[Bugfix][XPU] store the pointer raw bit pattern instead of its numeric value (#54514)

Signed-off-by: Lai, Yejing <yejing.lai@intel.com>
Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>
This commit is contained in:
Yejing Lai
2026-09-24 02:05:26 -07:00
committed by GitHub
co-authored by Kunshang Ji
parent c0c35ddd8e
commit 14f98cbe5d
@@ -37,8 +37,8 @@ class PromptEmbedsState:
# Indexed by req_state index. Stale entries after removal are
# harmless: add_request rewrites all fields for every index it claims.
self.embeds_ptrs = UvaBackedTensor(max_num_reqs, dtype=torch.int64)
self.mask_ptrs = UvaBackedTensor(max_num_reqs, dtype=torch.int64)
self.embeds_ptrs = UvaBackedTensor(max_num_reqs, dtype=torch.uint64)
self.mask_ptrs = UvaBackedTensor(max_num_reqs, dtype=torch.uint64)
self.embeds_lens = UvaBackedTensor(max_num_reqs, dtype=torch.int32)
def add_request(self, req_index: int, new_req_data: NewRequestData) -> None:
@@ -100,8 +100,8 @@ class PromptEmbedsState:
def _apply_prompt_embeds_kernel(
inputs_embeds_ptr,
inputs_embeds_stride,
embeds_ptrs_ptr, # int64 [max_num_reqs], device pointers (0-len = unused)
mask_ptrs_ptr, # int64 [max_num_reqs], 0 = no is-token-ids mask
embeds_ptrs_ptr, # uint64 [max_num_reqs], device pointers (0-len = unused)
mask_ptrs_ptr, # uint64 [max_num_reqs], 0 = no is-token-ids mask
embeds_lens_ptr, # int32 [max_num_reqs]
idx_mapping_ptr,
query_start_loc_ptr,