Skip to content

Pull requests: linkedin/Liger-Kernel

New pull request New
66 Open 886 Closed
66 Open 886 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

Add LigerAttnRes nn.Module wrapper for the AttnRes op
#1293 opened Jul 7, 2026 by abtonmoy Loading…
2 of 3 tasks
perf: remove device to host syncs in cross entropy forward and backward
#1292 opened Jul 7, 2026 by kashif Contributor Loading…
1
[CuteDSL] FLCE: 2-SM online-softmax fast path (Blackwell), ~5x over Triton
#1291 opened Jul 7, 2026 by justinhh4 Contributor Loading…
[CuteDSL] Add benchmarking, CE memory optimizations, and match Triton num_warps
#1290 opened Jul 6, 2026 by justinhh4 Contributor Loading…
Fix ZeroDivisionError in fused linear cross entropy when weight vocab dim is 0
#1289 opened Jul 4, 2026 by nileshpatil6 Loading…
1
Add Native Sparse Attention (NSA) Triton kernels and module
#1288 opened Jul 4, 2026 by abtonmoy Loading…
2 of 3 tasks
1
1
Add fused QK-Norm + RoPE kernel for Qwen3
#1286 opened Jul 2, 2026 by zhangxin81 Loading…
1 of 3 tasks
Add cuTile RMSNorm and shared DTensor handling updates
#1285 opened Jul 2, 2026 by Charlesrizzly • Draft
3 tasks
Add THD support for RoPE
#1284 opened Jul 2, 2026 by niyunsheng Contributor Loading…
3 tasks done
fix: align _input_chunk dtype in FLCE grad_weight addmm under AMP
#1283 opened Jul 2, 2026 by justinhh4 Contributor Loading…
2 of 3 tasks
Add cutedsl implementation of Liger Rope Kernel
#1282 opened Jul 1, 2026 by Celaena24 Contributor Loading…
3 tasks done
15
Implement RMSNORM in cuTile and cuTeDSL
#1280 opened Jul 1, 2026 by Charlesrizzly Loading…
3 tasks
3
Add CuTe DSL implementation of the swiglu liger kernel
#1277 opened Jun 29, 2026 by Celaena24 Contributor Loading…
3 tasks done
1
[transformers] Add tiled MLP patching via the transformers register_patch_mapping API
#1276 opened Jun 28, 2026 by kashif Contributor Loading…
1
1
perf(flce): fuse grad_weight accumulation via addmm_ (cuBLAS beta=1) + memory cleanups
#1270 opened Jun 26, 2026 by justinhh4 Contributor Loading…
1
[cuTile] Add rope/qwen2vl_mrope/kl_div/group_norm/multi_token_attention
#1269 opened Jun 26, 2026 by xjmxyt Contributor Loading…
1 of 3 tasks
1
perf(ce): single-shot dx_y gradient correction
#1268 opened Jun 25, 2026 by justinhh4 Contributor Loading…
Add DeepSeek 3.2 support
#1264 opened Jun 20, 2026 by michaelroyzen Contributor Loading…
1
feat: optimize fused linear JSD memory and runtime
#1261 opened Jun 14, 2026 by d7chong Loading…
3 tasks done
5
Cast GroupNorm fp16 backward gradient accumulators to fp16 (not bfloat16)
#1253 opened Jun 4, 2026 by lollinng Loading…
Fix softmax multi-block path for n_cols > 65536 (unreachable + non-compiling)
#1252 opened Jun 4, 2026 by lollinng Loading…
Add inplace=False to liger_cross_entropy to fix upstream gradient corruption (#272)
#1251 opened Jun 4, 2026 by lollinng Loading…
1
perf: extend RMSNorm block-row dispatch to hidden<=512, lower row threshold
#1238 opened May 25, 2026 by alifurkanstahl Loading…
2 of 3 tasks
fix LigerORPOTrainer bug in none fsdp setup
#1230 opened May 19, 2026 by Ola-Vish • Draft
1 task done
fix(glm4v): version-aware hidden_size/vocab_size for Glm4vConfig
#1214 opened May 5, 2026 by 1fanwang Member Loading…
1
ProTip! no:milestone will show everything without a milestone.