add supplementary content to the npu quick start doc #6727
Conversation
Summary of ChangesHello @addsubmuldiv, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed! 此拉取请求旨在通过添加补充内容来显著增强NPU快速入门文档。主要更新包括详细的环境安装指南、硬件支持信息，以及关于SFT和RL算法在NPU上支持现状的全面表格。此外，还明确指出了当前NPU不支持或未完全验证的模块。这些改进将为用户提供更清晰、更全面的NPU使用指南，帮助他们更有效地进行模型微调、推理和部署。 Highlights
Using Gemini Code AssistThe full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips. Invoking Gemini You can request assistance from Gemini at any point by creating a comment using either
Customization To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a Limitations & Feedback Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here. You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension. Footnotes
|
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request adds significant supplementary content to the NPU quick start guide, including installation instructions, hardware requirements, and support status tables. The additions are valuable for users looking to use NPUs. My review includes several suggestions to improve the document's structure, clarity, and formatting, such as filling in empty sections, correcting heading levels, fixing a potential typo, and tidying up excessive blank lines. Addressing these points will enhance the readability and professionalism of the documentation.
| ## 硬件支持 | ||
|
|
||
|
|
There was a problem hiding this comment.
The "硬件支持" (Hardware Support) section is currently empty. It would be helpful to list the supported hardware models (e.g., Ascend 910B) or add a placeholder like "待补充" (To be added) to clarify that this information is forthcoming and avoid confusion.
|
|
||
|
|
There was a problem hiding this comment.
There are multiple consecutive blank lines here and in several other places throughout the document (e.g., lines 40, 46, 57, 82, 118, etc.). It's recommended to remove these extra blank lines and use only a single blank line to separate sections and paragraphs. This will make the document more compact and improve readability.
| | torch_npu | == 2.7.1 | | ||
|
|
||
|
|
||
| 基础环境准备请参照这份[文档](https://gitcode.com/Ascend/pytorch) |
There was a problem hiding this comment.
The link text "文档" (document) is not very descriptive. Consider using more specific text, such as "Ascend PyTorch Installation Guide", to give users a better idea of what the link points to before they click it.
| 基础环境准备请参照这份[文档](https://gitcode.com/Ascend/pytorch) | |
| 基础环境准备请参照这份[Ascend PyTorch 安装文档](https://gitcode.com/Ascend/pytorch) |
|
|
||
| ## 支持现状 | ||
|
|
||
| ## 表 1：SFT 类算法 |
There was a problem hiding this comment.
The headings for "表 1" (Table 1), "表 2" (Table 2, line 286), and "表 3" (Table 3, line 299) use the same level (##) as the main section "支持现状" (Support Status, line 265). For better document structure and hierarchy, it's recommended to demote the table headings to a lower level (e.g., ###) to make them subsections of "支持现状".
| ## 表 1：SFT 类算法 | |
| ### 表 1：SFT 类算法 |
|
|
||
| | item | | ||
| | ---------------------- | | ||
| | Liger-kernal | |
There was a problem hiding this comment.
"Liger-kernal" appears to be a typo. Did you mean "Liger-kernel"?
| | Liger-kernal | | |
| | Liger-kernel | |
| | Liger-kernal | | ||
| | 量化/QLoRA相关 | | ||
| | Megatron相关 | | ||
| | 使用sglang作为推理引擎 | |
There was a problem hiding this comment.
The file is missing a final newline at the end. It's a common convention to end text files with a single newline character.
There was a problem hiding this comment.
Pull request overview
This PR adds supplementary content to the NPU quick start documentation, enhancing the guide for using ms-swift with Ascend NPU hardware. The changes provide more comprehensive information about installation requirements, supported algorithms, and hardware compatibility.
Key Changes:
- Added introduction section explaining NPU support capabilities
- Added installation section with software version requirements table
- Added support status section with three detailed tables covering SFT algorithms, RL algorithms, and unsupported modules
- Improved formatting consistency throughout the document with additional spacing between sections
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
|
|
||
|
|
There was a problem hiding this comment.
The "硬件支持" (Hardware Support) section is empty. Consider adding content describing the supported NPU hardware models, or remove this section if it's not needed.
| 当前已支持以下昇腾 NPU 硬件型号： | |
| - 昇腾 910B3 64G | |
| 如需支持其他型号，请参考昇腾官方文档或联系我们。 |
|
|
||
| | item | | ||
| | ---------------------- | | ||
| | Liger-kernal | |
There was a problem hiding this comment.
Spelling error: "Liger-kernal" should be "Liger-kernel" (kernel not kernal).
| | Liger-kernal | | |
| | Liger-kernel | |
| ## 表 1：SFT 类算法 | ||
|
|
||
| | algorithm | model families | strategy | hardware | | ||
| | --------- | --------------------------- | --------------------- | ----------------- | | ||
| | SFT | Qwen2.5-0.5B-Instruct | FSDP1/FSDP2/deepspeed | Atlas 900 A2 PODc | | ||
| | SFT | Qwen2.5-1.5B-Instruct | FSDP1/FSDP2/deepspeed | Atlas 900 A2 PODc | | ||
| | SFT | Qwen2.5-7B-Instruct | FSDP1/FSDP2/deepspeed | Atlas 900 A2 PODc | | ||
| | SFT | Qwen2.5-VL-3B-Instruct | FSDP1/FSDP2/deepspeed | Atlas 900 A2 PODc | | ||
| | SFT | Qwen2.5-VL-7B-Instruct | FSDP1/FSDP2/deepspeed | Atlas 900 A2 PODc | | ||
| | SFT | Qwen2.5-Omni-3B | FSDP1/FSDP2/deepspeed | Atlas 900 A2 PODc | | ||
| | SFT | Qwen3-8B | FSDP1/FSDP2/deepspeed | Atlas 900 A2 PODc | | ||
| | SFT | Qwen3-32B | FSDP1/FSDP2/deepspeed | Atlas 900 A2 PODc | | ||
| | SFT | Qwen3-VL-30B-A3B-Instruct | FSDP1/FSDP2/deepspeed | Atlas 900 A2 PODc | | ||
| | SFT | Qwen3-Omni-30B-A3B-Instruct | FSDP1/FSDP2/deepspeed | Atlas 900 A2 PODc | | ||
| | SFT | InternVL3-8B | FSDP1/FSDP2/deepspeed | Atlas 900 A2 PODc | | ||
| | SFT | Ovis2.5-2B | FSDP1/FSDP2/deepspeed | Atlas 900 A2 PODc | | ||
|
|
||
| ------ | ||
|
|
||
| ## 表 2：RL 类算法 | ||
|
|
||
| | algorithm | model families | strategy | rollout engine | hardware | | ||
| | --------- | ------------------- | --------- | -------------- | ----------------- | | ||
| | **GRPO** | Qwen2.5-7B-Instruct | deepspeed | vllm-ascend | Atlas 900 A2 PODc | | ||
| | **GRPO** | Qwen3-8B | deepspeed | vllm-ascend | Atlas 900 A2 PODc | | ||
| | **DPO** | Qwen2.5-7B-Instruct | deepspeed | vllm-ascend | Atlas 900 A2 PODc | | ||
| | **DPO** | Qwen3-8B | deepspeed | vllm-ascend | Atlas 900 A2 PODc | | ||
| | **PPO** | Qwen2.5-7B-Instruct | deepspeed | vllm-ascend | Atlas 900 A2 PODc | | ||
| | **PPO** | Qwen3-8B | deepspeed | vllm-ascend | Atlas 900 A2 PODc | | ||
|
|
||
| --- | ||
|
|
||
| ## 表 3：当前 NPU 暂不支持 / 未完全验证的模块 |
There was a problem hiding this comment.
[nitpick] The section heading structure is inconsistent. "表 1：SFT 类算法" (Table 1: SFT Algorithms), "表 2：RL 类算法" (Table 2: RL Algorithms), and "表 3：当前 NPU 暂不支持 / 未完全验证的模块" (Table 3: Currently Unsupported/Unverified Modules) should be subsections (###) under "支持现状" (Support Status) rather than top-level sections (##). This would better organize the document hierarchy.
| | **PPO** | Qwen2.5-7B-Instruct | deepspeed | vllm-ascend | Atlas 900 A2 PODc | | ||
| | **PPO** | Qwen3-8B | deepspeed | vllm-ascend | Atlas 900 A2 PODc | | ||
|
|
||
| --- |
There was a problem hiding this comment.
[nitpick] Inconsistent horizontal rule formatting. Line 284 uses "------" (6 dashes) while line 297 uses "---" (3 dashes). For consistency, use the same number of dashes throughout the document.
| --- | |
| ------ |
addsubmuldiv
force-pushed
the
npu_quick_start_doc
branch
from
4fba40f to
f5dc65d
Compare
2 participants
PR type
PR information
Write the detail information belongs to this PR.
Experiment results
Paste your experiment result here(if needed).