From d743be016967ba8c5928cada43b53d4f1d0b2f93 Mon Sep 17 00:00:00 2001 From: Yuan Fang Date: Tue, 7 Jul 2026 23:33:40 +0800 Subject: [PATCH] docs: clarify Ascend soft slicing resource requests Signed-off-by: Yuan Fang --- README.md | 3 ++- README_cn.md | 2 +- 2 files changed, 3 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index a713c3b..b3d88bb 100644 --- a/README.md +++ b/README.md @@ -139,13 +139,14 @@ metadata: huawei.com/Ascend910B: "1" # if you don't specify Ascend910B-memory, it will use a whole NPU. huawei.com/Ascend910B-memory: "4096" + huawei.com/Ascend910B-core: "40" ``` For more examples, see [examples](https://github.com/Project-HAMi/ascend-device-plugin/tree/main/examples) ### Soft Slicing Configuration (HAMi) -Use the annotation below whenever you intend **soft** slicing; omitting it keeps **template-based vNPU** behavior (see the note under [Usage in HAMi](#usage-in-hami)). +Use the annotation below whenever you intend **soft** slicing; omitting it keeps **template-based vNPU** behavior (see the note under [Usage in HAMi](#usage-in-hami)). Soft-sliced workloads should request both **memory** and **core** resources; if only memory is requested, memory may be calculated by hard-slicing template matching, so the actual allocated value may be greater than or equal to the requested value instead of strictly equal to it. ```yaml apiVersion: v1 diff --git a/README_cn.md b/README_cn.md index 7acfcc5..35834ef 100644 --- a/README_cn.md +++ b/README_cn.md @@ -135,7 +135,7 @@ devices: ### 软切分配置 (HAMi) -需要 **软切分** 时请显式加上下文中的注解；不加则仍为 **模板硬切分 vNPU**（与上一节「在 HAMi 中使用」中的说明一致）。 +需要 **软切分** 时请显式加上下文中的注解；不加则仍为 **模板硬切分 vNPU**（与上一节「在 HAMi 中使用」中的说明一致）。软切分任务应同时申请**显存**和**算力**资源；如果只申请显存，显存可能按硬切分模板匹配方式计算，实际分配值大于或等于申请值，而不是严格等于申请值。 ```yaml apiVersion: v1