Support speculative decoding on CPU (#27862)

Co-authored-by: Valentine233 <xuan.liao@intel.com>
2026-07-09 10:27:09 +08:00
parent 177c048c68
commit 3b43df5b6d
36 changed files with 3499 additions and 138 deletions
@@ -305,3 +305,17 @@ def transfer_kv_all_layer_mla_lf_pf(
        block_quota,
        num_warps_per_block,
    )
+
+
+def copy_all_layer_kv_cache_cpu(
+    data_ptrs: torch.Tensor,
+    strides: torch.Tensor,
+    tgt_loc: torch.Tensor,
+    src_loc: torch.Tensor,
+):
+    torch.ops.sgl_kernel.copy_all_layer_kv_cache_cpu(
+        data_ptrs,
+        strides,
+        tgt_loc,
+        src_loc,
+    )