apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-23 08:21:53 +08:00
Code Issues Actions 19 Packages Projects Releases Wiki Activity

[XPU] modify speculate_verify (#5522)

Browse Source
This commit is contained in:
RuohengMa
2025-12-23 14:50:30 +08:00
committed by GitHub
parent 945a1bc4e2
commit 2c3c983b96
7 changed files with 278 additions and 354 deletions
Download Patch File Download Diff File Expand all files Collapse all files
custom_ops/xpu_ops/src/plugin/include/xpu/plugin.h
+4 -1
View File
@@ -384,6 +384,7 @@ DLL_EXPORT int mtp_free_and_dispatch_block(Context* ctx,
template <bool ENABLE_TOPP, bool USE_TOPK>
DLL_EXPORT int speculate_verify(Context* ctx,
const int64_t* sampled_token_ids,
int64_t* accept_tokens,
int* accept_num,
int64_t* step_idx,
@@ -409,7 +410,9 @@ DLL_EXPORT int speculate_verify(Context* ctx,
const int max_candidate_len,
const int verify_window,
const bool prefill_one_step_stop,
const bool benchmark_mode);
const bool benchmark_mode,
const bool accept_all_drafts,
const bool use_target_sampling);
DLL_EXPORT int speculate_clear_accept_nums(Context* ctx,
int* accept_num,