AReaL
Python MediuminclusionAI/AReaL
5,005 stars
454 forks
20 open issues
Active Apr 2026
Beginner-Friendly Issues 20
Issues tagged for new contributors
[Feature] Support multi-node inference workers in inference gateway
#1149 · Apr 8, 2026
good first issue call-for-contribution
DPO algo and WeChat QR-code support
#1137 · Apr 4, 2026
question
[BUG] FSDP compute_logp fails for Hugging Face Qwen3.5 with dict attention_mask
#1132 · Apr 2, 2026
bug
feat(engine): sparse delta compression for disk-based weight updates
#1125 · Apr 1, 2026
[Feature] `RTensor` optimization: caching and IPC backend
#1117 · Mar 31, 2026
enhancement good first issue call-for-contribution
question
[Feature] Feature Proposal: Multi-Agent Training Framework (Dr. MAS Integration)
#1114 · Mar 30, 2026
[Feature] Load dataset distributedly in the single-controller mode
#1111 · Mar 30, 2026
[Feature] Async Tensor Parallelism for Archon Engine
#1110 · Mar 30, 2026
refactor(archon): adopt Pydantic models for RTensor data proxy endpoints
#1106 · Mar 30, 2026
good first issue call-for-contribution
[Feature] Elastic weight update setup and acceleration
#1101 · Mar 27, 2026
[BUG]kl_ctl大于0时(使用reference model),训练报错
#1099 · Mar 27, 2026
bug npu
[Feature] Add Trackio as a New Backend for Experiment Visualization
#1089 · Mar 25, 2026
good first issue call-for-contribution
[Feature] Auto-scaling Inference Engines in Single-Controller Mode for AReaL
#1086 · Mar 25, 2026
[Question] The Wechat QR code is not working
#1066 · Mar 20, 2026
question stale
[Feature] Sequence/token level rejection sampling on async training
#1052 · Mar 18, 2026
enhancement good first issue call-for-contribution
[Feature] Fixed bugs in Archon LoRA Backend
#1040 · Mar 15, 2026
stale
[Feature] Support LLaVA-OneVision-1.5 RL Training
#1028 · Mar 13, 2026
good first issue call-for-contribution
[Doc]微信群二维码已过期
#1016 · Mar 11, 2026
documentation stale
[Feature] Integrating aenvironment for environment handling
#1014 · Mar 10, 2026
stale