MindSpeed-RL/mindspeed_rl · Ascend/MindSpeed-RL

ascend-robot[copyright] add copyright head for some code files

文件	最后提交记录	最后更新时间
config_cls	Modify the document using the docs-ci tool Co-authored-by: nuerxiati<738457498@qq.com> # message auto-generated for no-merge-commit merge: !919 merge fix_docs into master Modify the document using the docs-ci tool Created-by: NurxatAbilmit Commit-by: NurxatAbilmit;nuerxiati Merged-by: ascend-robot Description: What this PR does / why we need it? 1. We use the docs-ci tool to scan and standardize the document。 2. Added the parameter tranformer impl: "local" to fix the error that occurred when MindSpeed-LLM ran RL 3. The requirements.txt has added packages bytecode that are missing in the environment Does this PR introduce any user-facing change? update docs, requirements and MindSpeed-LLM parameter How was this patch tested? The environment and the MindSpeed-LLM script were tested locally See merge request: Ascend/MindSpeed-RL!919	3 个月前
datasets	master资料修改 Co-authored-by: xiazhahe<nieshiyu1@huawei.com> # message auto-generated for no-merge-commit merge: !767 merge master into master master资料修改 Created-by: xiazhahe Commit-by: xiazhahe Merged-by: ascend-robot Description: master资料修改 See merge request: Ascend/MindSpeed-RL!767	7 个月前
models	[requirement] update ray requirement module to v2.54.0 Co-authored-by: zjchenn<zjchenn@gmail.com> # message auto-generated for no-merge-commit merge: !923 merge update-ray into master [requirement] update ray requirement module to v2.54.0 Created-by: zjchenn Commit-by: zjchenn Merged-by: ascend-robot Description: [requirement] update ray requirement module to v2.54.0 See merge request: Ascend/MindSpeed-RL!923	3 个月前
tools	[copyright] add copyright head for some code files Co-authored-by: zjchenn<zjchenn@gmail.com> # message auto-generated for no-merge-commit merge: !924 merge add-copyright into master [copyright] add copyright head for some code files Created-by: zjchenn Commit-by: zjchenn Merged-by: ascend-robot Description: [copyright] add copyright head for some code files 安全开源证书整改，为与开源代码雷同的文件增加 copyright 头 See merge request: Ascend/MindSpeed-RL!924	3 个月前
trainer	Refactor comments for folder models Co-authored-by: psyloy<zhangjianxiang3@huawei.com> # message auto-generated for no-merge-commit merge: !921 merge master into master Refactor comments for folder models Created-by: McFlurry100 Commit-by: psyloy Merged-by: ascend-robot Description: 添加模块化注释，未修改非注释代码，修改前后功能等效 See merge request: Ascend/MindSpeed-RL!921	3 个月前
utils	TransferQueue 兼容TensorDock 方案 Co-authored-by: zhiqiang024<zhiqiang5@h-partners.com> Co-authored-by: jianingdong<jianingstat@outlook.com> # message auto-generated for no-merge-commit merge: !902 merge tq-merge into master TransferQueue 兼容TensorDock 方案 Created-by: zhiqiang024 Commit-by: zhiqiang024;jianingdong Merged-by: ascend-robot Description: 精度验证结果（GRPO、DAPO、PPO、GRPO多模、partial rollout(同步异步)）： https://clouddocs.huawei.com/wapp/share/ac9ab6f8-3c47-4f46-b7c9-a44557fb6b29 特性： 1.支持现有transferdock的全部功能，可以通过配置的方式实现tq和td 切换，并实现对主流程的低侵入式修改。 2.新增可配置的分布式存储，使用Ray控制面+ZMQ数据面，兼容现有基于Ray的代码的同时，提供了ZMQ的快速数据传输能力。 TransferQueueClient: 线程安全的客户端，提供访问TransferQueue的全部接口。 TransferQueueManger: 提供控制面的分片、索引分配、状态位维护等功能，全局唯一，支持并发访问。 TransferQueueShard：数据存储后端，与Client通过ZMQ直连传输数据；统一存取接口，后续可提供外接Redis等存储后端的功能。 3.新增多topic数据存储，单一TransferQueue实例支持多套 schema 并存，服务多模态、动态采样、agent等场景。 transferdock、sampling_transferdock、mm_transferdock可以被一个TransferQueue替换。基于topic的存储，支持多agent训练场景下的经验数据隔离 4.完成为partial rollout提供的not-exactly-once消费、data staleness（age）的通用支持。 5.支持init时配置所有RL阶段所需的column，TQ自动通过RL阶段名来取出对应column。 Contributor: 韦亦迅、党政、植强、董佳宁 See merge request: Ascend/MindSpeed-RL!902	5 个月前
workers	Refactor comments for folder models Co-authored-by: psyloy<zhangjianxiang3@huawei.com> # message auto-generated for no-merge-commit merge: !921 merge master into master Refactor comments for folder models Created-by: McFlurry100 Commit-by: psyloy Merged-by: ascend-robot Description: 添加模块化注释，未修改非注释代码，修改前后功能等效 See merge request: Ascend/MindSpeed-RL!921	3 个月前
__init__.py	【FEAT】MindSpeed-RL 增加 swanlab logger Co-authored-by: wangqilin-0603<wangqilin16@huawei.com> # message auto-generated for no-merge-commit merge: !821 merge master into master 【FEAT】MindSpeed-RL 增加 swanlab logger Created-by: wangqilin-0603 Commit-by: wangqilin-0603 Merged-by: ascend-robot Description: SWANLAB 是机器学习领域广泛使用的实验跟踪与可视化工具。功能全面，可以展示丰富的数据类型，包括训练曲线，图片，视频，表格，html，matplotlib图像等，SwanLab对比WandB部署更容易，更安全，更易用，并且可以离线。 See merge request: Ascend/MindSpeed-RL!821	6 个月前