Commit Graph

  • 0a01e2a453 fix the attn wangbluo 2024-09-13 03:33:08 +00:00
  • 216d54e374 [pre-commit.ci] auto fixes from pre-commit.com hooks pre-commit-ci[bot] 2024-09-13 02:38:39 +00:00
  • fdd84b9087 fix the sp wangbluo 2024-09-13 02:32:03 +00:00
  • 9bc3b6e220 [feat] moehybrid support zerobubble; duanjunwen 2024-09-12 02:51:46 +00:00
  • a35a078f08 [doc] update sp doc (#6055) flybird11111 2024-09-11 17:25:14 +08:00
  • 13946c4448 [fp8] hotfix backward hook (#6053) Hongxin Liu 2024-09-11 16:11:25 +08:00
  • 11ae6848c6 [zerobubble]Support ZeroBubble Pipeline (#6034) duanjunwen 2024-09-10 17:33:09 +08:00
  • c54c4fcd15 [hotfix] moe hybrid parallelism benchmark & follow-up fix (#6048) botbw 2024-09-10 17:30:53 +08:00
  • 8fd25d6e09 [Feature] Split cross-entropy computation in SP (#5959) Wenxuan Tan 2024-09-10 12:06:50 +08:00
  • b3db1058ec [release] update version (#6041) v0.4.3 Hongxin Liu 2024-09-10 10:31:09 +08:00
  • 6c2a120bed [fix] add testcase with microbatch 4; duanjunwen 2024-09-09 10:16:03 +00:00
  • 8366a7855f [fix] update optim state dict assert (include param group & state); fix mem assert after add optim; duanjunwen 2024-09-09 09:27:13 +00:00
  • ce58d8e8bf [fix] add output_obj_grad assert None at bwd b step; replace input_obj.require_grad_ with treemap; duanjunwen 2024-09-09 08:19:58 +00:00
  • 7568b34626 [fix] fix redundant detach & clone; add buffer assertation in the end; duanjunwen 2024-09-09 08:04:28 +00:00
  • fed8b1587d [fix] fix model zoo import; duanjunwen 2024-09-09 06:39:33 +00:00
  • a5ec3d4285 [fix] fix mem; use a new model shape; only assert mem less and equal than theo; duanjunwen 2024-09-09 06:38:31 +00:00
  • 5ce6dd75bf [fp8] disable all_to_all_fp8 in intranode (#6045) Hanks 2024-09-09 13:47:17 +08:00
  • 35a7b636b3 [fix] fix mem assertation duanjunwen 2024-09-09 05:41:39 +00:00
  • 400e5e5b23 [fix] mem assertation' duanjunwen 2024-09-09 02:58:06 +00:00
  • 4a358348c7 [fix] fix mem check; duanjunwen 2024-09-04 10:57:38 +00:00
  • 2f09c374f3 [feat] add memory assertation; duanjunwen 2024-09-04 06:34:18 +00:00
  • e6e1a97a6d [fix] fix requir grad position and detach position and input&output local buffer append position; duanjunwen 2024-09-04 03:31:08 +00:00
  • 20503cdfdf [fix] rm requir_grad for output; duanjunwen 2024-09-03 09:24:40 +00:00
  • b4103f125c [fix] fix detach output & release output; duanjunwen 2024-09-03 09:09:41 +00:00
  • 4c1f81c683 [fix] fix bwd step if condition; remove useless comments and format info; duanjunwen 2024-09-03 08:56:08 +00:00
  • 26e553937b [fp8] fix linear hook (#6046) Hongxin Liu 2024-09-03 16:37:16 +08:00
  • c3b5caff0e [fp8] optimize all-gather (#6043) Hongxin Liu 2024-09-03 15:45:17 +08:00
  • ab643c9af7 [fix] rm output.data after send fwd; duanjunwen 2024-09-03 14:12:17 +08:00
  • a48afc4a66 [fix] fix optim bwd; duanjunwen 2024-09-03 02:40:26 +00:00
  • c650a906db [Hotfix] Remove deprecated install (#6042) Tong Li 2024-09-03 10:33:18 +08:00
  • 591a13bf7e [fix] fix optim bwd; duanjunwen 2024-09-02 11:19:42 +00:00
  • 77fe44286c [fix] rm zbv in hybridplugin duanjunwen 2024-09-02 10:00:43 +00:00
  • 6d18d38d5c [feat] update test; rm comments; duanjunwen 2024-09-02 09:50:47 +00:00
  • e9032fb0b2 [colossalai/checkpoint_io/...] fix bug in load_state_dict_into_model; format error msg (#6020) Gao, Ruiyuan 2024-09-02 16:56:35 +08:00
  • a7b767b071 [fix] fix communication_map; duanjunwen 2024-08-30 05:56:02 +00:00
  • 8eb6eac225 [fix] fix optim bwd; add license for v_schedule; remove redundant attributes; fix schedule loop "while"--> "for"; add communication dict; duanjunwen 2024-08-30 05:42:43 +00:00
  • 6af81d8c0d [feat] add fwd_bwd_step, run_fwd_only; duanjunwen 2024-08-30 02:47:52 +00:00
  • 48ba22dbfd [feat] fix optimizer bwd b & w; support return accum loss & output duanjunwen 2024-08-29 08:54:45 +00:00
  • e96a0761ea [FP8] unsqueeze scale to make it compatible with torch.compile (#6040) Guangyao Zhang 2024-08-29 14:49:23 +08:00
  • 4c4b01b859 [feat] add optim backward_b_by_grad duanjunwen 2024-08-29 03:16:59 +00:00
  • 0d3a85d04f add fused norm (#6038) Tong Li 2024-08-28 17:12:51 +08:00
  • 4a68efb7da [Colossal-LLaMA] Refactor latest APIs (#6030) Tong Li 2024-08-28 17:01:58 +08:00
  • b1419ef76a [fix] fix poc test; add comments in poc; duanjunwen 2024-08-28 05:47:53 +00:00
  • 582ba0d6ff [feat] fix func name & ci; add comments; duanjunwen 2024-08-28 03:40:50 +00:00
  • b5f7b4d228 [feat] fix poc format duanjunwen 2024-08-28 03:08:35 +00:00
  • d6e3d7d2a3 [feat] fix ci; add assert; duanjunwen 2024-08-28 02:41:05 +00:00
  • 29383b2de0 [fix] update duanjunwen 2024-08-28 02:33:42 +00:00
  • cc1b0efc17 [plugin] hotfix zero plugin (#6036) Hongxin Liu 2024-08-28 10:16:48 +08:00
  • fe209164f1 [feat] add apply v_schedule graph; p & p.grad assert err exist; duanjunwen 2024-08-27 10:29:39 +00:00
  • 8b37323f16 [feat] add run_fwd_bwd_with_microbatch (replace input) & test; add p&p.grad assert close test & all pass; duanjunwen 2024-08-27 09:31:38 +00:00
  • 9e0bd1af00 [fix] fix ci test; add pytest; duanjunwen 2024-08-27 08:00:23 +00:00
  • 283c9ff5d2 [fix] rm useless assign and comments; duanjunwen 2024-08-27 07:31:58 +00:00
  • 1b4bb2beeb [feat] add comments for ZBV func; duanjunwen 2024-08-27 07:11:50 +00:00
  • f1c1a87246 [feat] add test for p & p grad; duanjunwen 2024-08-27 06:37:26 +00:00
  • 5e09c8b4e1 [feat] split communication and calculation; fix pop empty send_bwd_buffer error; duanjunwen 2024-08-27 06:29:13 +00:00
  • d383449fc4 [CI] Remove triton version for compatibility bug; update req torch >=2.2 (#6018) Wenxuan Tan 2024-08-27 10:12:21 +08:00
  • 17904cb5bf Merge pull request #6012 from hpcaitech/feature/fp8_comm Hongxin Liu 2024-08-27 10:09:43 +08:00
  • 1d75045c37 [feat] add test run_fwd_bwd automatic scheduling; duanjunwen 2024-08-26 11:21:56 +00:00
  • 4a6f31eb0c Merge pull request #6033 from wangbluo/fix Wang Binluo 2024-08-26 14:06:06 +08:00
  • fd5526b76e Merge branch 'main' into dev/zero_bubble duanjunwen 2024-08-26 04:03:20 +00:00
  • 107230d27a [update] update text; duanjunwen 2024-08-26 04:00:51 +00:00
  • 80d24ae519 [pre-commit.ci] auto fixes from pre-commit.com hooks pre-commit-ci[bot] 2024-08-26 03:48:42 +00:00
  • dae39999d7 fix wangbluo 2024-08-26 03:45:42 +00:00
  • 203033ea16 [fix] fix weight not close; duanjunwen 2024-08-23 08:57:27 +00:00
  • 7cf9df07bc [Hotfix] Fix llama fwd replacement bug (#6031) Wenxuan Tan 2024-08-23 15:44:27 +08:00
  • c18ef060cf [feat] add dw test; duanjunwen 2024-08-23 06:04:12 +00:00
  • 3568df498a [pre-commit.ci] auto fixes from pre-commit.com hooks flybird11111-patch-1 pre-commit-ci[bot] 2024-08-23 05:50:36 +00:00
  • 0bf46c54af Merge pull request #6029 from hpcaitech/flybird11111-patch-1 Wang Binluo 2024-08-23 13:50:04 +08:00
  • 9e767643dd Update low_level_zero_plugin.py flybird11111 2024-08-23 13:49:53 +08:00
  • 3b0df30362 [pre-commit.ci] auto fixes from pre-commit.com hooks pre-commit-ci[bot] 2024-08-23 05:48:11 +00:00
  • 0bc9a870c0 Update train_dpo.py flybird11111 2024-08-23 13:47:13 +08:00
  • ee9baedadf [feat] add zerobubble pp (just a frame now); add POC test for dx_dw; add test for zerobubble; duanjunwen 2024-08-22 10:25:34 +00:00
  • caab4a307f Merge branch 'main' into feature/fp8_comm Hongxin Liu 2024-08-22 15:14:38 +08:00
  • afe845ff15 Merge pull request #6024 from wangbluo/fix_merge Wang Binluo 2024-08-22 11:07:04 +08:00
  • a292554179 [pre-commit.ci] auto fixes from pre-commit.com hooks pre-commit-ci[bot] 2024-08-22 03:04:43 +00:00
  • 971b16a74f fix wangbluo 2024-08-22 03:00:40 +00:00
  • d77e66a577 Merge pull request #6023 from wangbluo/fp8_merge Wang Binluo 2024-08-22 10:32:13 +08:00
  • eea37da6fa [fp8] Merge feature/fp8_comm to main branch of Colossalai (#6016) Wang Binluo 2024-08-22 09:21:34 +08:00
  • 8b8e282441 fix wangbluo 2024-08-21 09:18:45 +00:00
  • 698c8b9804 fix wangbluo 2024-08-21 03:58:21 +00:00
  • 6aface9316 fix wangbluo 2024-08-21 03:51:25 +00:00
  • 193030f696 fix wangbluo 2024-08-21 03:21:49 +00:00
  • eb5ba40def fix the merge wangbluo 2024-08-21 02:58:23 +00:00
  • 39e2597426 [ColossalChat] Add PP support (#6001) Tong Li 2024-08-21 10:47:39 +08:00
  • 0d3b0bd864 [plugin] add cast inputs option for zero (#6003) (#6022) Hongxin Liu 2024-08-21 10:21:26 +08:00
  • 2d362ac090 fix merge wangbluo 2024-08-20 09:26:04 +00:00
  • 2e4cbe3a2d fix wangbluo 2024-08-20 09:11:02 +00:00
  • 2ee6235cfa fix wangbluo 2024-08-20 06:48:16 +00:00
  • f7acfa1bd5 fix wangbluo 2024-08-20 05:07:58 +00:00
  • 53823118f2 fix wangbluo 2024-08-20 03:20:13 +00:00
  • dcc44aab8d [misc] Use dist logger in plugins (#6011) Edenzzzz 2024-08-20 10:32:41 +08:00
  • 1f703e0ef4 fix wangbluo 2024-08-19 10:15:16 +00:00
  • 88b3f0698c fix the merge wangbluo 2024-08-19 10:11:27 +00:00
  • 2eb36839c6 fix wangbluo 2024-08-19 09:23:10 +00:00
  • 12b44012d9 fix wangbluo 2024-08-19 09:02:16 +00:00
  • 0d8e82a024 Merge branch 'fp8_merge' of https://github.com/wangbluo/ColossalAI into fp8_merge wangbluo 2024-08-19 08:10:27 +00:00
  • 4c82bfcc54 fix the merge wangbluo 2024-08-19 08:09:34 +00:00
  • 64aad96723 [pre-commit.ci] auto fixes from pre-commit.com hooks pre-commit-ci[bot] 2024-08-19 08:08:45 +00:00
  • 3353042525 fix the merge wangbluo 2024-08-19 08:07:51 +00:00
  • f1c3266a94 overlap kv comm with output rescale (#6017) Edenzzzz 2024-08-19 14:08:17 +08:00