-03-22 12:11:41 +0800
  • a9ecb4b244
    [zero] polish sharded optimizer v2 (#490) ver217 2022-03-22 15:53:48 +0800
  • 9d12199893 polish code ver217 2022-03-22 15:11:22 +0800
  • c82bbd67e0 add cpu adamW lclgy 2022-03-22 14:59:37 +0800
  • 62b0a8d644
    [zero] sharded optim support hybrid cpu adam (#486) ver217 2022-03-22 14:56:59 +0800
  • b334822163
    [zero] polish sharded param name (#484) Jiarui Fang 2022-03-22 14:36:16 +0800
  • 9caa8b6481
    docs get correct release version (#489) ver217 2022-03-22 14:24:41 +0800
  • c944b07f07 docs get correct release version ver217 2022-03-22 14:22:50 +0800
  • 03278b1df6 polish jiaruifang 2022-03-22 13:59:12 +0800
  • 2bc4b33a54 polish docstring ver217 2022-03-22 13:33:46 +0800
  • 4720507332 polsih jiaruifang 2022-03-22 13:27:38 +0800
  • c48ce58b43 update unit test ver217 2022-03-22 13:17:18 +0800
  • b050d51c54 sharded optim support hybrid cpu adam ver217 2022-03-22 12:09:03 +0800
  • 12567a74db polish jiaruifang 2022-03-22 11:56:28 +0800
  • 7508bc0fdd polish code jiaruifang 2022-03-22 11:38:04 +0800
  • 06a1d0e6bd [test] fixed rerun_on_exception and adapted test cases FrankLeeeee 2022-03-22 11:00:18 +0800
  • d7ea63992b
    [MOE] add FP32LinearGate for MOE in NaiveAMP context (#480) HELSON 2022-03-22 10:50:20 +0800
  • 8ed8afe540 polish jiaruifang 2022-03-22 10:48:32 +0800
  • 3945148e81 add FP32LinearGate for MOE in NaiveAMP context 1SAA 2022-03-21 16:41:57 +0800
  • 6c97e2fc16 polish code jiaruifang 2022-03-22 09:35:36 +0800
  • 353566c198
    Automated submodule synchronization (#483) github-actions[bot] 2022-03-22 09:34:26 +0800
  • 53417b8777 [zero] polish sharded param name jiaruifang 2022-03-22 09:27:31 +0800
  • b9c3abd316 Automated submodule synchronization github-actions 2022-03-22 00:01:23 +0000
  • 65c0f380c2
    [format] polish name format for MOE (#481) Jiarui Fang 2022-03-21 23:19:47 +0800
  • 8085ab7f95 polish jiaruifang 2022-03-21 21:59:11 +0800
  • 758c882458 polish jiaruifang 2022-03-21 21:23:22 +0800
  • 82f2388632 Merge branch 'main' of github.com:hpcaitech/ColossalAI into jiaruifang/polish_moe_name jiaruifang 2022-03-21 21:21:34 +0800
  • 961498af7c [MOE] linear gating jiaruifang 2022-03-21 18:23:42 +0800
  • 12941797f2 polish code jiaruifang 2022-03-21 17:35:42 +0800
  • 7fccffc745 polish code jiaruifang 2022-03-21 17:13:07 +0800