Skip to content

Eval bug: Video transcribe crash, Qwen3-Omni-30B-A3B #24322

Description

@thomas-0816

Name and Version

./llama-b9568$ ./llama-server -hf ggml-org/Qwen3-Omni-30B-A3B-Instruct-GGUF:Q4_K_M --threads -1 --parallel 1 --ctx-size 16384 --temp 0.6 --top-p 0.95 --top-k 20 --predict 4096 --direct-io --log-verbosity 9999

Input mp4 file 11 minutes long over the web ui.

log output of first error: corrupted double-linked list

0.22.805.922 D === Differential analysis complete ===
0.22.806.329 D common_chat_templates_apply_jinja: generated parser:
Sequence(Literal(<|im_start|>assistant
), Space, Repetition(Sequence(Epsilon, Literal(<think>), Repetition(Literal(
), 0, 1), Tag(reasoning, Until(</think>)), Epsilon, Repetition(Literal(
), 0, 1), Literal(</think>), Repetition(Literal(
), 0, 1), Repetition(Literal(
), 0, 1)), 0, 1), Tag(content, Until()), End)

parser generation prompt: <|im_start|>assistant
<think>

</think>


0.22.829.787 D probe: launching:0.22.829.792 D  ffprobe0.22.829.792 D  -v0.22.829.793 D  quiet0.22.829.793 D  -show_entries0.22.829.794 D  stream=width,height,r_frame_rate,nb_frames,duration0.22.829.794 D  -select_streams0.22.829.794 D  v:00.22.829.794 D  -of0.22.829.794 D  default=noprint_wrappers=10.22.829.795 D  pipe:00.22.829.795 D 
0.22.832.198 D feed_stdin: feeding 22449655 bytes to stdin
0.22.885.258 D feed_stdin: wrote 22449655 bytes, closing stdin
corrupted double-linked list
Aborted

Operating systems

Linux

GGML backends

Vulkan

Hardware

ryzen 7840U

Models

No response

Problem description & steps to reproduce

add a mp4 file for transcription

First Bad Commit

No response

Relevant log output

Logs

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions