Skip to content

[BugFix][Qwen3-Omni]Fixed the issue of incorrect answers for single words. #2239

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
Gaohan123 merged 1 commit into from
Mar 27, 2026
Merged

[BugFix][Qwen3-Omni]Fixed the issue of incorrect answers for single words. #2239

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
3 changes: 1 addition & 2 deletions vllm_omni/benchmarks/patch/patch.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -376,9 +376,8 @@ async def benchmark(
limit_per_host=max_concurrency or 0,
ttl_dns_cache=300,
use_dns_cache=True,
keepalive_timeout=60,
enable_cleanup_closed=True,
force_close=False,
force_close=True,
Copy link
Copy Markdown

@chatgpt-codex-connector chatgpt-codex-connector Bot Mar 26, 2026

Choose a reason for hiding this comment

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

P2 Badge Keep benchmark HTTP connections reusable

In benchmark() this change sets aiohttp.TCPConnector(..., force_close=True), which closes every connection after each request. That disables connection reuse while this block is explicitly configured to benchmark request throughput/latency and even comments that connections are reused, so HTTPS/TCP handshakes get added to nearly every request and the measured model-serving metrics become systematically inflated/noisy under load.

Useful? React with 👍 / 👎.

Copy link
Copy Markdown
Collaborator

@gcanlin gcanlin Mar 26, 2026

Choose a reason for hiding this comment

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

And why do we need this change?

Copy link
Copy Markdown
Contributor Author

@amy-why-3459 amy-why-3459 Mar 27, 2026

Choose a reason for hiding this comment

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

To avoid server connect errors during high-concurrency benchmark runs

Comment on lines -379 to +380
Copy link
Copy Markdown
Contributor

@LJH-LBJ LJH-LBJ Mar 26, 2026

Choose a reason for hiding this comment

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

When a benchmark sends a large number of concurrent requests, the connection can be reused. Why delete it?

Copy link
Copy Markdown
Contributor Author

@amy-why-3459 amy-why-3459 Mar 27, 2026

Choose a reason for hiding this comment

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

Reusing connections may result in a server disconnect error.

ssl=ssl_setting,
)

Expand Down
4 changes: 1 addition & 3 deletions vllm_omni/model_executor/models/qwen3_omni/qwen3_omni.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1047,9 +1047,7 @@ def _get_talker_assistant_parts(
dim=0,
)
else:
trailing_text_hidden = torch.zeros(
tts_eos_embed.shape, device=tts_eos_embed.device, dtype=tts_eos_embed.dtype
)
trailing_text_hidden = tts_eos_embed

input_embeds = assistant_text_hidden + assistant_codec_hidden
input_ids = torch.full(
Expand Down
Loading