ColossalAI/colossalai
Wenxuan Tan 8fd25d6e09
[Feature] Split cross-entropy computation in SP (#5959)
* halfway

* fix cross-PP-stage position id length diff bug

* fix typo

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* unified cross entropy func for all shardformer models

* remove redundant lines

* add basic ring attn; debug cross entropy

* fwd bwd logic complete

* fwd bwd logic complete; add experimental triton rescale

* precision tests passed

* precision tests passed

* fix typos and remove misc files

* update softmax_lse shape by new interface

* change tester name

* remove buffer clone; support packed seq layout

* add varlen tests

* fix typo

* all tests passed

* add dkv_group; fix mask

* remove debug statements

* adapt chatglm, command-R, qwen

* debug

* halfway

* fix cross-PP-stage position id length diff bug

* fix typo

* fix typo

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* unified cross entropy func for all shardformer models

* remove redundant lines

* add basic ring attn; debug cross entropy

* fwd bwd logic complete

* fwd bwd logic complete; add experimental triton rescale

* precision tests passed

* precision tests passed

* fix typos and remove misc files

* add sp_mode to benchmark; fix varlen interface

* update softmax_lse shape by new interface

* add varlen tests

* fix typo

* all tests passed

* add dkv_group; fix mask

* remove debug statements

* add comments

* q1 index only once

* remove events to simplify stream sync

* simplify forward/backward logic

* 2d ring forward passed

* 2d ring backward passed

* fixes

* fix ring attn loss

* 2D ring backward + llama passed

* merge

* update logger

* fix typo

* rebase

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fix typo

* remove typos

* fixes

* support GPT

---------

Co-authored-by: Edenzzzz <wtan45@wisc.edu>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
2024-09-10 12:06:50 +08:00
..
_C Clean up 2024-06-07 09:09:29 +00:00
_analyzer [test] Fix/fix testcase (#5770) 2024-06-03 15:26:01 +08:00
accelerator [hotfix] fix typo change MoECheckpintIO to MoECheckpointIO (#5335) 2024-03-05 21:52:30 +08:00
amp [npu] change device to accelerator api (#5239) 2024-01-09 10:20:05 +08:00
auto_parallel [pre-commit.ci] pre-commit autoupdate (#5572) 2024-07-01 17:16:41 +08:00
autochunk [hotfix] Fix examples no pad token & auto parallel codegen bug; (#5606) 2024-04-18 18:15:50 +08:00
booster [fp8] fix linear hook (#6046) 2024-09-03 16:37:16 +08:00
checkpoint_io [colossalai/checkpoint_io/...] fix bug in load_state_dict_into_model; format error msg (#6020) 2024-09-02 16:56:35 +08:00
cli [devops] fix extention building (#5427) 2024-03-05 15:35:54 +08:00
cluster [FP8] rebase main (#5963) 2024-08-06 16:29:37 +08:00
context [Fix]: implement thread-safety singleton to avoid deadlock for very large-scale training scenarios (#5625) 2024-04-25 14:45:52 +08:00
device [Feature] Distributed optimizers: Lamb, Galore, CAME and Adafactor (#5694) 2024-05-14 13:52:45 +08:00
fx [test] Fix/fix testcase (#5770) 2024-06-03 15:26:01 +08:00
inference [colossalai/checkpoint_io/...] fix bug in load_state_dict_into_model; format error msg (#6020) 2024-09-02 16:56:35 +08:00
interface [Feature] Distributed optimizers: Lamb, Galore, CAME and Adafactor (#5694) 2024-05-14 13:52:45 +08:00
kernel [NFC] Fix code factors on inference triton kernels (#5743) 2024-05-21 22:12:15 +08:00
lazy [fp8] Merge feature/fp8_comm to main branch of Colossalai (#6016) 2024-08-22 09:21:34 +08:00
legacy [fp8] Merge feature/fp8_comm to main branch of Colossalai (#6016) 2024-08-22 09:21:34 +08:00
logging [fp8] Merge feature/fp8_comm to main branch of Colossalai (#6016) 2024-08-22 09:21:34 +08:00
moe [fp8]Moe support fp8 communication (#5977) 2024-08-09 18:26:02 +08:00
nn [misc] fix dist logger (#5782) 2024-06-05 15:04:22 +08:00
pipeline [fp8] Merge feature/fp8_comm to main branch of Colossalai (#6016) 2024-08-22 09:21:34 +08:00
quantization [fp8] disable all_to_all_fp8 in intranode (#6045) 2024-09-09 13:47:17 +08:00
shardformer [Feature] Split cross-entropy computation in SP (#5959) 2024-09-10 12:06:50 +08:00
tensor [fp8] support fp8 amp for hybrid parallel plugin (#5975) 2024-08-07 18:21:08 +08:00
testing [misc] update compatibility (#6008) 2024-08-16 18:49:14 +08:00
utils Merge pull request #5310 from hpcaitech/feature/npu 2024-01-29 13:49:39 +08:00
zero [fp8] Merge feature/fp8_comm to main branch of Colossalai (#6016) 2024-08-22 09:21:34 +08:00
__init__.py [devops] remove post commit ci (#5566) 2024-04-08 15:09:40 +08:00
initialize.py [FP8] rebase main (#5963) 2024-08-06 16:29:37 +08:00