-
Notifications
You must be signed in to change notification settings - Fork 602
Pull requests: linkedin/Liger-Kernel
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
fix(monkey_patch): patch Q/K norms on existing Qwen models
#1470
opened Sep 16, 2026 by
luca-888
Contributor
Loading…
1 of 3 tasks
fix(cutedsl): fuse Gemma RMSNorm weight offset in FP32
#1462
opened Sep 11, 2026 by
luca-888
Contributor
Loading…
1 of 3 tasks
fix(grpo): respect independent token and mask strides
#1461
opened Sep 11, 2026 by
LOGO127
Loading…
3 tasks
fix(mlp): preserve packed activations across backward replay
#1460
opened Sep 11, 2026 by
arde171
Collaborator
Loading…
3 tasks
fix(cutile): respect native shape and backward contracts
#1459
opened Sep 11, 2026 by
arde171
Collaborator
Loading…
3 tasks
test(cutedsl): gap-op dispatcher tests + fused_add_rms_norm 3-D backward + shape fallbacks
#1457
opened Sep 10, 2026 by
arde171
Collaborator
Loading…
feat(cutile): register six backends and harden CE/SwiGLU contracts
#1456
opened Sep 10, 2026 by
arde171
Collaborator
Loading…
3 tasks
feat(cutile): chunk FLCE with Triton sizing and cuBLAS GEMMs
#1455
opened Sep 10, 2026 by
vaibhavjindal
Collaborator
•
Draft
2 of 3 tasks
feat(ops): support partial RoPE in Triton kernel
#1451
opened Sep 9, 2026 by
piyushumate
Contributor
Loading…
5 tasks done
feat(cutedsl): add CuTe-DSL Newton-Schulz operator and LigerMuon optimizer
#1449
opened Sep 9, 2026 by
piyushumate
Contributor
•
Draft
2 tasks done
[WIP] Add Qwen4Exp text model support
#1447
opened Sep 9, 2026 by
alifurkanstahl
•
Draft
3 of 7 tasks
fix(llama4): honour skip_logits in lce_forward
#1445
opened Sep 7, 2026 by
yupengtang
Contributor
Loading…
fix(monkey_patch): let the norm flags cover the whole vision tower
#1443
opened Sep 7, 2026 by
yupengtang
Contributor
Loading…
perf(norm): skip dW/dB calculation when parameters are frozen (#1067)
#1440
opened Sep 4, 2026 by
piyushumate
Contributor
Loading…
feat(flce): expose direct chunk-size control via LIGER_FLCE_CHUNK_SIZE
#1439
opened Sep 4, 2026 by
jin-five
Loading…
3 tasks done
fix: request float32 logits from the fused JSD projection GEMM
#1433
opened Sep 1, 2026 by
KyleMylonakisProtopia
Contributor
Loading…
perf(flce): write the weight-gradient product directly into grad_weight
#1429
opened Sep 1, 2026 by
ananthv26
Loading…
2 of 3 tasks
fix(peft): patch the LayerNorm copies, not the ModulesToSaveWrapper itself
#1425
opened Aug 30, 2026 by
yurekami
Loading…
[Feature] Support apply_liger_kernel_to_deepseek_v3
#1424
opened Aug 30, 2026 by
Yulong-Cauli
Contributor
Loading…
3 tasks done
docs: fix input/target shapes in DPO and ORPO forward docstrings
#1406
opened Aug 25, 2026 by
Aryan95614
Loading…
1 of 3 tasks
Load unsupported model types instead of raising KeyError in AutoLigerKernelForCausalLM
#1404
opened Aug 23, 2026 by
nileshpatil6
Loading…
Previous Next
ProTip!
Exclude everything labeled
bug with -label:bug.