| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
fix(torch/cp): use sbnd format before all2all Co-authored-by: clc2025<chenlucong@huawei.com> # message auto-generated for no-merge-commit merge: !3282 merge fixbug_ulysses_tnd into master fix(torch/cp): use sbnd format before all2all Created-by: clc2025 Commit-by: clc2025 Merged-by: ascend-robot Description: fixbug for ulysses tnd See merge request: Ascend/MindSpeed!3282 | 7 个月前 | |
| 1 年前 | ||
Add offline pad_data Co-authored-by: wuweiqiang24<wuweiqiang11@huawei.com> # message auto-generated for no-merge-commit merge: !2938 merge revise_preprocess_data into master Add offline pad_data Created-by: wuweiqiang24 Commit-by: wuweiqiang24 Merged-by: ascend-robot Description: 增加离线预处理pack数据集功能，可提前将数据padding到2\*CP倍，在线使用CP功能时可节约padding部分耗时 * 精度与非离线padding版本存在一定差异  * Llama2-7b，单机16k，GBS=8场景下，性能提升4.8%  See merge request: Ascend/MindSpeed!2938 | 10 个月前 | |
perf(verl ckpt): ckpt load and save acceleration Co-authored-by: 李鸣沼<lmztju@126.com> # message auto-generated for no-merge-commit merge: !3074 merge verl_load_and_save_ckpt into master perf(verl ckpt): ckpt load and save acceleration Created-by: lmztju Commit-by: 李鸣沼;l30057177 Merged-by: ascend-robot Description: **测试场景1：** verl+megatron后端dapo-qwen3-30b 910A3双机加载和保存ckpt 本地存储  See merge request: Ascend/MindSpeed!3074 | 9 个月前 | |
docs:fix docs/zh mistakes Co-authored-by: Keilo_W<wangkaiyu11@h-partners.com> # message auto-generated for no-merge-commit merge: !3318 merge master into master docs:fix docs/zh mistakes Created-by: Keilo_W Commit-by: Keilo_W Merged-by: ascend-robot Description: 修改了一些被误操作的注释及代码 See merge request: Ascend/MindSpeed!3318 | 6 个月前 | |
fix: adapt device_capability for PT212 Co-authored-by: yulelanmei<huangyijie8@huawei.com> # message auto-generated for no-merge-commit merge: !3555 merge master into master fix: adapt device_capability for PT212 Created-by: yulelanmei Commit-by: yulelanmei Merged-by: ascend-robot Description: ## What this PR does / why we need it? Unable to run training with Pytorch2.12 because interface is not compatible. ## Does this PR introduce any user-facing change? N/A ## How was this patch tested? Run any core test case with Pytorch2.12 version. See merge request: Ascend/MindSpeed!3555 | 3 个月前 | |
fix hccl buffer errors for verl cases Co-authored-by: quancs001<quancs@qq.com> # message auto-generated for no-merge-commit merge: !3478 merge fix_hccl_buffer_for_verl into master fix hccl buffer errors for verl cases Created-by: quancs001 Commit-by: quancs001 Merged-by: ascend-robot Description: What this PR does / why we need it? When running RL exps with verl, several errors are raised, e.g.: 1. megatron.training.get_args raises an exception 2. the args for hccl_buffer with ";" could not be parsed by hydra, and the error LexerNoViableAltException is raised. This PR is proposed to solve the errors. Does this PR introduce any user-facing change? No. How was this patch tested? The code is tested and verified locally. See merge request: Ascend/MindSpeed!3478 | 4 个月前 | |
fix: TE LayerNormLinear init weight order align NVTE Co-authored-by: clc2025<chenlucong@huawei.com> # message auto-generated for no-merge-commit merge: !3569 merge fix_lnliner_initweight into master fix: TE LayerNormLinear init weight order align NVTE Created-by: clc2025 Commit-by: clc2025 Merged-by: ascend-robot Description: ## What this PR does / why we need it? 具体查看关联issue ## Does this PR introduce any user-facing change? NA ## How was this patch tested? 基于脚本用例，从GPU上保存权重，NPU加载权重后断点续训，精度能对齐且无功能报错 See merge request: Ascend/MindSpeed!3569 | 3 个月前 | |
Add swap layer input Co-authored-by: JialiZheng<jializheng@huawei.com> # message auto-generated for no-merge-commit merge: !3529 merge master into master Add swap layer input Created-by: JialiZheng1 Commit-by: JialiZheng Merged-by: ascend-robot Description: Add swap input RFC：https://gitcode.com/Ascend/MindSpeed/issues/176 验证wiki：https://wiki.huawei.com/wiki/WIKI2026060411356524 See merge request: Ascend/MindSpeed!3529 | 3 个月前 | |
quant fp8 optimizer | 10 个月前 | |
disttrain intervl2 适配开箱 Co-authored-by: gcw_amOUPDs9<fuyuefeng@huawei.com> # message auto-generated for no-merge-commit merge: !2915 merge master into master disttrain intervl2 适配开箱 Created-by: gcw_amOUPDs9 Commit-by: gcw_amOUPDs9 Merged-by: ascend-robot Description: disttrain intervl2 适配开箱 disttrain intervl2 适配开箱 See merge request: Ascend/MindSpeed!2915 | 10 个月前 | |
fix: TE LayerNormLinear init weight order align NVTE Co-authored-by: clc2025<chenlucong@huawei.com> # message auto-generated for no-merge-commit merge: !3569 merge fix_lnliner_initweight into master fix: TE LayerNormLinear init weight order align NVTE Created-by: clc2025 Commit-by: clc2025 Merged-by: ascend-robot Description: ## What this PR does / why we need it? 具体查看关联issue ## Does this PR introduce any user-facing change? NA ## How was this patch tested? 基于脚本用例，从GPU上保存权重，NPU加载权重后断点续训，精度能对齐且无功能报错 See merge request: Ascend/MindSpeed!3569 | 3 个月前 | |
| 1 年前 | ||
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!3496 | 4 个月前 | |
add w8a16 quant Co-authored-by: wangjunhang<wangjunhang7@huawei.com> # message auto-generated for no-merge-commit merge: !3460 merge Dev_W8A16 into master add w8a16 quant Created-by: goodflower9 Commit-by: wangjunhang Merged-by: ascend-robot Description: What this PR does / why we need it? This PR adds W8A16 MXFP8 QAT support based on the existing QAT flow. Does this PR introduce any user-facing change? Yes. Users can enable W8A16 MXFP8 QAT with: --qat-scheme w8a16-mxfp8 Related doc: docs/zh/features/qat_quant.md How was this patch tested? Enable fake quantization by adding the parameter --qat-scheme w8a16-mxfp8 https://wiki.huawei.com/domains/76578/wiki/233229/WIKI2026051211063362 See merge request: Ascend/MindSpeed!3460 | 4 个月前 | |
feat(ut/qos/torch): 补充ut，修复代码遗漏BUG Co-authored-by: Klayyy<wanglei886@h-partners.com> # message auto-generated for no-merge-commit merge: !3309 merge master into master feat(ut/qos/torch): 补充ut，修复代码遗漏BUG Created-by: Klayyy Commit-by: Klayyy Merged-by: ascend-robot Description: 1.补充AI QOS特性feature UT 2.ut补充过程中，自检代码，修复BUG 2.1 torch_npu._C._distributed_c10d.ProcessGroupHCCL.Options()调用名称修改 2.2 qos_feature.py 中 raiseValueError 提示词完善 2.3 qos.py中对于最小冲突度组合中优先级的赋值部分，去掉重复代码，去掉无用库导入，_PARALLEL_TYPES中有逗号未添加 2.4 qos.py中 应是sdma qos 部分的处理，误使用roce 3.补充H2D QOS 对于 PCIE异步通道的使用，对于DCMI接口新建set_h2d_qos接口，提供给python调用 4.修改aiQos Readme中关于DCMI接口的调用，补充DCMI接口SO编译方法 See merge request: Ascend/MindSpeed!3309 | 6 个月前 | |
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!3162 | 8 个月前 | |
fix: use MindSpeed full args in MoE utils Co-authored-by: Zhangminjie<eddy_zmj@qq.com> # message auto-generated for no-merge-commit merge: !3575 merge zmj/fix-moe-get-args-initialization into master fix: use MindSpeed full args in MoE utils Created-by: MengyuyangEddy Commit-by: Zhangminjie Merged-by: ascend-robot Description: ## What this PR does / why we need it? This PR fixes an argument initialization issue in the MindSpeed MoE router path. When running the MoE router with the MindSpeed runtime, topk_softmax_with_capacity() calls get_args() through mindspeed/core/transformer/moe/moe_utils.py. The previous implementation imported get_args from Megatron: python from megatron.training import get_args However, in the MindSpeed/veRL runtime path, Megatron global args may not be initialized through the native Megatron training entrypoint. As a result, calling Megatron's get_args() can trigger the following error: text AssertionError: args is not initialized This PR replaces the import with MindSpeed's full-args helper: python from mindspeed.args_utils import get_full_args as get_args This keeps the existing get_args() call sites unchanged while avoiding the dependency on Megatron's uninitialized global _GLOBAL_ARGS. The change is limited to the argument source used by MoE utils. It does not change the MoE router algorithm, top-k routing logic, capacity calculation, token dispatching behavior, or expert computation. ## Does this PR introduce any user-facing change? No. This PR does not introduce any user-facing API or configuration change. Existing scripts and configs can continue to call the same MoE/MindSpeed paths without modification. The only behavior change is that MoE utils now obtain runtime arguments from MindSpeed's get_full_args() instead of Megatron's global get_args(), which fixes the args is not initialized failure in MindSpeed runtime scenarios. ## How was this patch tested?  pre-commit:  See merge request: Ascend/MindSpeed!3575 | 3 个月前 | |
feat: add last word feature Co-authored-by: Zhaiwenxuan<zhaiwenxuan4@h-partners.com> # message auto-generated for no-merge-commit merge: !3542 merge master into master feat: add last word feature Created-by: Zhaiwenxuan Commit-by: Zhaiwenxuan Merged-by: ascend-robot Description: ## What this PR does / why we need it? 新增适配verl训练场景的临终遗言功能 ## Does this PR introduce any user-facing change? YES ++actor_rollout_ref.actor.megatron.override_transformer_config.ttp_enabled=True 用户可通过增加++actor_rollout_ref.actor.megatron.override_transformer_config.ttp_enabled=True，开启适配verl训练场景的临终遗言功能 ## How was this patch tested? https://wiki.huawei.com/domains/181442/wiki/383165/WIKI2026060911405364 See merge request: Ascend/MindSpeed!3542 | 3 个月前 | |
| 2 年前 | ||
feat: mxfp8-32x32 quant Co-authored-by: kyle_zhangchi<zhangchi158@huawei.com> # message auto-generated for no-merge-commit merge: !3471 merge feat_mxfp8-32x32 into master feat: mxfp8-32x32 quant Created-by: kyle_zhangchi Commit-by: kyle_zhangchi Merged-by: ascend-robot Description: ## What this PR does / why we need it? 在Megatron框架下新增mxfp8-32x32量化算子，降低权重显存占用 ## Does this PR introduce *any* user-facing change? --fp8-recipe新增mxfp8-32x32选项 https://gitcode.com/Ascend/MindSpeed/commit/e065cbca6873bfc02661d088b07d90224333e87d?ref=feat_mxfp8-32x32&prId=3471 ## How was this patch tested? 验证文档 https://wiki.huawei.com/domains/170864/wiki/367830/WIKI2026051111046509 See merge request: Ascend/MindSpeed!3471 | 4 个月前 | |
| 1 年前 | ||
| 1 年前 | ||
| 1 年前 | ||
| 1 年前 | ||
| 1 年前 | ||
fixbug for auto_settings Co-authored-by: wuweiqiang24<wuweiqiang11@huawei.com> # message auto-generated for no-merge-commit merge: !3003 merge auto_settings into master fixbug for auto_settings Created-by: wuweiqiang24 Commit-by: wuweiqiang24 Merged-by: ascend-robot Description: 1. fixbug for auto_settings 2. add readme See merge request: Ascend/MindSpeed!3003 | 9 个月前 | |
| 1 年前 |
| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
| 7 个月前 | ||
| 1 年前 | ||
| 10 个月前 | ||
| 9 个月前 | ||
| 6 个月前 | ||
| 3 个月前 | ||
| 4 个月前 | ||
| 3 个月前 | ||
| 3 个月前 | ||
| 10 个月前 | ||
| 10 个月前 | ||
| 3 个月前 | ||
| 1 年前 | ||
| 4 个月前 | ||
| 4 个月前 | ||
| 6 个月前 | ||
| 8 个月前 | ||
| 3 个月前 | ||
| 3 个月前 | ||
| 2 年前 | ||
| 4 个月前 | ||
| 1 年前 | ||
| 1 年前 | ||
| 1 年前 | ||
| 1 年前 | ||
| 1 年前 | ||
| 9 个月前 | ||
| 1 年前 |