apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-04 00:06:38 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[Code Simplification] Refactor Post-processing in VL Model Forward Method (#2937)

Browse Source 
* rm sth useless

* refactor model forward

* mv bool index to kernel
This commit is contained in:
Ryan
2025-08-01 17:28:07 +08:00
committed by GitHub
parent 3a4db15765
commit 94264bbf60
3 changed files with 25 additions and 38 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
custom_ops/gpu_ops/cpp_extensions.cc
View File

@@ -323,7 +323,7 @@ std::vector<paddle::Tensor> ExtractTextTokenOutput(
const paddle::Tensor &max_seq_len, const paddle::Tensor &max_seq_len_index,
const paddle::Tensor &mm_token_num_len,
const paddle::Tensor &seq_lens_this_time,
const paddle::Tensor &cu_seqlens_q, const paddle::Tensor &score_text);
const paddle::Tensor &cu_seqlens_q, const paddle::Tensor &hidden_states);
std::vector<paddle::Tensor> MoEDeepGEMMPermute(const paddle::Tensor &x,
const paddle::Tensor &topk_idx,