Skip to content

Comments

[BugFix] token2wav code out of range#655

Merged
Gaohan123 merged 1 commit intovllm-project:mainfrom
Bounty-hunter:token_bug
Jan 6, 2026
Merged

[BugFix] token2wav code out of range#655
Gaohan123 merged 1 commit intovllm-project:mainfrom
Bounty-hunter:token_bug

Conversation

@Bounty-hunter
Copy link
Contributor

@Bounty-hunter Bounty-hunter commented Jan 5, 2026

PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS (AT THE BOTTOM) HAVE BEEN CONSIDERED.

Purpose

As show in #476

Talker with vocab_size 8448, while token2wav with 8193, and it will result out of range in token2wav.

Modify:
we suppress the token id [8194:8448] in talker stage.

Test Plan

Run the benchmark again!

Test Result

All Request Success

============ Serving Benchmark Result ============
Successful requests:                     100       
Maximum request concurrency:             1         
Request rate configured (RPS):           1.00      
Benchmark duration (s):                  862.52    
Total input tokens:                      16721     
Total text input tokens:                 1000      
Total generated tokens:                  6714      
Request throughput (req/s):              0.12      
Audio throughput (num/s):                0.12      
Output token throughput (tok/s):         7.78      
Peak output token throughput (tok/s):    1.00      
Peak concurrent requests:                2.00      
Total Token throughput (tok/s):          27.17     

Essential Elements of an Effective PR Description Checklist
  • The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
  • The test plan, such as providing test command.
  • The test results, such as pasting the results comparison before and after, or e2e results
  • (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
  • (Optional) Release notes update. If your change is user facing, please update the release notes draft.

BEFORE SUBMITTING, PLEASE READ https://github.com/vllm-project/vllm-omni/blob/main/CONTRIBUTING.md (anything written below this line will be removed by GitHub Actions)

Signed-off-by: dengyunyang <[email protected]>
@hsliuustc0106
Copy link
Collaborator

@tzhouam @Gaohan123 is the range diffference a known problem?

@hsliuustc0106
Copy link
Collaborator

for qwen3-omni, does this happen?

@Bounty-hunter
Copy link
Contributor Author

for qwen3-omni, does this happen?

Qwen3-omni may not have this problem:
talker_config.code_predictor_config.vocab_size = 2048
talker_config.code_predictor_config.num_code_groups = 16

code2wav_config.codebook_size = 2048
code2wav_config.num_quantizers = 16

It same.

@Gaohan123 Gaohan123 added the ready label to trigger buildkite CI label Jan 6, 2026
Copy link
Collaborator

@Gaohan123 Gaohan123 left a comment

Choose a reason for hiding this comment

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

LGTM. Thanks!

@Gaohan123 Gaohan123 merged commit b600300 into vllm-project:main Jan 6, 2026
7 checks passed
Shirley125 pushed a commit to Shirley125/vllm-omni that referenced this pull request Jan 9, 2026
princepride pushed a commit to princepride/vllm-omni that referenced this pull request Jan 10, 2026
sniper35 pushed a commit to sniper35/vllm-omni that referenced this pull request Jan 10, 2026
ZJY0516 pushed a commit to LawJarp-A/vllm-omni that referenced this pull request Jan 10, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ready label to trigger buildkite CI

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants