You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

Edenzzzz f5c84af0b0 [Feature] Zigzag Ring attention (#5905) ... * 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 * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add sp_mode to benchmark; fix varlen interface * 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 --------- Co-authored-by: Edenzzzz <wtan45@wisc.edu> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>		3 months ago
..
__init__.py	[MoE/ZeRO] Moe refactor with zero refactor (#5821)	5 months ago
checkpoint_io_base.py	[lora] add lora APIs for booster, support lora for TorchDDP (#4981)	7 months ago
general_checkpoint_io.py	[lora] add lora APIs for booster, support lora for TorchDDP (#4981)	7 months ago
hybrid_parallel_checkpoint_io.py	[Feature] Zigzag Ring attention (#5905)	3 months ago
index_file.py	[misc] update pre-commit and run all files (#4752)	1 year ago
moe_checkpoint.py	[moe] implement tp	4 months ago
utils.py	[MoE/ZeRO] Moe refactor with zero refactor (#5821)	5 months ago