文件最后提交记录最后更新时间
!2332 core_r0.12.0 add v2 ut Merge pull request !2332 from wangruiqi/master 11 个月前
fixbug for coc print Co-authored-by: clc2025<chenlucong@huawei.com> # message auto-generated for no-merge-commit merge: !3052 merge fixbugcoc into master fixbug for coc print Created-by: clc2025 Commit-by: clc2025 Merged-by: ascend-robot Description: fixbug for coc print 问题背景:https://wiki.huawei.com/domains/84233/wiki/312004/WIKI202512019235085 See merge request: Ascend/MindSpeed!30525 个月前
fixbug for coc print Co-authored-by: clc2025<chenlucong@huawei.com> # message auto-generated for no-merge-commit merge: !3052 merge fixbugcoc into master fixbug for coc print Created-by: clc2025 Commit-by: clc2025 Merged-by: ascend-robot Description: fixbug for coc print 问题背景:https://wiki.huawei.com/domains/84233/wiki/312004/WIKI202512019235085 See merge request: Ascend/MindSpeed!30525 个月前
!2200 refactor: mc2 ut Merge pull request !2200 from yangcheng/master 1 年前
fix reapeated allgather & tp2d forward_only Co-authored-by: zhao-yifan222<17801034608@163.com> # message auto-generated for no-merge-commit merge: merge master into master fix reapeated allgather & tp2d forward_only Created-by: zhao-yifan27 Commit-by: zhao-yifan222 Merged-by: ascend-robot Description: 1fix reapeated allgather 该问题导致moe 场景 edp>1 且开启reuse fp32时,精度有问题 2 fix tp2d forward_only 场景跑不通报错 See merge request: Ascend/MindSpeed!28548 个月前
!2659 fix: correct parameter naming in _initialize_affine_weight_gpu Merge pull request !2659 from 邓佳/core_r0.12.1_fix_dj 10 个月前
!2290 fix: ripipe v2 Merge pull request !2290 from 邓佳/master_fix_ripipe 1 年前
!359 change ascendspeed to mindspeed Merge pull request !359 from 邓佳/master 1 年前
!571 Megatron 0.7.0适配 Merge pull request !571 from 闻江/master 1 年前
!1323 tp-2d内存优化-master Merge pull request !1323 from liujianxing/tp_2d_mem_3 1 年前
!699 2d张量并行 Merge pull request !699 from liujianxing/2d_tensor_0824 1 年前
!2425 rm: unused code Merge pull request !2425 from 邓佳/master_rm_v2 11 个月前
!2729 [bugfix!!!]shared_expert_gate remove&amp; logits adjust &amp; overlap readme append Merge pull request !2729 from yangjie/master 9 个月前
quant fp8 optimizer 6 个月前
!620 【Fix Bug】修复使用开关采集Profiling数据异常的问题以及ND_MatMul的精度问题 Merge pull request !620 from robert/master 1 年前
feature(fp8): te checkpoint Co-authored-by: Muu<koimuu@163.com> # message auto-generated for no-merge-commit merge: !3162 merge feature_checkpoint into master feature(fp8): te checkpoint Created-by: Muuyo Commit-by: Muu Merged-by: ascend-robot Description: 1. 引入 te checkpoint消除重计算中冗余的量化操作 2. refactor(blockwise): 删除128*128的blockwise策略, 保留1 * 128|128 * 128策略替换 3. perf(hif8): 删除多余的cast 4. fix(delayed): 修复delayed算法 5. refactor(recipe 2x): 重构blockwise和mxfp8策略数据存取, 简化后续算子适配 6. 消除字符串字面量, 采用枚举替代 验证报告: https://wiki.huawei.com/domains/76578/wiki/233229/WIKI202601139775970 See merge request: Ascend/MindSpeed!31624 个月前