c721ce33创建于 2025年6月4日历史提交

文件	最后提交记录	最后更新时间
README.md	!2377 【Readme update!!!】资料修正 Merge pull request !2377 from yangjie/2.0.0_core_r0.8.0	1 年前
ffn.md	!2377 【Readme update!!!】资料修正 Merge pull request !2377 from yangjie/2.0.0_core_r0.8.0	1 年前
fusion_attention.md	!2377 【Readme update!!!】资料修正 Merge pull request !2377 from yangjie/2.0.0_core_r0.8.0	1 年前
gmm.md	!2377 【Readme update!!!】资料修正 Merge pull request !2377 from yangjie/2.0.0_core_r0.8.0	1 年前
lcal_coc.md	!2377 【Readme update!!!】资料修正 Merge pull request !2377 from yangjie/2.0.0_core_r0.8.0	1 年前
npu_all_to_all_all_gather_bmm.md	!2377 【Readme update!!!】资料修正 Merge pull request !2377 from yangjie/2.0.0_core_r0.8.0	1 年前
npu_apply_fused_ema_adamw.md	!1689 update readme for ema adam op Merge pull request !1689 from wangyuansheng8/master	1 年前
npu_bmm_reduce_scatter_all_to_all.md	!1703 mc2 moe operator-shard H doc Merge pull request !1703 from 邵燕波/master	1 年前
npu_dropout_add_layer_norm.md	!371 MoE MLP not use mc2 linear and remove ops from init Merge pull request !371 from Jializheng/master	2 年前
npu_fused_moe_token_permute.md	!822 docs(docs/ops/npu_fused_moe_token_permute.md) Merge pull request !822 from 邓佳/master_docs	1 年前
npu_fused_moe_token_unpermute.md	!2377 【Readme update!!!】资料修正 Merge pull request !2377 from yangjie/2.0.0_core_r0.8.0	1 年前
npu_grouped_mat_mul_all_reduce.md	!2377 【Readme update!!!】资料修正 Merge pull request !2377 from yangjie/2.0.0_core_r0.8.0	1 年前
npu_groupmatmul_add.md	!1288 GMM+Add 适配 Merge pull request !1288 from zengshu/master	1 年前
npu_matmul_add.md	!1288 GMM+Add 适配 Merge pull request !1288 from zengshu/master	1 年前
npu_mm_all_reduce_add_rms_norm.md	!2377 【Readme update!!!】资料修正 Merge pull request !2377 from yangjie/2.0.0_core_r0.8.0	1 年前
npu_mm_all_reduce_add_rms_norm_.md	!2377 【Readme update!!!】资料修正 Merge pull request !2377 from yangjie/2.0.0_core_r0.8.0	1 年前
npu_ring_attention_update.md	!2319 MindSpeed 版本维护策略更新 Merge pull request !2319 from wangruiqi/2.0.0_core_r0.8.0	1 年前
npu_rotary_position_embedding.md	!2377 【Readme update!!!】资料修正 Merge pull request !2377 from yangjie/2.0.0_core_r0.8.0	1 年前
quant_gmm.md	!2377 【Readme update!!!】资料修正 Merge pull request !2377 from yangjie/2.0.0_core_r0.8.0	1 年前
rms_norm.md	!2377 【Readme update!!!】资料修正 Merge pull request !2377 from yangjie/2.0.0_core_r0.8.0	1 年前
swiglu.md	!359 change ascendspeed to mindspeed Merge pull request !359 from 邓佳/master	2 年前
weight_quant_gmm.md	!2377 【Readme update!!!】资料修正 Merge pull request !2377 from yangjie/2.0.0_core_r0.8.0	1 年前

How to run the ops?

previous installation

CANN
CANN-NNAL(Ascend-Transformer-Boost)
torch_npu

compile and install

1. set the environment variables

# Default path, change it if needed.
source /usr/local/Ascend/ascend-toolkit/set_env.sh

if use Ascend-Transformer-Boost

# Default path, change it if needed.
source /usr/local/Ascend/nnal/atb/set_env.sh

2. include header files

newest torch_npu
newest cann

3. install scripts

python3 setup.py build
python3 setup.py bdist_wheel
pip3 install dist/*.whl --force-reinstall