Skip to content

【开源实习】针对任务类型Text-to-Audio,开发可在香橙派AIpro开发板运行的应用 #1854 - #137

Open
uMemory wants to merge 3 commits into
candle-org:masterfrom
uMemory:feature/EnglishTextToAudio
Open

【开源实习】针对任务类型Text-to-Audio,开发可在香橙派AIpro开发板运行的应用 #1854#137
uMemory wants to merge 3 commits into
candle-org:masterfrom
uMemory:feature/EnglishTextToAudio

Conversation

@uMemory

@uMemory uMemory commented Apr 27, 2026

Copy link
Copy Markdown

Description: 开源实习任务#1854

在香橙派 AIpro (Ascend 310B) 开发板上实现了基于 MindNLP 框架的英文文本转语音(Text-to-Audio)应用。

项目背景:
利用 SpeechT5 模型与 HiFi-GAN 声码器,实现高质量的英文语音合成。针对 Ascend 310B 的硬件特性与算子限制,本 PR 提供了针对 MindNLP 0.4.1 库文件的关键适配补丁,确保在 NPU 环境下能够顺畅运行。

Related Issue: Fixes #1854 (Ref: mindspore/community#1854).

Changes:

  1. New Module: 路径为 Online/community/EnglishTextToAudio
  2. Core Artifacts:
    • mindspore_EnglishTextToAudio.ipynb: 包含推理流程、库文件适配说明及交互式界面的 Notebook。
    • README.md: 提供详细的部署指南、环境准备说明及 4 处关键源码适配补丁的详细指引。
  3. Environment Fixes: 文档中记录了对 mindnlp/core/ops/blas.pymodeling_speecht5.pymindnlp/core/ops/array.py 的修改逻辑,解决了 BatchMatMulExt 形状检查错误及 nonzero 算子参数不兼容问题。

Testing & Benchmark:

  • 硬件: Orange Pi AIpro 20T24G
  • 软件环境: CANN 8.1.RC1, MindSpore 2.6.0, MindNLP 0.4.1
  • 功能验证: 经实机测试,应用可成功加载模型,输入英文文本后可生成自然流畅的音频。支持切换 7931 种说话人音色,交互式 Demo 运行正常。
  • 性能表现: 模型总参数量约 158.3M,内存占用稳定,推理速度满足端侧实时交互需求。

Checklist

  • Read and followed the Contributing Guidelines.
  • Self-tested locally to ensure the code runs correctly and achieves expected results (all CI checks expected to pass).
  • Updated documentation if needed.
  • Verified accuracy or performance benchmarks if applicable.

@uMemory
uMemory force-pushed the feature/EnglishTextToAudio branch from 0a4b57c to 5c066e6 Compare April 27, 2026 15:30
@cui0523

cui0523 commented May 18, 2026

Copy link
Copy Markdown
Contributor

ipynb的名称辛苦规范成蛇形,小写+下划线的方式

@cui0523

cui0523 commented May 18, 2026

Copy link
Copy Markdown
Contributor

readme文件中的mindnlp修改成Mindspore NLP

@uMemory

uMemory commented May 18, 2026

Copy link
Copy Markdown
Author

readme文件中的mindnlp修改成Mindspore NLP

已按要求规范命名.ipynb文件,mindnlp除源码适配部分的路径说明外,均已修改为Mindspore NLP

@cui0523

cui0523 commented May 19, 2026

Copy link
Copy Markdown
Contributor

验收通过,可以合入

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants