apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-09-27 12:52:29 +08:00
Code Issues Actions 5 Packages Projects Releases Wiki Activity

[Inference, rename] remove padding_offsets from atten use batch_id_per_token (#2880)

Browse Source 
* remove padding_offsets from atten
This commit is contained in:
周周周
2025-07-17 18:41:31 +08:00
committed by GitHub
parent d49f8fb30a
commit ddb10ac509
50 changed files with 311 additions and 288 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
fastdeploy/model_executor/pre_and_post_process.py
View File

@@ -72,7 +72,7 @@ def pre_process(
Return:
ids_remove_padding:
cum_offsets:
padding_offset:
batch_id_per_token:
cu_seqlens_q:
cu_seqlens_k:
"""
@@ -85,7 +85,7 @@ def pre_process(
(
ids_remove_padding,
cum_offsets,
padding_offset,
batch_id_per_token,
cu_seqlens_q,
cu_seqlens_k,
) = speculate_get_padding_offset(
@@ -115,12 +115,12 @@ def pre_process(
(
ids_remove_padding,
cum_offsets,
padding_offset,
batch_id_per_token,
cu_seqlens_q,
cu_seqlens_k,
) = get_padding_offset(input_ids, cum_offsets_now, token_num,
seq_lens_this_time)
return (ids_remove_padding, cum_offsets, padding_offset, cu_seqlens_q,
return (ids_remove_padding, cum_offsets, batch_id_per_token, cu_seqlens_q,
cu_seqlens_k, output_cum_offsets, output_padding_offset)