-
Notifications
You must be signed in to change notification settings - Fork 35
Pull requests: RL-Align/RL-Kernel
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[FEAT][kernels] route LM-head vocab projection through deterministic GEMM
#194
opened Jun 26, 2026 by
inaniloquentee
Collaborator
Loading…
feat: add embedding invariance and LM-head linear_logp routing
#190
opened Jun 25, 2026 by
inaniloquentee
Collaborator
Loading…
[FEAT][kernels] Add tensor-parallel linear_logp path
#189
opened Jun 24, 2026 by
inaniloquentee
Collaborator
Loading…
:feat(ws1): add NativeAttentionOp pure-PyTorch standard-softmax reference
#188
opened Jun 24, 2026 by
maxiaosong1124
Collaborator
Loading…
feat(kernels): add fused masking + variable-length pack-and-pad op
needs-gpu-ci
#182
opened Jun 23, 2026 by
Chen-BUPT
Loading…
[WS2][distributed] Add deterministic all-reduce contract and smoke tests
#181
opened Jun 23, 2026 by
CyberSecurityErial
Loading…
[WS1][kernels] Batch-invariant deterministic GEMM (fwd + bwd)
needs-gpu-ci
#180
opened Jun 22, 2026 by
Flink-ddd
Collaborator
Loading…
feat(ws1):Add LogProb reference operator interface(reuse)
needs-gpu-ci
#179
opened Jun 22, 2026 by
a-kaa
Collaborator
Loading…
feat(ws1): NativeLMHeadOp pure-PyTorch ground-truth reference + numerical contract tests
needs-gpu-ci
#170
opened Jun 22, 2026 by
maxiaosong1124
Collaborator
Loading…
feat(ws1): NativeEmbeddingOp pure-PyTorch ground-truth reference + numerical contract tests
needs-gpu-ci
#169
opened Jun 22, 2026 by
maxiaosong1124
Collaborator
Loading…
6 tasks done
feat(ws1): Add PyTorch matmul reference operator
needs-gpu-ci
#168
opened Jun 21, 2026 by
a-kaa
Collaborator
Loading…
feat(ws1): Add PyTorch RoPE reference operator
needs-gpu-ci
#167
opened Jun 21, 2026 by
a-kaa
Collaborator
Loading…
feat(ws1): NativeSiLUOp + NativeSwiGLUOp pure-PyTorch ground-truth references + numerical contract tests
needs-gpu-ci
#166
opened Jun 21, 2026 by
maxiaosong1124
Collaborator
Loading…
6 tasks done
feat(ws1): NativeRMSNormOp pure-PyTorch ground-truth reference + numerical contract tests
needs-gpu-ci
platform: cuda
Specific optimizations or bugs in NVIDIA graphics cards (such as FlashInfer, TMA optimizations)
priority: high
Severe congestion issues require the highest priority for resolution.
sprint-0615
#160
opened Jun 20, 2026 by
maxiaosong1124
Collaborator
Loading…
docs: define vime integration design
#126
opened Jun 16, 2026 by
inaniloquentee
Collaborator
Loading…
docs: map vime architecture hook points
#125
opened Jun 16, 2026 by
inaniloquentee
Collaborator
Loading…
feat: add logprob cross-engine benchmark
#107
opened Jun 13, 2026 by
inaniloquentee
Collaborator
Loading…
feat(testing): add TP-invariant reduction references
#103
opened Jun 12, 2026 by
inaniloquentee
Collaborator
Loading…
Add batch-invariant deterministic CUDA logp
needs-gpu-ci
#98
opened Jun 11, 2026 by
inaniloquentee
Collaborator
Loading…
ProTip!
Add no:assignee to see everything that’s not assigned.