Skip to content

Pull requests: togethercomputer/xorl

Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort

Pull requests list

Add the exact GLM-5.2 training contract
#35 opened Aug 3, 2026 by kiddyboots216 Contributor Draft
Add server execution and orchestration
#33 opened Jul 31, 2026 by kiddyboots216 Contributor 9/11 Loading…
Harden server transport and weight synchronization
#32 opened Jul 31, 2026 by kiddyboots216 Contributor 8/11 Loading…
Add RL objectives and distillation primitives
#31 opened Jul 31, 2026 by kiddyboots216 Contributor 7/11 Loading…
Add FP8, NVFP4, and QARL training and export
#30 opened Jul 31, 2026 by kiddyboots216 Contributor 6/11 Loading…
Add DeepSeek-V4 and GLM-5 support
#29 opened Jul 31, 2026 by kiddyboots216 Contributor 5/11 Loading…
Add common MoE, MiniMax-M3, and Nemotron-H support
#28 opened Jul 31, 2026 by kiddyboots216 Contributor 4/11 Loading…
Extend distributed training primitives
#27 opened Jul 31, 2026 by kiddyboots216 Contributor 3/11 Loading…
Add linear-attention and SSM kernels
#26 opened Jul 31, 2026 by kiddyboots216 Contributor 2/11 Loading…
Establish the CUDA and Quack runtime
#25 opened Jul 31, 2026 by kiddyboots216 Contributor 1/11 Loading…
Add the portable XoRL training simulator
#24 opened Jul 30, 2026 by kiddyboots216 Contributor Loading…
Add architecture-specific MoE and zero-K3 training support
#23 opened Jul 30, 2026 by kiddyboots216 Contributor 11/11 Loading…
Add the consolidated train-serve numerical contract
#22 opened Jul 30, 2026 by kiddyboots216 Contributor 10/11 Loading…
ProTip! no:milestone will show everything without a milestone.