Skip to content

[Feature] Support set PREEMPTED_TOKEN_ID in GET_SAVE_OUTPUT_V1 #7159

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Jiang-Jia-Jun merged 3 commits into from
Apr 8, 2026
+11 −1
Merged

[Feature] Support set PREEMPTED_TOKEN_ID in GET_SAVE_OUTPUT_V1 #7159

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
8dd4eec
[Feature] Support set PREEMPTED_TOKEN_ID in GET_SAVE_OUTPUT_V1
rainyfly Apr 2, 2026
d940eb6
[Feature] Support set PREEMPTED_TOKEN_ID in GET_SAVE_OUTPUT_V1
rainyfly Apr 2, 2026
556c78a
fix
rainyfly Apr 3, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 11 additions & 1 deletion fastdeploy/worker/gpu_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,7 +27,7 @@
from paddle import nn
from paddleformers.utils.log import logger

from fastdeploy.config import FDConfig
from fastdeploy.config import PREEMPTED_TOKEN_ID, FDConfig
from fastdeploy.engine.pooling_params import PoolingParams
from fastdeploy.engine.request import ImagePosition, Request, RequestType
from fastdeploy.model_executor.graph_optimization.utils import (
Expand Down Expand Up @@ -2404,6 +2404,16 @@ def _postprocess(

# 5.1. Async cpy
post_process_event = paddle.device.cuda.create_event()
if envs.FD_USE_GET_SAVE_OUTPUT_V1:
# If one query is preempted, there is no sampled token for it, we use token_id PREEMPTED_TOKEN_ID to signal server, abort is finished.
paddle.assign(
paddle.where(
self.share_inputs["last_preempted_idx"][: sampler_output.sampled_token_ids.shape[0]] == 1,
PREEMPTED_TOKEN_ID,
sampler_output.sampled_token_ids,
),
sampler_output.sampled_token_ids,
)
# if not self.speculative_decoding:
self.share_inputs["sampled_token_ids"].copy_(sampler_output.sampled_token_ids, False)
if self.speculative_decoding:
Expand Down
Loading