Search results

Qwen-ASR 0.6B (Streaming) Hotword Feature Causes Repetitive Output of Hotword List for Similar Speech Input

Status: Open.

#140 In QwenLM/Qwen3-ASR;

· linghugoogle opened on Mar 30, 2026

Support transformers >= 5.3 , vllm >= 0.17 and some PR.

Status: Open.

#138 In QwenLM/Qwen3-ASR;

· One-sixth opened on Mar 26, 2026

the evaluation scripts and dataset?

Status: Open.

#134 In QwenLM/Qwen3-ASR;

· vincentLehighYe opened on Mar 23, 2026

set min_length=1, remove audio less than 150ms. following errors stil stop training

Status: Open.

#130 In QwenLM/Qwen3-ASR;

· housebaby opened on Mar 19, 2026

Endless repetition in output

Status: Open.

#129 In QwenLM/Qwen3-ASR;

· andypotato opened on Mar 19, 2026

Evaluation procedure？

Status: Open.

#128 In QwenLM/Qwen3-ASR;

· vincentLehighYe opened on Mar 18, 2026

Using VLLM for backend inference in qwen3-asr may cause it to freeze (GPU memory is not released and utilization is 0%).

Status: Open.

#126 In QwenLM/Qwen3-ASR;

· Lee-xeo opened on Mar 16, 2026

不再适配 vllm 0.16.0+

Status: Open.

#119 In QwenLM/Qwen3-ASR;

· Lee-xeo opened on Mar 9, 2026

slow inference after finetuning

Status: Open.

#118 In QwenLM/Qwen3-ASR;

· binnybn98 opened on Mar 9, 2026

Why can't vllm be found after installation?

Status: Open.

#110 In QwenLM/Qwen3-ASR;

· SunYong0821 opened on Mar 5, 2026

ASRTranscription output contains context.

Status: Open.

#106 In QwenLM/Qwen3-ASR;

· huanchen-1 opened on Mar 4, 2026

Eager nagisa import adds ~3.5s to startup even when ForcedAligner is unused

Status: Open.

#105 In QwenLM/Qwen3-ASR;

· Add1ct1ve opened on Mar 3, 2026