apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-23 00:17:25 +08:00
Code Issues Actions 19 Packages Projects Releases Wiki Activity

remove speculate_get_padding_offset op (#6308)

Browse Source
This commit is contained in:
周周周
2026-02-03 15:18:12 +08:00
committed by GitHub
parent 39dc4b0c2e
commit 8277b95fa6
7 changed files with 44 additions and 324 deletions
Download Patch File Download Diff File Expand all files Collapse all files
fastdeploy/model_executor/pre_and_post_process.py
+2 -6
View File
@@ -59,7 +59,6 @@ elif current_platform.is_maca():
save_output_topk,
set_stop_value_multi_ends,
speculate_get_output_padding_offset,
speculate_get_padding_offset,
speculate_get_seq_lens_output,
speculate_limit_thinking_content_length_v1,
speculate_limit_thinking_content_length_v2,
@@ -86,7 +85,6 @@ else:
save_output_topk,
set_stop_value_multi_ends,
speculate_get_output_padding_offset,
speculate_get_padding_offset,
speculate_get_seq_lens_output,
speculate_save_output,
speculate_save_output_topk,
@@ -226,7 +224,7 @@ def pre_process(
if specific_platform and not speculative_decoding:
# Note(ZKK): This case's code is very simple!
ids_remove_padding, batch_id_per_token, cu_seqlens_q, cu_seqlens_k = get_padding_offset(
input_ids, seq_lens_this_time, token_num_cpu
input_ids, seq_lens_this_time, None, None, token_num_cpu
)
return (
ids_remove_padding,
@@ -247,9 +245,7 @@ def pre_process(
batch_id_per_token,
cu_seqlens_q,
cu_seqlens_k,
) = speculate_get_padding_offset(
input_ids, draft_tokens, cum_offsets_now, seq_lens_this_time, seq_lens_encoder, token_num_cpu
)
) = get_padding_offset(input_ids, seq_lens_this_time, draft_tokens, seq_lens_encoder, token_num_cpu)
seq_lens_output = speculate_get_seq_lens_output(
seq_lens_this_time,
seq_lens_encoder,