Merge pull request #1051 from InfiniTensor/issue/1050

issue/1050 - fix paged caching and paged prefill on metax

Merge pull request #1051 from InfiniTensor/issue/1050
issue/1050 - fix paged caching and paged prefill on metax
8d99a8f5 · pengcheng888 · GitHub · 3e5cad10 · 9a0f2505 · 8d99a8f5
Unverified Commit 8d99a8f5 authored Mar 05, 2026 by pengcheng888 Committed by GitHub Mar 05, 2026
3 changed files
--- a/src/infiniop/ops/paged_attention_prefill/metax/paged_attention_prefill_metax.maca
+++ b/src/infiniop/ops/paged_attention_prefill/metax/paged_attention_prefill_metax.maca
--- a/src/infiniop/ops/paged_attention_prefill/nvidia/paged_attention_prefill_nvidia.cu
+++ b/src/infiniop/ops/paged_attention_prefill/nvidia/paged_attention_prefill_nvidia.cu
--- a/src/infiniop/ops/paged_caching/metax/paged_caching_metax.maca
+++ b/src/infiniop/ops/paged_caching/metax/paged_caching_metax.maca
@@ -12,7 +12,7 @@ INFINIOP_METAX_KERNEL pagedCaching(
    const ptrdiff_t k_src_stride, const ptrdiff_t v_src_stride,
    const ptrdiff_t k_cache_block_stride, const ptrdiff_t v_cache_block_stride,
    const ptrdiff_t k_cache_head_stride, const ptrdiff_t v_cache_head_stride,
-    const ptrdiff_t k_cache_slot_stride, const ptrdiff_t v_cache_slot_strid) {
+    const ptrdiff_t k_cache_slot_stride, const ptrdiff_t v_cache_slot_stride) {
    op::paged_caching::cuda::pagedCachingKernel<Tdata, NUM_THREADS>(
        k_cache, v_cache, k, v, slot_mapping, head_size,
        block_size, k_src_stride, v_src_stride,