Repository navigation

All pull requests

Labels (23)Milestones (0)New pull request

Open

(146)

Closed

(1,376)

[WIP] Restore env vars after colocate/disagg weight-update tests
#2165PengchengShi00 opened Oct 13, 2026Collaborator
[Fix] Remesh FP32 ignored params onto world-wide replicas
#2162YanhuiDua opened Oct 12, 2026Collaborator3 tasks
6
[Fix] Unify Muon matrix layout for NS and lr_ratio
#2161YanhuiDua opened Oct 12, 2026Collaborator4 tasks done
8
test: add trainer-free AutoModel to existing HF/XT parity test
#2152xs1997zju opened Oct 10, 2026Collaborator
[Enhance] extend sandbox creation args
#2142braisedpork1964 opened Oct 9, 2026Collaborator
[Docs] Propose sandbox agent loop environment interfaces
#2135matrix72c opened Oct 8, 2026Contributor
fix bug for cache file
#2133YifanHe-ailab opened Sep 30, 2026Collaborator
[WIP] support PPO
#2132PengchengShi00 opened Sep 29, 2026Collaborator
[Feature] Bind NPU training ranks to NUMA-local CPU slices (XTUNER_NPU_CPU_BIND)
#2131longboat2010 opened Sep 27, 2026
[Feature] Route FSDP2 unshard to the compute stream (XTUNER_FSDP_UNSHARD_ASYNC_OP)
#2130longboat2010 opened Sep 27, 2026
[Feature] NPU liger fused-linear-cross-entropy loss (XTUNER_LOSS_MODE=liger)
#2129longboat2010 opened Sep 27, 2026
[Feature] Fused SwiGLU Triton kernels for MLP activation (XTUNER_NPU_FUSED_SWIGLU)
#2128longboat2010 opened Sep 26, 2026
fix(datasets): build the DP all_reduce tensor on the mesh device instead of assuming CUDA
#2127li-lizhe opened Sep 25, 2026
1
[Feature] Keep Muon and AdamW optimizer state on host
#2125jayhenry opened Sep 24, 2026Collaborator
4
[Fix] Restore Muon optimizer states on DCP resume
#2124jayhenry opened Sep 24, 2026Collaborator
3
[Fix] Use at least num_sms DeepEP QPs so internode EP16 passes the IBGDA check
#2123jayhenry opened Sep 24, 2026Collaborator
[ ci ] add interns2-preivew cases on npu
#2121kkscilife opened Sep 24, 2026Collaborator
[Feature] NPU AscendC op stack for GatedDeltaNet (causal conv1d / chunk delta rule / fused gated norm)
#2120longboat2010 opened Sep 24, 2026
[Feature] NPU fused zero-centered RMSNorm (XTUNER_NPU_FUSED_ZC_NORM)
#2119longboat2010 opened Sep 24, 2026
[ ci ] add qwen3.5 cases on npu
#2117kkscilife opened Sep 24, 2026Collaborator
[Feature] Add offline token and packing statistics for text datasets
#2115bingshao333 opened Sep 24, 2026
4
[Fix]: stabilize Qwen3.5 Dense gradients across sequence parallel sizes
#2113JT-Ushio opened Sep 24, 2026Contributor
6
[Feature] Add GLM-5.3-Flash F6 core: text model + MTP + compose model
#2111jayhenry opened Sep 23, 2026Collaborator
[Feature] Add GLM-5.3-Flash F2: vision tower + projector (eager)
#2110jayhenry opened Sep 23, 2026Collaborator
1
[Feature] Add GLM-5.3-Flash F1: VL data preprocessing pipeline
#2109jayhenry opened Sep 23, 2026Collaborator
3

Pull requests