[Refactor] Remove dead minimax allreduce rms kernel (#46842)

Signed-off-by: yewentao256 <zhyanwentao@126.com>
This commit is contained in:
Wentao Ye
2026-06-30 08:29:21 -07:00
committed by GitHub
parent d8f483dc30
commit 9e84ec8648
4 changed files with 0 additions and 52 deletions
-5
View File
@@ -449,10 +449,6 @@ STABLE_TORCH_LIBRARY_FRAGMENT(_C, ops) {
"int cache_block_size) -> ()");
#ifndef USE_ROCM
ops.def(
"minimax_allreduce_rms("
"Tensor input, Tensor norm_weight, Tensor workspace, "
"int rank, int nranks, float eps) -> Tensor");
ops.def(
"minimax_allreduce_rms_qk("
"Tensor qkv, Tensor norm_weight_q, Tensor norm_weight_k, "
@@ -705,7 +701,6 @@ STABLE_TORCH_LIBRARY_IMPL(_C, CUDA, ops) {
"fused_deepseek_v4_qnorm_rope_kv_rope_full_cache_fp8_insert",
TORCH_BOX(&fused_deepseek_v4_qnorm_rope_kv_rope_full_cache_fp8_insert));
#ifndef USE_ROCM
ops.impl("minimax_allreduce_rms", TORCH_BOX(&minimax_allreduce_rms));
ops.impl("minimax_allreduce_rms_qk", TORCH_BOX(&minimax_allreduce_rms_qk));
#endif
ops.impl("fused_minimax_m3_qknorm_rope_kv_insert",