0
代码
代码
Issues
Pull Requests
讨论
Wiki
分析
0
ascend-robotascend-robotfeat: add custom pp layout
94e8ac1c创建于 10 小时前历史提交
文件最后提交记录最后更新时间
__init__.py
feat: add custom pp layout Co-authored-by: wuweiqiang24<wuweiqiang11@huawei.com> # message auto-generated for no-merge-commit merge: !3496 merge add_pp_layout into master feat: add custom pp layout Created-by: wuweiqiang24 Commit-by: wuweiqiang24 Merged-by: ascend-robot Description: 新增pipeline-model-parallel-layout功能，支持自定义PP每个stage的层排布 验证链接：https://wiki.huawei.com/domains/137239/wiki/268925/WIKI2026052611233549 issue: https://gitcode.com/Ascend/MindSpeed/issues/166 See merge request: Ascend/MindSpeed!349610 小时前
dualpipev_feature.py
fix【问题】use_custom_fsdp特性和dualpipev不兼容 Co-authored-by: lizhi<lizhi166@huawei.com> # message auto-generated for no-merge-commit merge: !3080 merge bug-master into master fix【问题】use_custom_fsdp特性和dualpipev不兼容 Created-by: liz_hw Commit-by: lizhi Merged-by: ascend-robot Description: fix【问题】use_custom_fsdp特性和dualpipev不兼容 See merge request: Ascend/MindSpeed!30805 个月前
multi_parameter.py
[test]Add fsdp2 feature to ST Co-authored-by: yulelanmei<huangyijie8@huawei.com> # message auto-generated for no-merge-commit merge: !3086 merge add_st into master [test]Add fsdp2 feature to ST Created-by: yulelanmei Commit-by: yulelanmei Merged-by: ascend-robot Description: 1. 为ST用例新增目前缺少的需要落A5转测的加速特性，包含 fsdp2 在内 2. 统一更新 acl.get_soc_name() 为 torch_npu.npu.get_device_name() ，后者在A2/A3/A5上都能使用，前者在A5不支持 3. multi_parameter_pipeline用例修复，删除重复用例文件 See merge request: Ascend/MindSpeed!30865 个月前
noop_layers.py
!2788 feat: verl noop layer adaptor Merge pull request !2788 from YE ZHENYUAN/master0828 8 个月前
optimize_p2p_comm.py
feat: add custom pp layout Co-authored-by: wuweiqiang24<wuweiqiang11@huawei.com> # message auto-generated for no-merge-commit merge: !3496 merge add_pp_layout into master feat: add custom pp layout Created-by: wuweiqiang24 Commit-by: wuweiqiang24 Merged-by: ascend-robot Description: 新增pipeline-model-parallel-layout功能，支持自定义PP每个stage的层排布 验证链接：https://wiki.huawei.com/domains/137239/wiki/268925/WIKI2026052611233549 issue: https://gitcode.com/Ascend/MindSpeed/issues/166 See merge request: Ascend/MindSpeed!349610 小时前
optimize_send_recv_comm.py
!2279 bugfix: 修复pp和megatron_basic获取args参数的错误 Merge pull request !2279 from liurong1995/bugfix_args 1 年前
pipeline_model_parallel_layout_feature.py
feat: add custom pp layout Co-authored-by: wuweiqiang24<wuweiqiang11@huawei.com> # message auto-generated for no-merge-commit merge: !3496 merge add_pp_layout into master feat: add custom pp layout Created-by: wuweiqiang24 Commit-by: wuweiqiang24 Merged-by: ascend-robot Description: 新增pipeline-model-parallel-layout功能，支持自定义PP每个stage的层排布 验证链接：https://wiki.huawei.com/domains/137239/wiki/268925/WIKI2026052611233549 issue: https://gitcode.com/Ascend/MindSpeed/issues/166 See merge request: Ascend/MindSpeed!349610 小时前
ripipe_schedules_feature.py
【问题单】补充特性限制说明 Co-authored-by: libaokui<libaokui@huawei.com> # message auto-generated for no-merge-commit merge: !2998 merge wtd into master 【问题单】补充特性限制说明 Created-by: libaokui Commit-by: libaokui Merged-by: ascend-robot Description: 1. 暂不支持Megatron原生特性defer-embedding-wgrad-compute 2. recompute-in-bubble不支持与swap-attention、no-overlap-p2p-communication、overlap-grad-reducer不兼容 3. recompute-in-advance不支持no-overlap-p2p-communication、overlap-grad-reduce不兼容 详见以下报告： https://wiki.huawei.com/domains/73366/wiki/264623/WIKI202511139013241?title=4----overlap-grad-reduce---overlap-param-gather See merge request: Ascend/MindSpeed!29986 个月前
unaligned_pineline_feature.py
!2788 feat: verl noop layer adaptor Merge pull request !2788 from YE ZHENYUAN/master0828 8 个月前
variable_seq_length.py
!2801 fix: transformer config validate bug Merge pull request !2801 from YE ZHENYUAN/master0902 8 个月前