[CUDA] Fuse MoE router bias into MatMulNBits GEMV #29170
[CUDA] Fuse GPT-OSS router bias into MatMulNBits GEMV
2a2ce830
tianleiwu
force pushed
from
651acb5a
to
2a2ce830
21 days ago
Merge remote-tracking branch 'origin/main' into tlwu/20260619/qmoe_ro…
e3f75417
Require static M==1 for CUDA router bias fusion
13e6ffb4
Fix -Wshorten-64-to-32 error in MatMulNBitsBiasFusion test
f69cb5fd
Address review: qualify std::optional and add explicit int casts in t…
5acf8a99
add fallback
b0a4d86e
tianleiwu
marked this pull request as ready for review 20 days ago
extend to qwen and gemma
25789d98
add doc
4f64dc59
tianleiwu
changed the title [CUDA] Fuse GPT-OSS router bias into MatMulNBits GEMV [CUDA] Fuse MoE router bias into MatMulNBits GEMV 20 days ago
Address review: restrict router GEMV to GPT-OSS, cap bias-add grid, r…
fbff53a3
Merge remote-tracking branch 'origin/main' into tlwu/20260619/qmoe_ro…
5926f730
docs: fix relative source links in matmul_nbits.md
4d2bc950
docs+test: align router GEMV scope with code; cover block_size 64
9b23007d
kunal-vaishnavi
deleted the tlwu/20260619/qmoe_router_bias_gemv branch 16 days ago
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub