Search results

[HOW TO Optimize]: the delay of the first frame increases too quickly for Qwen3-TTS with Concurrently

bugSomething isn't workingSomething isn't working

Status: Open.

#3136 In vllm-project/vllm-omni;

· liujian19911023 opened on Apr 25, 2026

[Bug][Qwen3-Omni]: Launching individual stages (stage-based CLI) resulted in an OutOfMemoryError (OOM) error.

bugSomething isn't workingSomething isn't working

Status: Open.

#3135 In vllm-project/vllm-omni;

· amy-why-3459 opened on Apr 25, 2026

[Chore]: TTS/omni: use vllm.vllm_flash_attn instead of flash-attn package for 5 models

Status: Open.

#3131 In vllm-project/vllm-omni;

· zhangj1an opened on Apr 25, 2026

[Bug]: Qwen3-TTS, voice clone, Code2Wav input_ids length 1 not divisible by num_quantizers 16

bugSomething isn't workingSomething isn't working

Status: Open.

#3124 In vllm-project/vllm-omni;

· liujian19911023 opened on Apr 25, 2026

[CI Failure]: Diffusion X2I(&A&T) · Doc Test, test_text_to_image.py, pydantic_core._pydantic_core.ValidationError: 1 validation error for DiffusionParallelConfig

bugSomething isn't workingSomething isn't working

ci-failureCI failure issues, expected to be solved asapCI failure issues, expected to be solved asap

Status: Open.

#3123 In vllm-project/vllm-omni;

· yenuo26 opened on Apr 25, 2026

[CI Failure]: Simple Unit Test, tests/benchmarks/test_seed_tts_dataset_variants.py, ValueError: max() iterable argument is empty

bugSomething isn't workingSomething isn't working

ci-failureCI failure issues, expected to be solved asapCI failure issues, expected to be solved asap

Status: Open.

#3122 In vllm-project/vllm-omni;

· yenuo26 opened on Apr 25, 2026

[Bug] CUDNN_ATTN crashes on LTX-2.0 audio attention under torch.compile (symbolic head_dim)

help wantedExtra attention is neededExtra attention is needed

Status: Open.

#3121 In vllm-project/vllm-omni;

· lishunyang12 opened on Apr 25, 2026

[Bug]: diffusers backend running Wan model, output different from raw diffusers library

bugSomething isn't workingSomething isn't working

Status: Open.

#3116 In vllm-project/vllm-omni;

· fhfuih opened on Apr 24, 2026

[RFC]: LVSA — Long Video Sparse Attention for vllm-omni

Status: Open.

#3115 In vllm-project/vllm-omni;

· gglorian opened on Apr 24, 2026

[Bug]: Qwen3-TTS, when set load_format:dummy in qwen3-tts.yaml, ValueError: This checkpoint does not provide speaker_encoder weights; cannot compute ref_spk_embedding from ref_audio.

bugSomething isn't workingSomething isn't working

Status: Open.

#3112 In vllm-project/vllm-omni;

· yenuo26 opened on Apr 24, 2026

[Performance]: Qwen3-Omni text-only output inference speed is slower than vllm

Status: Open.

#3093 In vllm-project/vllm-omni;

· zsl24 opened on Apr 24, 2026

OmniGenerationScheduler closes request before stateful streaming vocoder's finalize chunk is emitted (async_chunk path, affects CosyVoice3 and VoxCPM)

Status: Open.

#3090 In vllm-project/vllm-omni;

· npuichigo opened on Apr 24, 2026