master

ascend-robotfix(pytorch): fix duplicate transports in mla and rope

b00643a0创建于 3月19日历史提交

文件	最后提交记录	最后更新时间
datasets	!3282 [mindspore][master]fix ringattentionupdate Merge pull request !3282 from 杨承翰/ring_master	8 个月前
distributed	!2744 【训练性能】兼容层named_modules递归获取参数性能优化 Merge pull request !2744 from zhangyihui/master	11 个月前
models	!3352 [mindspore][master][bugfix]patch_apply_llama3_scaling Merge pull request !3352 from xinyuan/patch_apply_llama3_scaling	8 个月前
pipeline_parallel	fix(pytorch): fix duplicate transports in mla and rope Co-authored-by: zhyebin01<zhangyebin@h-partners.com> # message auto-generated for no-merge-commit merge: !4309 merge bugfix2 into master fix(pytorch): fix duplicate transports in mla and rope Created-by: zhyebin01 Commit-by: zhyebin01 Merged-by: ascend-robot Description: ## What this PR does / why we need it? fix duplicate transports in mla and rope ## Does this PR introduce any user-facing change? no ## How was this patch tested? pipeline test passed See merge request: Ascend/MindSpeed-LLM!4309	2 个月前
tensor_parallel	[mindspore][model] optimize dsv3 performance Co-authored-by: wanglijun55<wanglijun54@huawei.com> # message auto-generated for no-merge-commit merge: !3871 merge master-ds3 into master [mindspore][model] optimize dsv3 performance Created-by: wanglijun55 Commit-by: wanglijun55 Merged-by: ascend-robot Description: 关联问题单：https://e.gitee.com/mind_spore/issues/table?issue=ID75AB 修改说明： 1、优化1：get_batch中for循环里多次调用cpu tensor的item、gt、bool等操作，上patch先转numpy再过for循环 自验证结果： 1、性能达到0.99x pta，精度对齐   See merge request: Ascend/MindSpeed-LLM!3871	5 个月前
transformer	[mindspore][bugfix][master] fix shared expert by removing stream Co-authored-by: ffmh<fengminghao2@huawei.com> # message auto-generated for no-merge-commit merge: !3917 merge fix_shared_master into master [mindspore][bugfix][master] fix shared expert by removing stream Created-by: ffmh Commit-by: ffmh Merged-by: ascend-robot Description: fix shared expert by removing stream See merge request: Ascend/MindSpeed-LLM!3917	5 个月前