alibaba / alibaba/ROLL

🚀 [2026/06/18] Recent Updates Summary for ROLL Project

Open
#464 1 comment 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
3.4k
Forks
312
Avg merge
1h 2m
Merged PRs (30d)
2

Description

大家好!感谢大家对ROLL的关注。ROLL发布了v0.3.0版本,新增Video RLVR、AgentRunner 2.0、MTP训练、Router Replay、Multi-Teacher OPD等重要特性;新增OpenTelemetry可观测性支持;强化mcore_adapter能力;扩展NPU/AMD硬件适配。以下是近期更新的一些梳理,我们将持续对ROLL进行迭代更新,欢迎加入ROLL的社区。

## 🚀 亮点

- 新增 Video/Audio RLVR 训练支持(Video-R1 reward)
- 新增 AgentRunner 2.0 抽象,解耦Agent交互逻辑,支持更灵活的多轮Agent场景
- 新增 RemoteBatch 惰性数据传输机制,优化大规模 image/video/long_context logits 跨Worker传输
- 新增 MoE Router Replay (R3),MoE模型训练时复用rollout阶段的路由决策
- 支持 Qwen3.5/3.6 MTP (Multi-Token Prediction) SFT/RL 训练
- 新增 OpenTelemetry 分布式追踪,提供端到端可观测性

## 🚀 主要新特性

### Pipeline

- 新增 [Video](examples/qwen3-vl-30BA3B-video-rlvr/video_r1.yaml)/[Audio](examples/qwen3-omni/audio_test_80G.yaml) RLVR Pipeline,支持视频、音频理解场景的强化学习训练(Qwen3Omni系列模型)
- 新增 Multi-Teacher On-Policy Distillation 支持,[文档](docs_roll/i18n/zh-Hans/docusaurus-plugin-content-docs/current/User Guides/Pipeline/on_policy_distill_pipeline_start.md)
- 新增 LLM-as-Judge Server 模式,支持独立部署 judge 服务,[示例配置](examples/qwen2.5-7B-rlvr_megatron/rlvr_config_8gpus_llm_as_judge_server.yaml)

### Agent Native 2.0

- 新增 AgentRunner 抽象,解耦"Agent如何与环境交互"与"训练样本构造",[设计文档](docs_roll/i18n/zh-Hans/docusaurus-plugin-content-docs/current/User Guides/Agentic/agent_runner.md)
- 新增 ProxyEnvManager / MessageTracker,支持更复杂的Agent交互模式,[设计文档](docs_roll/i18n/zh-Hans/docusaurus-plugin-content-docs/current/User Guides/Agentic/prefix_aggregation.md)
- 新增 Atropos 环境集成,[示例配置](examples/agentic_demo/atropos_gsm8k_grpo_qwen25_0.5b.yaml)
- 新增 OpenReward 环境集成,[示例配置](examples/agentic_demo/openreward_endless_terminals_reinforce_qwen35_2b.yaml)

### mcore_adapter

- 新增 MTP (Multi-Token Prediction) 训练支持(standalone/joint两种模式),[使用文档](docs_roll/i18n/zh-Hans/docusaurus-plugin-content-docs/current/User Guides/Advanced Features/mtp_training.md)
- 新增 Router Replay (R3),MoE训练时复用rollout路由决策,减少重计算开销,[使用文档](docs_roll/docs/User Guides/Advanced Features/router_replay.md)
- 新增 Fused Entropy CE kernel,TP=1场景下加速交叉熵计算
- 新增 PP Stage Compile Warmup,Pipeline并行编译预热
- Qwen3.5/3.6系列 VLM sequence packing 优化

### RemoteBatch 传输优化

- 新增惰性数据传输后端,支持 image/video/long_context 场景下大规模数据高效传输,[使用文档](docs_roll/i18n/zh-Hans/docusaurus-plugin-content-docs/current/User Guides/Advanced Features/remote_batch_transfer.md)
- 基于[TransferQueue](https://github.com/Ascend/TransferQueue)优化 Ray Worker 间存储管理

### Observability

- 新增 OpenTelemetry 集成,支持分布式追踪,[使用文档](docs_roll/i18n/zh-Hans/docusaurus-plugin-content-docs/current/User Guides/Advanced Features/opentelemetry_tracing.md)
- 新增 OTEL Receiver,pipeline各阶段端到端tracing

### FSDP2

- 新增 Qwen3 MoE patch,支持 MoE 模型 FSDP2 训练
- LoRA模型支持优化
- FSDP2 / EP并行支持

### Docker

- 新增 NPU A2/A3 Docker 镜像
- 新增 AMD torch2.8.0/torch2.10 Docker 镜像

### Hardware

- NPU:新增 A2/A3 适配,修复 FSDP2 相关问题,新增 Ascend 全流程文档
- AMD:新增 torch2.10 支持,ROCm参数同步优化

### Models

- 支持 [Qwen3.5 Dense (27B)](examples/qwen3.5-27B-rlvr_megatron/rlvr_megatron_80GB.yaml) / [MoE (35B-A3](examples/qwen3.5-35BA3-rlvr_megatron/rlvr_megatron_80GB.yaml), [122B-A10](examples/qwen3.5-122A10-rlvr_megatron/rlvr_megatron_80GB.yaml), [397A-17](examples/qwen3.5-397A17-rlvr_megatron/rlvr_megatron_80GB.yaml)) 系列模型
- 提供 Megatron + FSDP2 多种规格配置示例

### Performance 优化

- do_checkpoint pin_memory 优化
- GC 优化
- low-memory checkpoint convert

### Bug Fix

- fix sglang & vllm 偶现 port conflict
- fix reward worker metrics 透出
- fix vllm GDN attention mixed decode/spec-decode crash(vllm < 0.17.2)

## Deprecated

- DeepSpeed Strategy(third_party代码已移除)
- Wan RewardRL(生成模型的RL训练重构中)

## TODOs

- Multi Agent 支持
- Full vocab version Multi-Teacher OPD

Contributor guide

No contributing guide indexed for this repository

Research direction

This issue is a release and update summary covering multiple features, documentation pages, examples, and bug fixes, but it does not identify a requested change. Review the linked documentation and example paths if a specific follow-up is assigned; the issue itself provides no clear completion criterion or test to run.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
documentation
Issue type
Documentation
Difficulty
1/5
Estimated time
Under an hour
Activity status
Quiet
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.