Repository Issue Activity (beta feature)

#6078 [RFC] Alternative design for vllm-omni rollout LoRA sync: tmpfs + cudaHostRegister preserving file-loading semantics

#

#5836 [roadmap] verl 26Q2 roadmap

#

#6104 freeze vision model (megatron backend, mbridge)

#

#

#

#6100 【Ascend】已提交被覆盖了

#4815 qw3 8B 同时开启mbridge和VPP时报错IndexError: too many indices for tensor of dimension 3

#

#

#

#

#6083 Megatron-Bridge环境问题

#

#6094 Sequence Parallelism (SP) leads to tensor shape mismatch in Qwen-3.5 but works in Qwen-3

#6096 RLHFDataset does not support OpenAI-style per-sample tools / tool_choice (function calling payload)

#

#

#5998 [RFC] Skip Everything in RL workflow

#

#5889 torch.distributed.DistNetworkError: The server socket has failed to listen on any local network address. port: 18071, useIpv6: false, code: -98, name: EADDRINUSE, message: address already in use

#

#5755 [RFC] Multi-modal Generation RL 2026Q2 Roadmap