🚀 [2026/02/04] Recent Updates Summary for ROLL Project

[PanAndy]

大家好！感谢大家对ROLL的关注。
ROLL近期更新了大量新功能，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:

• 新增模型支持：Qwen3-VL、Qwen3-MoE-VL、Qwen3-Omni、GLM-4.7
• agentic 训练与 Rollout GPU部分重叠，训练空闲GPU切换为Rollout
• DynamicSamplingScheduler协程化重构
• 新增: FSDP2 Strategy
• 训练支持 Sequence packing 和 Dynamic batching

🚀主要新特性：

• Rollout
  • DynamicSamplingScheduler协程化重构
  • 自定义rollout pre/post process， 支持动态samping param、多阶段生成、ThinkingBudget控制
  • Sglang： Strategy重构，支持server模式，onload/offload native化，inflight FP8 quant rollout，跨机多节点部署
  • vLLM：DP/EP 支持， 支持vllm==0.12.0
  • 提供AgentNative Rollout范式，AgentNativeStepEnvManager + SokobanNativeEnv，完全由env进行上下文管理
  • Async Rollout Hang Detect：增加异步 Rollout 卡死检测，快速定位问题env
  • 支持rollout dump & mock，提高forward/train阶段精度对齐效率
  • agentic pipeline支持 train-val/rollout overlap
• Training
  • FSDP2
  • Megatron support LoRA, LoRA RL blogs:https://macaron.im/mindlab/research/building-trillion-parameter-reasoning-rl-with-10-gpus
  • megatron训练时在线保存hf格式的模型参数
  • support FP8 training for Megatron Strategy
  • Sequence packing，微调loss_func接口定义
  • Dynamic batching
  • Add DeepSpeed SFT support
• Model Update实现优化：消除机间冗余、权重转换和nccl broadcast overlap、优化host to device、多pp串行同步调整为lock模式同时同步
• Asynchronous Feature
  • 训练与 Rollout GPU部分重叠，训练空闲GPU切换为Rollout，report: https://arxiv.org/abs/2512.24873
  • agentic off policy loss 与 IS 修正
• Pipeline recipe
  • VLM image tool use： DeepEyes，工具调用与reward计算overlap
• Models：新增模型支持 Qwen3-VL、Qwen3-MoE-VL、Qwen3-Omni-Thinker、GLM-4.7

[PanAndy]

Hello everyone! Thanks for your interest in ROLL.
ROLL has recently introduced a large number of new features. Below is a summary of recent updates. We will continue to iterate on ROLL, and you are very welcome to join the ROLL community.

🚀 Highlights:

• New model support: Qwen3-VL, Qwen3-MoE-VL, Qwen3-Omni, GLM-4.7
• Agentic training and Rollout GPU partial overlap, switching idle training GPUs to Rollout
• DynamicSamplingScheduler coroutine refactoring
• New: FSDP2 Strategy
• Training supports Sequence packing and Dynamic batching

🚀 Major New Features:

• Rollout
  • DynamicSamplingScheduler coroutine refactoring
  • Custom rollout pre/post process, supporting dynamic sampling params, multi-stage generation, ThinkingBudget control
  • Sglang: Strategy refactoring, supporting server mode, native onload/offload, inflight FP8 quant rollout, cross-machine multi-node deployment
  • vLLM: DP/EP support, supports vllm==0.12.0
  • Provides AgentNative Rollout paradigm, AgentNativeStepEnvManager + SokobanNativeEnv, fully managed context by env
  • Async Rollout Hang Detect: Added asynchronous Rollout hang detection to quickly locate problematic envs
  • Supports rollout dump & mock, improving forward/train phase precision alignment efficiency
  • Agentic pipeline supports train-val/rollout overlap
• Training
  • FSDP2
  • Megatron support LoRA, LoRA RL blogs: https://macaron.im/mindlab/research/building-trillion-parameter-reasoning-rl-with-10-gpus
  • Save model parameters in HF format online during Megatron training
  • Support FP8 training for Megatron Strategy
  • Sequence packing, fine-tuned loss_func interface definition
  • Dynamic batching
  • Add DeepSpeed SFT support
• Model Update implementation optimization: Eliminate inter-machine redundancy, weight conversion and nccl broadcast overlap, optimize host to device, adjust multiple pp serial synchronization to lock mode for simultaneous synchronization
• Asynchronous Feature
  • Training and Rollout GPU partial overlap, switching idle training GPUs to Rollout, report: https://arxiv.org/abs/2512.24873
  • Agentic off policy loss with IS correction
• Pipeline recipe
  • VLM image tool use: DeepEyes, tool invocation and reward calculation overlap
• Models: New model support for Qwen3-VL, Qwen3-MoE-VL, Qwen3-Omni-Thinker, GLM-4.7