ModelZoo-PyTorch:基于昇腾生态的AI模型平台项目

用户可获取多方向AI模型及昇腾机器实操案例,用于模型训练与推理。项目涵盖计算机视觉、自然语言处理等领域,提供详细使用指导,持续增加典型网络和预训练模型。【此简介由AI生成】

分支2Tags0
文件最后提交记录最后更新时间
add .gitee/PULL_REQUEST_TEMPLATE.en.md. Signed-off-by: 王姜奔 <wangjiangben@huawei.com>3 年前
Canary-1B推理适配 Co-authored-by: peichengbo<peichengbo@huawei.com> # message auto-generated for no-merge-commit merge: !7592 merge master into master Canary-1B推理适配 Created-by: peichengbo Commit-by: peichengbo Merged-by: ascend-robot Description: ## Motivation Canary-1B推理适配代码合入,适用于A2 ## Modification 包含所有适配的代码,以及使用文档 ## Self-test (Optional) 性能测试: | 硬件 | 数据集 | RTF | RTFx | WER| |---|---|---:|---:|---:| | Atlas 800I A2 | LibriSpeech test-clean | 0.003553452 | 281.416 | 0.014728| ![图片.png](https://raw.gitcode.com/user-images/assets/8112803/c8179ff2-e5a2-402a-876d-e8ee3de60989/图片.png '图片.png') 精度测试: | 任务类型 | 语言 | 数据集 | 指标 | 得分 | 竞品 | | -------- | ----- | ------------------------ | ------ | ----- | ----- | | ASR | de | Multilingual LibriSpeech | WER(%) | 3.83 | 4.19 | | ASR | es | Multilingual LibriSpeech | WER(%) | 2.30 | 3.15 | | ASR | fr | Multilingual LibriSpeech | WER(%) | 3.69 | 4.12 | | AST | en-de | Fleurs | BLEU | 31.41 | 32.15 | | AST | en-es | Fleurs | BLEU | 22.69 | 22.66 | | AST | en-fr | Fleurs | BLEU | 39.84 | 40.76 | | AST | de-en | Fleurs | BLEU | 33.50 | 33.98 | | AST | es-en | Fleurs | BLEU | 21.78 | 21.80 | | AST | fr-en | Fleurs | BLEU | 30.29 | 30.95 | ![图片.png](https://raw.gitcode.com/user-images/assets/8112803/2550537a-c613-41b0-bfba-10931578d88d/图片.png '图片.png') ![图片.png](https://raw.gitcode.com/user-images/assets/8112803/c9cd644e-915b-479b-9623-2c0aade2b9a4/图片.png '图片.png') ![图片.png](https://raw.gitcode.com/user-images/assets/8112803/14af1c69-1377-4257-aa46-b16be3419c66/图片.png '图片.png') ![图片.png](https://raw.gitcode.com/user-images/assets/8112803/406c578d-c81c-47c1-8c4f-942d5031c10a/图片.png '图片.png') ![图片.png](https://raw.gitcode.com/user-images/assets/8112803/15cb864a-19e7-4dfd-9e88-bc3835379a86/图片.png '图片.png') ![图片.png](https://raw.gitcode.com/user-images/assets/8112803/2bab6346-59f2-4c11-aaab-1a4986336ecc/图片.png '图片.png') ![图片.png](https://raw.gitcode.com/user-images/assets/8112803/5748fd26-a101-4d3b-8083-639b7fd34dae/图片.png '图片.png') ![图片.png](https://raw.gitcode.com/user-images/assets/8112803/aa781eec-bd65-4cb3-bf62-5be5e0e2edae/图片.png '图片.png') ![图片.png](https://raw.gitcode.com/user-images/assets/8112803/3d0c8753-e5f7-4d56-be91-78e0e57aa2a3/图片.png '图片.png') 官方精度:https://huggingface.co/nvidia/canary-1b ![图片.png](https://raw.gitcode.com/user-images/assets/8112803/dd50e286-c795-4097-acdb-f048d7f0c5e0/图片.png '图片.png') ![图片.png](https://raw.gitcode.com/user-images/assets/8112803/6cdadec8-cfbc-4f4c-8cfa-8d1246f1a611/图片.png '图片.png') ## PR 信息 https://gitcode.com/Ascend/ModelZoo-PyTorch/issues/2673 ## BC-breaking (Optional) 不涉及 ## Checklist **Before PR**: - [x] The new code needs to comply with the Clean Code specification. - [x] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [ ] CLA has been signed and all committers have signed the CLA in this PR. - [ ] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/ModelZoo-PyTorch!75921 天前
适配FlashVSR-Pro模型 Co-authored-by: socrahow<suzihao4@h-partners.com> # message auto-generated for no-merge-commit merge: !7526 merge szh-pro into master 适配FlashVSR-Pro模型 Created-by: socrahow1 Commit-by: socrahow1;socrahow Merged-by: ascend-robot Description: ## Motivation 适配FlashVSR-Pro模型。 ## Modification 适配FlashVSR-Pro模型。 ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist **Before PR**: - [ ] The new code needs to comply with the Clean Code specification. - [ ] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [ ] CLA has been signed and all committers have signed the CLA in this PR. - [ ] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/ModelZoo-PyTorch!75261 个月前
更新MAPPO模型FAQ Co-authored-by: mamba-chen<chenhao388@huawei.com> # message auto-generated for no-merge-commit merge: !7597 merge mappo into master 更新MAPPO模型FAQ Created-by: mamba-chen Commit-by: mamba-chen Merged-by: ascend-robot Description: ## Motivation 更新MAPPO模型FAQ ## Modification 更新MAPPO模型FAQ ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist **Before PR**: - [ ] The new code needs to comply with the Clean Code specification. - [ ] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [ ] CLA has been signed and all committers have signed the CLA in this PR. - [ ] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/ModelZoo-PyTorch!75973 天前
!7390 add DeepSeek-R1-Safe Merge pull request !7390 from ZJUAISafety/master 8 个月前
Update model migration documentation Co-authored-by: xvxuopop<xvxuopop@gmail.com> # message auto-generated for no-merge-commit merge: !7545 merge docs into master Update model migration documentation Created-by: jin-yongxu Commit-by: xvxuopop Merged-by: ascend-robot Description: ## Motivation Update model migration documentation ## Checklist **Before PR**: - [x] The new code needs to comply with the Clean Code specification. - [x] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [x] CLA has been signed and all committers have signed the CLA in this PR. - [ ] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/ModelZoo-PyTorch!754520 天前
update readme Co-authored-by: frozenleaves<frozen@Mac.local> # message auto-generated for no-merge-commit merge: !7496 merge master into master update readme Created-by: frozenn Commit-by: frozenleaves Merged-by: ascend-robot Description: ## Motivation Please describe the motivation of this PR and the goal you want to achieve through this PR. ## Modification Please briefly describe what modification is made in this PR. ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist **Before PR**: - [ ] The new code needs to comply with the Clean Code specification. - [ ] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [ ] CLA has been signed and all committers have signed the CLA in this PR. - [ ] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/ModelZoo-PyTorch!74962 个月前
init 4 年前
init 4 年前
!7380 add owner reviewer Merge pull request !7380 from lijian/add_owner_reviewer 8 个月前
Update model migration documentation Co-authored-by: xvxuopop<xvxuopop@gmail.com> # message auto-generated for no-merge-commit merge: !7545 merge docs into master Update model migration documentation Created-by: jin-yongxu Commit-by: xvxuopop Merged-by: ascend-robot Description: ## Motivation Update model migration documentation ## Checklist **Before PR**: - [x] The new code needs to comply with the Clean Code specification. - [x] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [x] CLA has been signed and all committers have signed the CLA in this PR. - [ ] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/ModelZoo-PyTorch!754520 天前
!6900 修改ModelZoo-PyTorch中的Third_Party_Open_Source_Software_Notice文件 Merge pull request !6900 from 王凯宇/master 1 年前

ModelZoo-PyTorch

简介

ModelZoo-PyTorch,昇腾旗下的开源AI模型平台,涵盖计算机视觉、自然语言处理、语音、推荐、多模态、大语言模型等方向的AI模型及其基于昇腾机器实操案例。平台的每个模型都有详细的使用指导,为方便更多开发者使用ModelZoo-PyTorch,我们将持续增加典型网络和相关预训练模型。如果您有任何需求,请在GitCode提交issue,我们会及时处理。

目录结构

关键目录如下,详细目录介绍参见如下说明:

├── ACL_PyTorch                    # 基于昇腾芯片的推理模型参考
├── MindIE                         # 基于昇腾芯片的推理引擎模型参考
├── PyTorch                        # 基于昇腾芯片的训练模型参考
├── CONTRIBUTING_TRAIN.md          # 模型训练阶段代码提交规范
├── LICENSE                        # 许可证文件
├── NOTICE                         # 声明文件
└── Third_Party_Open_Source_Software_Notice # 第三方开源软件声明

贡献指导

欢迎大家为社区做贡献。如果有任何疑问或建议,请提交Issues,我们会尽快回复。感谢您的支持。

  1. 提交错误报告:如果您在ModelZoo-PyTorch中发现了一个不存在安全问题的漏洞,请在仓库中的Issues中搜索,以防该漏洞被重复提交,如果找不到漏洞可以创建一个新的Issues。
  2. 解决现有问题:通过查看仓库的Issues列表可以发现需要处理的问题信息, 可以尝试解决其中的某个问题。
  3. 如何提出新功能:请使用Issues的Feature标签进行标记,我们会定期处理和确认开发。
  4. 开始贡献: a. Fork本项目的仓库。 b. Clone到本地。 c. 创建开发分支。 d. 本地自测:提交前请通过所有的单元测试。 e. 提交代码。 f. 新建Pull Request。 g. 代码检视:您需要根据评审意见修改代码,并再次推送更新。此过程可能会有多轮。 h. 当您的PR获得足够数量的检视者批准后,Committer会进行最终审核。 i. 审核和测试通过后,CI会将您的PR合并入到项目的主干分支。

【重要】模型训练阶段代码提交规范请阅读 CONTRIBUTING_TRAIN

安全声明

运行用户建议

出于安全性及权限最小化角度考虑,不建议使用root等管理员类型账户使用。

文件权限控制

  1. 建议用户在主机(包括宿主机)及容器中设置运行系统umask值为0027及以上,保障新增文件夹默认最高权限为750,新增文件默认最高权限为640。

  2. 建议用户对个人数据、商业资产、源文件、训练过程中保存的各类文件等敏感内容做好权限管控,管控权限可参考表1进行设置。

    表1 文件(夹)各场景权限管控推荐最大值

    类型 linux权限参考最大值
    用户主目录 750(rwxr-x---)
    程序文件(含脚本文件、库文件等) 550(r-xr-x---)
    程序文件目录 550(r-xr-x---)
    配置文件 640(rw-r-----)
    配置文件目录 750(rwxr-x---)
    日志文件(记录完毕或者已经归档) 440(r--r-----)
    日志文件(正在记录) 640(rw-r-----)
    日志文件目录 750(rwxr-x---)
    Debug文件 640(rw-r-----)
    Debug文件目录 750(rwxr-x---)
    临时文件目录 750(rwxr-x---)
    维护升级文件目录 770(rwxrwx---)
    业务数据文件 640(rw-r-----)
    业务数据文件目录 750(rwxr-x---)
    密钥组件、私钥、证书、密文文件目录 700(rwx—----)
    密钥组件、私钥、证书、加密密文 600(rw-------)
    加解密接口、加解密脚本 500(r-x------)

运行安全声明

建议用户结合运行环境资源状况编写对应训练脚本。若训练脚本与资源状况不匹配,如数据集加载内存大小超出内存容量限制、训练脚本在本地生成数据超过磁盘空间大小等情况,可能引发错误并导致进程意外退出。

公网地址声明

详见各模型目录下的public_address_statement.md

公开接口声明

ModelZoo-PyTorch仅提供模型代码,不涉及构建打包发布,无公开接口,需按模型文档介绍,通过模型内shell脚本使用。

免责声明

致ModelZoo使用者

  1. ModelZoo提供的模型仅供您用于非商业目的。
  2. 对于各模型,ModelZoo平台仅提示性地向您建议可用于训练的数据集,华为不提供任何数据集,如您使用这些数据集进行训练,请您特别注意应遵守对应数据集的License,如您因使用数据集而产生侵权纠纷,华为不承担任何责任。
  3. 如您在使用ModelZoo模型过程中,发现任何问题(包括但不限于功能问题、合规问题),请在Gitcode提交issue,我们将及时审视并解决。

致数据集所有者

如果您不希望您的数据集在ModelZoo中的模型被提及,或希望更新ModelZoo中的模型关于您的数据集的描述,请在Gitcode提交issue,我们将根据您的issue要求删除或更新您的数据集描述。衷心感谢您对ModelZoo的理解和贡献。

License

ModelZoo-PyTorch产品的使用许可证,具体请参见LICENSE
Ascend ModelZoo提供的模型,如模型目录下存在License的,以该License为准。如模型目录下不存在License的,以Apache 2.0许可证许可,对应许可证文本可查阅Ascend ModelZoo根目录。

Ascend ModelZoo产品docs目录下的文档适用CC-BY 4.0许可证,具体请参见LICENSE

项目介绍

用户可获取多方向AI模型及昇腾机器实操案例,用于模型训练与推理。项目涵盖计算机视觉、自然语言处理等领域,提供详细使用指导,持续增加典型网络和预训练模型。【此简介由AI生成】

定制我的领域

下载使用量

0

项目总下载次数(含Clone、Pull、 zip 包及 release 下载),每日凌晨更新

语言类型

Python88.73%
Shell6.71%
C++2.2%
Cuda0.92%
Jupyter Notebook0.4%