Skip to content

Pull requests: InternLM/xtuner

New pull request New
114 Open 1,268 Closed
114 Open 1,268 Closed
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
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

[Fix] Restore vLLM rollout topology
#2017 opened Aug 13, 2026 by YanhuiDua Collaborator Loading…
Fix/h200 cu130 heyifan
#2014 opened Aug 12, 2026 by YifanHe-ailab Collaborator Loading…
【CI】add cases about ep_size
#2011 opened Aug 12, 2026 by kkscilife Collaborator Loading…
[Fix] Align Qwen3.5 tests with Transformers 5.14.1
#2009 opened Aug 10, 2026 by jayhenry Collaborator Loading…
[Feature] Add MoE expert tensor parallelism
#2007 opened Aug 10, 2026 by jayhenry Collaborator Loading…
2
[Fix] Isolate rollout pause control requests
#2004 opened Aug 10, 2026 by 123penny123 Loading…
[Feat] Add off-policy masking for partial rollouts
#2003 opened Aug 10, 2026 by YanhuiDua Collaborator Loading…
46
docs: fix example config paths in English installation guide
#1994 opened Jul 31, 2026 by latent-9 Loading…
[Feature] Add Checkpoint Engine as a transport between train and rollout
#1993 opened Jul 31, 2026 by PengchengShi00 Collaborator Loading…
35
Train all Qwen3.5 assistant turns
#1992 opened Jul 31, 2026 by albertimff Loading…
[Perf] Add CuTe DSL DSA indexer
#1991 opened Jul 31, 2026 by jayhenry Collaborator Loading…
[Feature] Keep DSA top-k out of checkpoint replay
#1989 opened Jul 30, 2026 by jayhenry Collaborator • 5/5 Loading…
1
[Refactor] Pass GLM DSA top-k IDs explicitly
#1988 opened Jul 29, 2026 by jayhenry Collaborator • 4/5 Loading…
1
Prepare torch 2.12.1 / CUDA 13.2 / triton 3.7.1 / transformers 5.14.1 SFT compatibility
#1983 opened Jul 28, 2026 by ShilohYu Collaborator Loading…
1
[Feature] Add recompute_cfg and per-model recompute declarations (3/3)
#1981 opened Jul 28, 2026 by HAOCHENYE Collaborator • 3/5 Loading…
8
[Feature] Add the selective checkpointing engine (2/3)
#1980 opened Jul 28, 2026 by HAOCHENYE Collaborator • 2/5 Loading…
7
[Refactor] Switch gradient checkpointing to the non-reentrant implementation (1/3)
#1979 opened Jul 28, 2026 by HAOCHENYE Collaborator • 1/5 Loading…
17
[Refactor] Pass GLM DSA top-k IDs explicitly
#1978 opened Jul 28, 2026 by jayhenry Collaborator Loading…
[Feat] Add Sampled-token OPD Support to RL Training
#1977 opened Jul 27, 2026 by YanhuiDua Collaborator Loading…
[FSDP][Loss] Reduce-sum gradient reduction with per-batch aux-loss hub
#1968 opened Jul 20, 2026 by HAOCHENYE Collaborator • Draft
11
[Feat] Support immediate rollout recovery by export hf asynchronously
#1966 opened Jul 17, 2026 by YanhuiDua Collaborator Loading…
9
refine session server
#1953 opened Jul 10, 2026 by YanhuiDua Collaborator Loading…
[Fix] Improve MTP compatibility and mitigate MoE load imbalance by skipping pad tokens
#1952 opened Jul 10, 2026 by tina-wen Contributor Loading…
[Perf] Reduce communication overhead, improve NPU compatibility, and add CPU core binding
#1951 opened Jul 10, 2026 by tina-wen Contributor Loading…
[feature] Support SGLang disk weight recovery
#1950 opened Jul 9, 2026 by PengchengShi00 Collaborator Loading…
2
ProTip! Add no:assignee to see everything that’s not assigned.