Early exit for MoE LoRA kernels (#27131)

Signed-off-by: gnovack <gnovack@amazon.com>
Co-authored-by: Jee Jee Li <pandaleefree@gmail.com>
This commit is contained in:
gnovack
2025-11-03 20:22:17 +08:00
committed by GitHub
co-authored by Jee Jee Li
parent 40b69e33e7
commit 294c805f1d
11 changed files with 123 additions and 34 deletions
+3 -1
View File
@@ -44,7 +44,9 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, m) {
" int max_num_m_blocks, "
" Tensor !sorted_token_ids,"
" Tensor !experts_ids,"
" Tensor !num_tokens_post_pad) -> () ");
" Tensor !num_tokens_post_pad,"
" Tensor !adapter_enabled,"
" Tensor !lora_ids) -> () ");
m.impl("moe_lora_align_block_size", torch::kCUDA, &moe_lora_align_block_size);
#ifndef USE_ROCM