-
Notifications
You must be signed in to change notification settings - Fork 274
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
* Update 【Hackathon_8th】个人挑战赛—套件开发任务合集.md * Update 【Hackathon_8th】开源贡献个人挑战赛rfc需求列表.md * Create 【Hackathon 8th】FundableProject任务合集.md * fix --------- Co-authored-by: zachary sun <[email protected]>
- Loading branch information
1 parent
8331fdd
commit 1a6224d
Showing
4 changed files
with
139 additions
and
3 deletions.
There are no files selected for viewing
54 changes: 54 additions & 0 deletions
54
hackathon/hackathon_8th/【Hackathon_8th】FundableProject任务合集.md
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,54 @@ | ||
此文档展示 **PaddlePaddle Hackathon 第八期活动——Fundable Projects** 任务详细介绍。Fundable Projects 赛道定位硬核任务,要求高水平的开发者独立进行任务拆解和完成。 | ||
|
||
## 产出要求 | ||
|
||
- 任务拆解 tracking issue | ||
- 答辩 PPT | ||
- 书面的技术报告 | ||
- 代码运行无误,通过社区 maintainers 的评审并合入代码仓库。 | ||
|
||
## 任务详情 | ||
|
||
### 一、PaddleSpeech 套件能力建设 - 模型精度对齐 | ||
|
||
**任务背景**: | ||
|
||
PaddleSpeech 是基于飞桨 PaddlePaddle 的语音方向的开源套件,囊括语音识别、语音合成、语音唤醒、声纹识别等多种语音常用功能的支持。由于近期 Paddle 新版本的升级存在不兼容部分(如 `paddle.fluid` API 全面退场,PIR + predictor 升级, 0-d tensor,view 行为修改等),需要重新对 PaddleSpeech 中的模型进行适配开发与回归测试,保证套件正常运转,模型功能与精度不受损失。外部开发者需要做的事情包括: | ||
|
||
**详细描述:** | ||
|
||
基于 Paddle 3.0.0-beta 版本对 PaddleSpeech 进行适配升级,梳理已有堵点并解决。保证[example](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples) 目录下核心模型在 新 Paddle 版本 & 新其他深度学习框架版本下的正常运转。目前适配版本为 Paddle 2.5.1。 | ||
|
||
**验收说明:** | ||
|
||
PaddleSpeech 基于 Paddle 3.0.0-beta 版本,完成 10+ 原有模型的适配和精度对齐。 | ||
|
||
**技术要求:** | ||
|
||
- 熟悉 Python,工程能力强 | ||
- 对语音识别或合成有一定了解,有训练或者研发经验(加分项) | ||
- 对 PaddleSpeech 套件比较熟悉(加分项) | ||
|
||
**参考资料:** https://github.com/PaddlePaddle/PaddleSpeech | ||
|
||
### 二、PaddleSpeech 套件能力建设 - PIR 导出 | ||
|
||
**任务背景**: | ||
|
||
PaddleSpeech 是基于飞桨 PaddlePaddle 的语音方向的开源套件,囊括语音识别、语音合成、语音唤醒、声纹识别等多种语音常用功能的支持。由于近期 Paddle 新版本的升级存在不兼容部分(如 `paddle.fluid` API 全面退场,PIR + predictor 升级, 0-d tensor,view 行为修改等),需要重新对 PaddleSpeech 中的模型进行适配开发与回归测试,保证套件正常运转,模型功能与精度不受损失。外部开发者需要做的事情包括: | ||
|
||
**详细描述:** | ||
|
||
基于 Paddle 3.0.0-beta 版本对 PaddleSpeech 中支持转静的模型重新按照 PIR + predictor 的方式导出,并成功推理。 | ||
|
||
**验收说明:** | ||
|
||
PaddleSpeech 基于 Paddle 3.0.0-beta 版本,完成 20+ 原有静态图模型的重新导出和上传。 | ||
|
||
**技术要求:** | ||
|
||
- 熟悉 Python,工程能力强 | ||
- 对语音识别或合成有一定了解,有训练或者研发经验(加分项) | ||
- 对 PaddleSpeech 套件比较熟悉(加分项) | ||
|
||
**参考资料:** https://github.com/PaddlePaddle/PaddleSpeech |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters