Skip to content

[BugFix] Fix max-num-seqs affecting performance on small models. - #2998

Closed
lizexu123 wants to merge 1 commit into
PaddlePaddle:developfrom
lizexu123:work_process-1
Closed

[BugFix] Fix max-num-seqs affecting performance on small models.#2998
lizexu123 wants to merge 1 commit into
PaddlePaddle:developfrom
lizexu123:work_process-1

Conversation

@lizexu123

Copy link
Copy Markdown
Collaborator

修复推理时候max-num-seqs,而不是真实batch_size

@paddle-bot

paddle-bot Bot commented Jul 24, 2025

Copy link
Copy Markdown

Thanks for your contribution!

@@ -255,10 +255,9 @@ def event_loop_ep(self) -> None:
)
# Process prefill inputs
self.worker.preprocess_new_task(req_dicts)

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

我理解这个里面把seq_lens_this_time做一次slice应该就可以。

@paddle-bot paddle-bot Bot closed this Jul 28, 2026
@paddle-bot

paddle-bot Bot commented Jul 28, 2026

Copy link
Copy Markdown

Automatically closed by Paddle-bot.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants